Paper deep dive
Planning, Scheduling, and Behavior in EV Charging Systems: A Critical Survey and Trilemma Framework
Peiyan Xiao, Yuheng Li, Ayan Mukhopadhyay, Sai Krishna Ghanta, Sabur Baidya, Yanhai Xiong
Intelligence
Status: succeeded | Model: Gemma-4-26B-A4B | Prompt: intel-v1 | Confidence: 94%
Last extracted: 7/8/2026, 6:32:22 PM
Summary
This survey introduces a three-layer Planning-Scheduling-Behavior (PSB) framework to organize electric vehicle (EV) charging research. It identifies the 'PSB trilemma,' a fidelity-tractability tradeoff where computational hardness at each layer and their pairwise couplings (modeled as bilevel programs, MPECs, and EPECs) forces researchers to simplify at least one layer. The paper maps actors, objectives, and interface variables across layers, demonstrating that simplifying Planning, Scheduling, or Behavior sacrifices investment feedback, grid/emissions dynamics, or user equity/heterogeneity, respectively.
Entities (16)
Relation Signals (21)
PSB Framework → comprises → Scheduling Layer
confidence 98% · This survey develops a three-layer Planning-Scheduling-Behavior (PSB) framework to organize EV charging research
PSB Framework → comprises → Planning Layer
confidence 98% · This survey develops a three-layer Planning-Scheduling-Behavior (PSB) framework to organize EV charging research
PSB Framework → comprises → Behavior Layer
confidence 98% · This survey develops a three-layer Planning-Scheduling-Behavior (PSB) framework to organize EV charging research
PSB Trilemma → isa → Fidelity-tractability tradeoff
confidence 97% · We further identify a fidelity-tractability tradeoff, termed the PSB trilemma
Scheduling Layer → affects → Behavior Layer
confidence 95% · scheduling affects behavior through prices and waiting times
Planning Layer → constrains → Scheduling Layer
confidence 95% · Planning constrains scheduling through infrastructure capacity
Bilevel Program → models → Planning-Scheduling coupling
confidence 95% · When the two layers are coupled, the natural formulation is a bilevel program in which the planner optimizes infrastructure in the upper level, and the scheduler optimizes operations in the lower level
Behavior Layer → →
Cypher Suggestions (0)
No Cypher suggestions yet.
Abstract
Abstract:The rapid growth of electric vehicles is shifting the main constraint on transport electrification from vehicle adoption to the deployment and operation of charging infrastructure. Charging-network design requires decisions across three interdependent layers: Planning, which determines where and how much infrastructure to build; Scheduling, which governs charging dispatch, pricing, and grid interaction; and Behavior, which captures how users choose stations, charging times, and charging durations. Existing studies have advanced each layer substantially, but the literature remains fragmented, and cross-layer interactions are often treated through simplifying assumptions. This survey develops a three-layer Planning-Scheduling-Behavior (PSB) framework to organize EV charging research according to decision horizon, actor objective, and coupling structure. We further identify a fidelity-tractability tradeoff, termed the PSB trilemma: each layer is computationally difficult in isolation, and realistic integration across layers generally requires reducing the fidelity of at least one layer. Reviewing the three pairwise-coupling literatures - Planning-Scheduling, Scheduling-Behavior, and Planning-Behavior - we show that the omitted third layer is typically fixed exogenously or represented by a static aggregate surrogate. These simplifications enable tractability but impose distinct costs: they can obscure long-term investment feedback, temporal grid and emissions dynamics, or heterogeneous user response and equity outcomes. Building on this diagnosis, we identify open challenges in emerging charging technologies, behavioral incentives, equity metrics, and city-scale learning-based methods that balance fidelity, interpretability, and policy relevance.
Tags
Links
- Source: https://arxiv.org/abs/2605.21665v1
- Canonical: https://arxiv.org/abs/2605.21665v1
Trouble viewing inline? Open PDF directly →
Full Text
144,848 characters extracted from source content.
Expand or collapse full text
Planning, Scheduling, and Behavior in EV Charging Systems: A Critical Survey and Trilemma Framework Peiyan Xiao a , Yuheng Li a , Ayan Mukhopadhyay a , Sai Krishna Ghanta b , Sabur Baidya c , Yanhai Xiong a,∗ a The College of William & Mary, Williamsburg, VA, USA b University of Georgia, Athens, GA, USA c University of Louisville, Louisville, KY, USA Abstract The rapid growth of electric vehicles is shifting the main constraint on trans- port electrification from vehicle adoption to the deployment and operation of charging infrastructure. Charging-network design requires decisions across three interdependent layers: Planning, which determines where and how much infrastructure to build; Scheduling, which governs charging dispatch, pricing, and grid interaction; and Behavior, which captures how users choose stations, charging times, and charging durations. Existing studies have advanced each layer substantially, but the literature remains fragmented, and cross-layer interactions are often treated through simplifying assump- tions. This survey develops a three-layer Planning–Scheduling–Behavior (PSB) framework to organize EV charging research according to decision horizon, actor objective, and coupling structure. We further identify a fidelity–tractability tradeoff, termed the PSB trilemma: each layer is com- putationally difficult in isolation, and realistic integration across layers gen- erally requires reducing the fidelity of at least one layer. Reviewing the three pairwise-coupling literatures—Planning–Scheduling, Scheduling–Behavior, and Planning–Behavior—we show that the omitted third layer is typically fixed exogenously or represented by a static aggregate surrogate. These sim- ∗ Corresponding author Email addresses: pxiao@wm.edu (Peiyan Xiao), yli95@wm.edu (Yuheng Li), amukhopadhyay@wm.edu (Ayan Mukhopadhyay), sai.krishna@uga.edu (Sai Krishna Ghanta), sabur.baidya@louisville.edu (Sabur Baidya), yanhaixiong7@gmail.com (Yanhai Xiong) arXiv:2605.21665v1 [cs.MA] 20 May 2026 plifications enable tractability but impose distinct costs: they can obscure long-term investment feedback, temporal grid and emissions dynamics, or heterogeneous user response and equity outcomes. Building on this diagnosis, we identify open challenges in emerging charging technologies, behavioral in- centives, equity metrics, and city-scale learning-based methods that balance fidelity, interpretability, and policy relevance. Word count: 9,881 Keywords: Electric vehicle charging; Charging infrastructure planning; Smart charging; User charging behavior; Grid integration; Transport electrification; Sustainable transportation 1. Introduction Achieving net-zero emissions requires deep decarbonization of transport, which produces about one-quarter of global energy-related CO 2 emissions; road vehicles account for nearly three-quarters of that share [1]. Because transport emissions are dispersed and shaped by long vehicle lifetimes and heterogeneous travel patterns, passenger and light-duty vehicle electrification is viewed as a key near-term mitigation pathway [2]. With more than sixty countries adopting combustion-engine phase-out targets or EV mandates, global electric car stock reached nearly 58 million in 2024 and is projected to exceed 250 million by 2030 [3]. As deployment accelerates, the main constraint is shifting from EV availability to charging infrastructure. Charging-network deployment requires integrated decisions on long-term siting and capacity, short-term pricing and power dispatch, and user re- sponses to price, wait time, detour, and availability. These decisions involve different actors and time scales but are tightly coupled: infrastructure con- 2 strains operations, operations shape charging choices, and aggregate demand determines utilization and investment viability. Yet the literature has largely developed each layer separately. Planning studies have advanced from fixed-demand coverage models [4, 5, 6] to queueing, grid-coupled sizing [7, 8], and multi-period expansion under uncertainty [9, 10]. Scheduling has moved from deterministic station dis- patch [11] to V2G, renewables, and multi-vehicle coordination [12], then to predictive and learning-based control under stochastic arrivals and prices [13, 14]. Behavior models have progressed from exogenous demand [15, 16] to choice, equilibrium, and heterogeneous adaptive populations [17, 18, 19, 20]. Thus, each layer has matured in isolation, motivating cross-layer integration. Studies that jointly address two or more stages have begun to appear. Planning models incorporate operational dispatch into the investment deci- sion [21, 22]; scheduling frameworks anticipate user response to pricing sig- nals [23, 24, 25]; and siting models account for user choice behavior through equilibrium-based formulations [26]. However, these cross-cutting studies remain comparatively limited, and most achieve tractability only by simpli- fying one stage to a fixed assumption. Existing surveys mirror this pattern: some integrate planning with power-distribution coupling but do not en- gage scheduling or user response [27]; the most recent planning review does not address operational and behavioral feedback [7]; others focus on oper- ations [12, 28], deterministic scheduling [29], reinforcement-learning meth- ods [14], or the service and policy landscape [30]. 3 This pattern is not accidental. Each of the three decision layers is individ- ually a hard computational problem: facility location with integer variables is NP-hard, scheduling under stochastic arrivals and grid constraints leads to MDPs that suffer the curse of dimensionality, and computing user equilib- ria scales poorly with the number of user classes and alternatives. Coupling any two layers compounds the difficulty, producing problem classes (bilevel programs, stochastic games, mathematical programs with equilibrium con- straints) for which no general-purpose efficient algorithms exist. The result is a tradeoff between modeling fidelity and computational tractability: practi- cal joint formulations at a realistic scale generally simplify at least one layer. We call this the Planning-Scheduling-Behavior (PSB) trilemma. Existing reviews typically organize EV charging research by application or method; this survey instead uses cross-layer decision coupling. Its cen- tral claim is that fragmentation reflects not only a literature gap but also a fidelity–tractability tradeoff. By unifying Planning, Scheduling, and Be- havior under the PSB framework, the survey makes each stream’s assump- tions, simplifications, and blind spots explicit, linking modeling choices to the practical needs of infrastructure deployment. Figure 1.1 illustrates the three components and their information flows. We make three contributions: 1. A three-layer PSB framework with an actor taxonomy (Sec- tion 2). We define Planning, Scheduling, and Behavior by decision horizon, specify their coupling edges, and identify the actors and ob- jectives associated with each layer. This allows studies to be compared 4 by the layers they model, the couplings they retain, and the simplifica- tions they impose. 2. The PSB trilemma (Section 2.4). We identify a fidelity–tractability tradeoff caused by the computational difficulty of each layer and the added complexity of cross-layer coupling. The trilemma explains why realistic joint formulations usually simplify at least one layer, a pattern reflected in pairwise studies that often fix or aggregate the omitted third layer. 3. Context-dependent simplification guidance (Sections 6 and 7). We show that simplifying different layers sacrifices different objectives: Planning simplification loses investment feedback, Scheduling simpli- fication erases temporal grid and emissions dynamics, and Behavior simplification suppresses heterogeneity and equity. Thus, the choice of simplification should depend on the decision-maker’s objective, with the PSB framework making the cost of each choice explicit. 1.1. Review scope and methodology This survey is a structured critical review of EV charging research, or- ganized by decision layer, actor objective, and coupling assumption across Planning, Scheduling, and Behavior. We searched Scopus, Web of Science, ScienceDirect, IEEE Xplore, TRID, and Google Scholar using terms on in- frastructure planning, charging scheduling, user behavior, and cross-layer in- terfaces, including siting, sizing, smart charging, dynamic pricing, vehicle-to- 5 grid, station choice, user equilibrium, demand response, bilevel optimization, Stackelberg models, transportation-power networks, and pricing–behavior coupling. Studies were screened through title/abstract review, full-text review, and PSB-based coding. We included papers modeling at least one PSB layer or explicitly coupling multiple layers, and coded them by layer, actor class, decision horizon, objective, method, uncertainty, grid and behavior repre- sentation, and interface variables. For pairwise studies, we also recorded which layers were endogenous, how the omitted layer was simplified, and what fidelity was lost. We excluded work centered on battery chemistry, charger hardware, converter control, or general EV adoption without direct charging-system relevance. Section 2 introduces the PSB framework and trilemma. Sections 3–5 re- view the three layers, Section 6 examines pairwise coupling and omitted-layer simplifications, Section 7 discusses future directions, and Section 8 concludes. 6 PlanningLayer SchedulingLayer BehaviorLayer ChargingInfrastructurePlanning Infrastructure capacity constraint Utilization feedback Prices & wait times Arrival rates & user types Siting （MonthstoYears） （HourstoDays） （Minutes） Operational Scheduling & Dispatch User Charging Choice & Response SizingGridCoupling Multi-Period Expansion Multi-EV Dis patch V2G TOUPricing ChargingTiming ChargingLocation ChargingDuration Real-Time Control Figure 1: Planning–Scheduling–Behavior framework for EV charging systems. The ver- tical axis indicates the decision horizon. Solid blocks denote the three decision layers; arrows denote coupling variables exchanged across layers. Planning constrains scheduling through infrastructure capacity; scheduling affects behavior through prices and waiting times; behavior feeds back through arrival rates, user types, utilization, and long-run de- mand. 2. A Three-Layer Framework for EV Charging Systems: Structure, Complexity, and Coupling This section develops the Planning–Scheduling–Behavior (PSB) frame- work. Section 2.1 defines the three layers and their temporal scales. Sec- tion 2.2 introduces the actor taxonomy. Section 2.3 characterizes the cross- layer coupling via interface variables. Section 2.4 states the PSB trilemma. 7 2.1. Overview of the Three-Layer Decomposition We decompose the EV charging problem into three layers, each defined by a characteristic decision horizon and decision-maker class. The Planning layer (P) concerns where to build stations, at what capacity, and how to connect them to the distribution grid. These choices are capital-intensive and difficult to reverse once construction is complete, so they fix the physical boundaries within which scheduling and user decisions must operate. Infrastructure operators, utilities, and public agencies make these decisions over a horizon of months to years. The Scheduling layer (S) concerns how to allocate charging power and set prices and incentives in response to evolving vehicle arrivals and grid conditions, including day-ahead and real-time pricing. Station opera- tors manage individual sites and decide how to split available power among connected vehicles; aggregators coordinate multiple stations or flexible loads and trade energy or ancillary services with the grid on their behalf. Both make scheduling decisions over a horizon of hours to days. The Behavior layer (B) concerns how users choose where and when to charge in response to real-time signals such as prices, wait times, and station availability. Individual drivers, fleet operators, and ride-hailing drivers make these decisions over a horizon of minutes. 8 2.2. Actors and Objectives The PSB decomposition organizes the literature along the temporal- computational dimension, but a second dimension is equally important: who is making the decision at each layer. Each layer is populated by several distinct actor classes. We define an actor by the triple of decision rights, objective function, and information set. This triple, rather than the layer alone, determines which mathematical formulation to instantiate. For ex- ample, in the Planning layer, a private charge-point operator maximizes long-run profit [31], while a public planner maximizes spatial coverage under a budget constraint [6]. Both use facility-location models, but their different objectives yield different optimal networks for the same candidate sites. Recognizing the implicit actor behind each formulation is therefore essential for comparing results across studies. Table 1 lists nine actor classes across the three layers. A key pattern is that actors within the same layer can have conflicting objectives, which means that results from studies assuming different actors are not directly comparable. The implicit actor of each model reviewed in Sections 3–5 is flagged at the relevant subsubsection, so that the per-layer reviews can focus on model families without re-introducing the actor taxonomy. 9 Table 1: Actor taxonomy across the three PSB layers, with each actor defined by its decisions and objective. Layer ActorDecisionsObjective P Charge point operator (CPO) Siting, sizing, charger mix Maximize net present value or long-run profit Public plannerSiting, coverage targets Maximize coverage, equity, or refuelable flow Utility / distribution system operator (DSO) Siting jointly with feeder upgrades Minimize grid reinforcement cost S Charging station operator / aggregator Power allocation, charging price and rate setting Maximize revenue subject to grid and quality-of-service constraints Distribution system operator Managing distribution network capacity, limiting charging load when needed Minimize feeder overloading and voltage violations Fleet managerRouting, depot dwell, charging schedule (including individual driver assignments) Minimize fleet operating cost under service level agreements B Private driverStation choice, timing, en-route stop, price negotiation Maximize random utility over cost, convenience, range Ride-hailing / transportation company Charging timing, idle relocation Maintain service availability while minimizing charging cost 10 2.3. Cross-Layer Coupling and Computational Structure This subsection first summarizes the mathematical structure each layer takes in isolation, then examines the three pairwise couplings and the com- plexity each introduces. Table 2 summarizes the six coupling edges. Standalone Mathematical Structure In isolation, each layer admits a canonical formulation. Planning is a facil- ity location problem over a finite candidate set, producing mixed-integer pro- grams of increasing complexity as capacity, grid, and uncertainty constraints are added (Section 3) [4]. Scheduling is a sequential resource-allocation prob- lem: in the deterministic case, the operator minimizes cost subject to energy delivery and grid limits; under stochastic arrivals and price uncertainty, this extends to a Markov decision process (MDP) (Section 4) [13]. Behavior, once congestion feedback is introduced, becomes a fixed-point problem: each user optimizes individually given the congestion and prices implied by the current population distribution, and an equilibrium is reached when the re- sulting distribution reproduces itself (Section 5). Planning × Scheduling The P→S direction transmits hard capacity constraints: a station’s charger count and power rating cap the dispatcher’s feasible allocation. The S→P return transmits observed utilization and congestion levels, informing expansion decisions [32]. When the two layers are coupled, the natural formulation is a bilevel program in which the planner optimizes infrastruc- 11 ture in the upper level, and the scheduler optimizes operations in the lower level [33]. Replacing the lower-level problem with its optimality conditions yields a mathematical program with equilibrium constraints (MPEC). Even with a deterministic lower level, bilevel programs with integer upper-level variables are Σ p 2 -hard in general [33]. When the scheduling response is mod- eled as an MDP rather than a convex program, its value function lacks a closed-form representation, preventing this reformulation. Most existing formulations, therefore, replace the scheduling response with a surrogate: a fixed utilization factor, a representative-day deterministic dispatch [22], or a convex lower level amenable to optimality-condition reformulation [21]. Scheduling × Behavior The S→B direction transmits real-time price signals and wait times that shape user charging decisions. The B→S return delivers the realized arrival rates and user-type distributions that constitute the scheduling load. When the two layers are coupled, this feedback loop transforms the single-agent scheduling problem into a multi-agent stochastic game. The resulting prob- lem belongs to the class of equilibrium problems with equilibrium constraints (EPEC), whose feasible regions are generally nonconvex and for which con- straint qualifications commonly fail [33]. The literature responds by splitting into two camps: dynamic-pricing papers absorb user response into a paramet- ric demand function [34], while equilibrium papers freeze operator dispatch at its mean values [17]. Each endogenizes one direction of the loop while 12 reducing the other to a static aggregate. Planning × Behavior The P→B direction determines which stations are accessible to users; the B→P return aggregates individual choices into long-run spatial demand pat- terns that shape investment returns [17]. When the two layers are coupled, embedding the behavioral fixed point inside the planning bilevel program produces an MPEC whose feasible region is nonconvex due to the comple- mentarity conditions that define the equilibrium [35]. Standard nonlinear solvers are not guaranteed to find even a local optimum, and the difficulty scales with the number of user classes and station alternatives. As a result, behavior is typically represented as a fixed or parametric demand surface that does not respond to deployment decisions over calendar time [36]. Taken together, the three pairwise couplings form a closed feedback sys- tem: every input that one layer takes as given is, in principle, determined endogenously by one of the other two. Table 2: Interface variables across the six coupling edges of the PSB framework. Edge Physical Meaning P → S Capacity constrains feasible scheduling domain S → P Utilization feeds back into expansion decisions P → B Siting determines reachable stations B → P Equilibrium behavior sets long-run demand S → B Price and wait time drive charging decisions B → S User types and arrival rates shape scheduling load 13 2.4. The PSB Trilemma: A Fidelity-Tractability Tradeoff Section 2.3 showed that each pairwise coupling produces a computation- ally hard problem class. This subsection explains why these difficulties, taken together, constitute a systematic barrier to joint modeling at realistic fidelity. What is the trilemma?. The difficulty is not that joint modeling across three layers is impossible. A small instance with five candidate sites, determin- istic scheduling, and homogeneous users can be formulated and solved as a single optimization, but such instances bear little resemblance to real-world deployments involving hundreds of candidate sites, stochastic conditions, and heterogeneous user populations. The difficulty is that, at the scale and fi- delity each layer requires for practically relevant applications, the problem becomes computationally intractable. The trilemma is therefore a fidelity- tractability tradeoff: practical joint formulations at a realistic scale typically reduce the modeling fidelity of at least one layer below what that layer would require if studied in isolation. Why does the trilemma arise?. The tradeoff has two sources. First, each layer in isolation is already a hard problem: facility location with integer variables is NP-hard, MDPs with large state-action spaces suffer the curse of dimensionality, and computing population equilibria scales poorly with the number of user classes and alternatives. Second, coupling any two lay- ers compounds the difficulty: as shown in Section 2.3, each pairwise coupling produces a problem class (Σ p 2 -hard bilevel programs, nonconvex EPECs, non- 14 convex MPECs) for which no general-purpose efficient algorithms exist [33]. These complexity results are not artifacts of a particular algorithm but es- tablished conclusions in the optimization and game theory literatures. How hard is it to bridge the gap?. The trilemma is not an impossibility but a tradeoff. The literature has developed approximation strategies that bridge specific pairs, but a gap remains between what these methods can handle and what real-world deployment requires. Stochastic bilevel formulations embed scheduling into the planning decision [37], but a realistic problem involves hundreds of candidate sites, multi-period expansion, and joint uncertainty in demand, prices, and renewable generation; current formulations achieve tractability by restricting scenarios to a handful, capturing only a fraction of real-world variability. Stackelberg game models capture leader-follower interaction between operators and users [38, 23], but require the follower’s response to take an analytically tractable form such as a logit choice model; as Sections 5.3 and 5.4 show, real users satisfice, form habits, and learn from experience, none of which produce the closed-form best-response functions these formulations require. Mean-field approximations compress a large user population into a single representative agent [39], but a real charging network serves commuters, ride-hailing drivers, and fleet vehicles with fundamentally different price sensitivities, time constraints, and spatial patterns; replacing them with a population average eliminates the distributional information needed to assess equity or predict group-specific responses. This gap between method capacity and real-world complexity is reflected 15 in the literature as a whole. The vast majority of studies model a single layer; those that jointly address two layers remain a small fraction [7, 12, 40]; and formulations that integrate all three layers at realistic fidelity are essentially absent [41, 42, 10]. Prior surveys have noted this fragmentation [27, 7], but they describe the symptom without diagnosing the structural cause. The trilemma provides this missing diagnosis. How should the trilemma be managed?. Different institutions control dif- ferent layers, and some actors partially bridge specific pairs: a vertically integrated utility internalizes P↔S [43], and a fleet manager internalizes S↔B [44]. No single actor, however, spans all three layers. Joint three- layer optimization is therefore not a natural decision problem for any ex- isting institutional role. Joint three-layer analysis, by contrast, is essential: the outcomes that matter most to regulators and planners, such as emis- sions, equity, and grid reliability, are emergent properties of the three-layer interaction that no single layer can predict alone. The trilemma does not excuse analysts from considering all three layers; rather, it tells them that the choice of which layer to simplify should be driven by the objective they pursue. Three decision contexts illustrate why this choice matters. When Planning is simplified, the analyst gains a high-fidelity S×B feed- back loop for tariff design, but loses the ability to feed operational insights back into infrastructure decisions; if stations are in the wrong locations, no pricing strategy can compensate. When Scheduling is simplified, the analyst can jointly optimize siting and user equilibrium, but collapsing operations 16 into a utilization factor erases the temporal structure that determines emis- sion outcomes; a planner evaluating siting against an annual average grid may build stations that concentrate charging at peak hours, increasing emis- sions rather than reducing them. When Behavior is simplified, the analyst can jointly optimize siting and dispatch under grid constraints, but the model assumes users will charge as predicted regardless of network changes; the in- frastructure that looks optimal on paper may perform poorly once real users interact with it. Each simplification is appropriate for some objectives and damaging for others. The trilemma thus serves not as a barrier but as a guide: it tells analysts which simplification to choose given the objective at hand, and the PSB framework makes the cost of each choice explicit. Section 6 examines how each two-layer subliterature navigates this trade- off in practice. 3. The Planning Layer: From Basic Siting to Real-World Infras- tructure Design This section reviews the Planning literature through a complexity pro- gression: starting from the simplest siting formulation, each subsection adds one layer of real-world constraints (capacity limits, grid coupling, uncertainty, and multi-period dynamics) so that the escalation from tractable integer pro- grams to NP-hard stochastic MILPs becomes explicit. 17 3.1. Basic Siting: Where to Build The most elemental planning question is: given a set of candidate loca- tions and a known spatial demand distribution, which sites should be acti- vated? This is a classical facility location problem (FLP), and EV charging siting inherits its rich mathematical tradition. Basic siting models include maximal covering [6, 45], set covering [46, 47], p-median [48, 49, 50, 51], p-center [52, 53, 54, 55], and flow-refueling formulations [56, 57, 58, 59, 60, 61, 62]. These models differ in whether they prioritize binary coverage, average accessibility, worst-case accessibility, or path refuelability, but they share a common PSB simplification: demand is treated as fixed, station capacity is usually simplified or omitted, and both scheduling and behavioral response are externalized. 3.1.1. Assumptions and Limitations The basic siting models in this subsection share several simplifying as- sumptions: 1. No capacity constraints: every opened station can serve unlimited de- mand [63]; 2. Deterministic, exogenous demand: origin-destination (OD) flows or nodal demands are fixed inputs [64]; 3. No grid considerations: electrical infrastructure is implicitly assumed to be available at every candidate site [65] (relaxed in Section 3.3); 18 4. Single-period: the investment decision is made once, with no temporal staging [66]. These assumptions make the problem tractable as a pure integer program of moderate scale. However, each assumption will be relaxed in subsequent subsections, progressively bringing the model closer to real-world planning challenges. The first and most immediate extension is to recognize that stations have finite capacity—the subject of the next subsection. 3.1.2. Solution Approaches The resulting siting models are binary integer programs that can be solved exactly using branch-and-bound or branch-and-cut for moderate-scale in- stances with up to a few hundred candidate sites. For larger networks, LP relaxation with rounding can provide near-optimal solutions with provable approximation guarantees [67, 68], while Lagrangian relaxation can decom- pose MCLP and FRLM formulations by dualizing budget or coverage con- straints [69]. When the candidate set expands to thousands of nodes, scalable greedy and interchange heuristics offer practical alternatives with empirically small optimality gaps [70, 71]. 3.2. Joint Siting and Sizing: How Large and What Type In practice, the planning decision is not merely where to build but also how large and what type. Under-provisioning generates excessive waiting and degrades service quality; over-provisioning wastes capital on underutilized assets. The joint siting-and-sizing problem extends the basic formulations of 19 Section 3.1 by introducing integer capacity variables that determine station size and charger mix. 3.2.1. Capacity and Technology Decisions Siting-and-sizing models extend binary siting variables x j by deciding the number of chargers of each type, n (k) j , at station j, where k ∈ K captures differences in power, cost, and service capability. This allows CPOs to bal- ance service heterogeneity and investment efficiency across Level 2 AC, DC fast charging, and related technologies [31, 72]. Some work also chooses the charging paradigm itself, including plug-in [73], battery swapping [8], and wireless charging [74]. These models typically minimize site-opening and charger costs under coverage, service, and budget constraints, with linking constraints restricting chargers to opened sites. They are formulated as MILPs with linearized ser- vice constraints or MINLPs when nonlinear queueing relations are retained. 3.2.2. Service Quality and Queueing Approximations With finite station capacity, demand can exceed available chargers, caus- ing waiting or blocking. Planning models, therefore, often impose service- quality constraints on expected delay or blocking probability. A common ap- proach embeds steady-state queueing approximations that translate charger capacity and exogenous arrival rates, usually obtained from demand fore- casts [75] or traffic assignment, into service performance [76]. Endogenizing arrivals requires coupling with the Behavior layer (Section 5). 20 The literature mainly uses three queueing paradigms. M/M/c models are most common because Erlang-C yields tractable delay estimates, de- spite the mismatch between exponential service times and realistic C-CV charging profiles [77]. Higher-fidelity M/G/c approximations allow general service-time distributions but often require interpolation or simulation-based surrogates [78, 79]. Erlang-B loss models apply when users are blocked rather than queued if all chargers are occupied [80]. These constraints restrict feasible capacity choices and intensify tradeoffs among investment cost, coverage, waiting time, and fairness, often motivat- ing multi-objective formulations [81]. Under stochastic demand, queueing and siting-sizing decisions must be optimized jointly [77], as discussed in Section 3.3. Overall, these models add finite capacity, integer sizing, and nonlinear service constraints, while typically retaining deterministic exoge- nous demand, ignoring grid interactions, and using a single-period planning horizon. 3.2.3. Solution Approaches The addition of integer sizing variables and nonlinear queueing constraints makes joint siting-and-sizing substantially harder than basic siting mod- els. With linearized, piecewise-linear, or surrogate queueing expressions, the model becomes an MILP or ILP solvable for small to moderate instances [82]. For larger instances, Benders-type decomposition is often used in EV charg- ing location and stochastic infrastructure-expansion models by separating 21 strategic siting decisions from routing, assignment, or scenario-dependent feasibility subproblems [77]. In queueing-aware models, however, nonlin- ear waiting-time or blocking-probability dependence on integer charger ca- pacity makes clean decomposition difficult. Thus, metaheuristics such as genetic algorithms and particle swarm optimization are often used to re- tain nonlinear queueing constraints while improving scalability [83]. Hybrid simulation–optimization approaches further evaluate queueing performance through simulation while optimization guides the search, especially when closed-form expressions are unavailable [84]. 3.3. Real-World Complications: Grid Coupling and Uncertainty The siting-and-sizing models of Section 3.2 assume that electrical infras- tructure is always available and that all model parameters are known with certainty. In practice, neither assumption holds. This subsection addresses two qualitatively different sources of complexity: physical constraints im- posed by the power grid, and informational constraints arising from uncer- tainty in demand, costs, and technology trajectories. 3.3.1. Grid Constraints and Transportation–Power Coupling When grid constraints are modeled explicitly, the utility or DSO be- comes a co-actor whose goals, such as deferring feeder upgrades or mini- mizing reinforcement cost, shape station feasibility. Because charging sta- tions are high-power electrical loads, grid-aware planning becomes a cou- pled transportation–power network problem, where station nodes act both 22 as transportation service facilities and distribution-network loads. Key grid constraints include transformer capacity, feeder thermal limits modeled through linearized branch-flow approximations [85] or second-order cone relaxations [86], and voltage regulation limits affected by spatial charg- ing demand [87]. These constraints can shift siting decisions away from high-demand areas with insufficient grid capacity. Many formulations jointly minimize station investment and grid rein- forcement costs under transportation-side siting and coverage constraints and power-side flow and voltage constraints [88]. Coupling arises because charger count and type determine both electrical demand and service ca- pacity. Extensions consider transformer or feeder upgrades [32], as well as co-location with PV, storage, and vehicle-to-grid resources, often yielding large-scale MINLP or mixed-integer convex models. 3.3.2. Uncertainty-Aware Optimization Planning models often assume known demand, costs, and system parame- ters, but EV adoption, charging demand, prices, policies, and technologies are uncertain. Uncertainty-aware planning is typically handled through stochas- tic, robust, distributionally robust, or data-driven approaches. Stochastic programming. Stochastic programming represents uncertainty with weighted scenarios and optimizes expected performance. In two-stage models, siting and sizing are decided before uncertainty is realized, while scenario-dependent performance is evaluated afterward [89]. Extensions 23 consider electric bus systems with uncertain travel times and battery degra- dation [90]. The main limitation is scenario explosion, commonly addressed by sample average approximation or scenario reduction. Robust and distributionally robust optimization. Robust optimization seeks solutions feasible for all realizations in an uncertainty set, with the set de- sign governing conservatism and tractability [91]. In EV charging planning, it is used to hedge against demand and renewable generation uncertainty. Distributionally robust optimization instead protects against the worst-case distribution within an ambiguity set, requiring less information than stochas- tic programming while reducing the conservatism of classical robustness [92]. Data-driven and learning-based complements. Machine learning increasingly supports uncertainty-aware planning by forecasting spatiotemporal charging demand from transactions, GPS traces, or travel surveys [93, 94, 95], or by constructing empirically calibrated uncertainty sets [96]. Recent predict- then-optimize methods align forecasting loss with downstream planning ob- jectives [97]. However, EV applications remain early, focusing mostly on demand forecasting for existing stations rather than new siting decisions [98, 99]. Overall, uncertainty-aware models relax deterministic-demand assump- tions but add computational burden through scenario growth, worst-case constraints, or prediction-error propagation, while often retaining simplified grid and single-period planning assumptions. 24 3.3.3. Solution Approaches Real-world extensions require solution strategies beyond those in Sec- tions 3.1 and 3.2. Grid-coupled models are typically large-scale MINLPs; SOCP relaxations of power-flow equations [86] can reduce them to mixed- integer SOCPs solvable for moderate networks [100], while larger cases often use decomposition between a transportation-side siting/sizing master prob- lem and a power-side load-flow feasibility subproblem [101]. ADMM has also been used to distribute computation across transportation and grid opera- tors [102]. For uncertainty-aware models, SAA converts two-stage stochastic programs into deterministic MILPs using sampled scenarios [103, 104], sce- nario reduction lowers computational burden [105], and distributionally ro- bust models are commonly solved by column-and-constraint generation [106]. 3.4. Multi-Period Dynamic Expansion Planning All preceding formulations make investment decisions in a single period. In practice, EV adoption follows a gradual trajectory spanning one to two decades, and infrastructure investment must be staged accordingly: build- ing everything at once wastes capital on underutilized assets, while building too little creates bottlenecks that suppress adoption. Multi-period planning formalizes this temporal sequencing. 3.4.1. Staged Investment Decisions The planning horizon is divided into multiple periods (typically years), and the planner decides at each period which stations to open, expand, or 25 upgrade [9]. The objective minimizes the discounted sum of incremental in- vestment and operational costs across all periods, extending the single-period formulation of Section 3.2 with a temporal dimension and a discount rate. In the EV charging literature, multi-period models typically incorporate sev- eral domain-specific features: charger technology upgrades (e.g., replacing Level 2 chargers with DC fast chargers as demand grows), coordination with distribution network reinforcement schedules so that grid capacity is available when new stations come online [10], and demand trajectories tied to regional EV adoption forecasts rather than generic growth curves [107]. Some formu- lations also allow station relocation, relaxing the irreversibility assumption when demand patterns shift or when previously installed stations become underutilized [108]. The decisions across periods are not independent. Endogenous cou- pling arises from investment irreversibility (stations once built cannot be removed), monotonic capacity expansion (charger counts can only increase), technology path dependence (early charger choices constrain later upgrades), and time-phased budget constraints. Exogenous drivers include EV fleet growth, charging technology advancement, and the emergence of alternative paradigms such as battery swapping and wireless charging. 3.4.2. Interaction with Uncertainty When a multi-period structure is combined with uncertainty, the result is a multi-stage stochastic program [66]. The scenario tree branches at each 26 decision period, and the number of scenarios grows exponentially with the number of stages, creating a curse of dimensionality in both the decision and uncertainty spaces. Unlike the single-period uncertainty models of Sec- tion 3.3, where recourse decisions are made once after uncertainty is revealed, multi-stage formulations require a contingent plan at every period that re- spects non-anticipativity: the planner cannot use information from future periods when making current decisions. This temporal nesting of decisions under uncertainty is what distinguishes multi-period stochastic planning from its single-period counterpart and drives the need for specialized decomposi- tion methods. 3.4.3. Solution Approaches Even deterministic multi-period expansion planning leads to large MILPs whose size grows with periods, candidate sites, and charger types. Com- mon solution methods include temporal decomposition and rolling-horizon heuristics, which solve truncated planning windows and fix near-term deci- sions before advancing; such methods have been used for fast-charging siting under evolving EV adoption and demand [109]. With uncertainty, multi- stage stochastic programs become substantially harder [110]. Nested Ben- ders decomposes by stage and passes feasibility/optimality cuts [111], while progressive hedging decomposes by scenario and enforces non-anticipativity, often with SAA for stochastic charging demand and expansion planning [112]. SDDP can avoid explicit scenario enumeration under convex recourse by ap- 27 proximating cost-to-go functions [91], and deep reinforcement learning has recently been used to learn investment policies from simulated demand and grid trajectories [113]. These complexities arise within the Planning layer alone; Sections 4 and 5 revisit the Scheduling and Behavior layers when in- frastructure decisions are fixed. 4. The Scheduling Layer: From Basic Dispatch to Real-Time Adap- tive Control This section traces a complexity progression in the Scheduling literature: from deterministic offline dispatch, through bidirectional scheduling under grid constraints, to stochastic online decision-making where uncertainty and state feedback require sequential adaptive methods. Unlike the Planning layer, the Scheduling layer does not have a dedicated multi-period exten- sion: scheduling decisions are inherently short-horizon (hours to days), so the temporal staging that drives Section 3.4 does not arise here. 4.1. Multi-EV Scheduling Regimes The basic scheduling problem asks when each connected EV should charge and at what rate under limited station capacity. This foundation of smart charging extends from deterministic offline scheduling to bidirectional oper- ation, pricing-based coordination, and real-time adaptive control [12, 28]. Multi-EV scheduling differs by whether vehicles are centrally coordinated or uncoordinated consumers sharing capacity. In the coordinated case, charg- ing is coupled with routing and task assignment, creating a combinatorial 28 optimization problem. In the uncoordinated case, arrivals, dwell times, and energy needs are revealed only at plug-in, forcing operators to allocate ca- pacity under uncertainty about future demand. 4.1.1. Coordinated Multi-EV: Fleet Scheduling The coordinated regime corresponds to the fleet manager actor in Table 1. A fleet operator jointly manages vehicle dispatch and charging for tasks such as delivery routes, bus trips, or ride-hailing assignments. Because a charging detour affects downstream service, charging slots, route choices, and task assignments must be optimized together. The canonical model is the Electric Vehicle Routing Problem (EVRP), especially the E-VRPTW, which extends the classical VRP with battery state and recharging stations [114]. Variants incorporate partial recharging [115, 116], nonlinear charging curves [117, 118], and heterogeneous fleets [119, 120]; see [121] for a broader review. Related electric bus models jointly optimize trip assignment and depot or network charging under capacity constraints [122, 123], with broader reviews in [124]. Here, complexity is mainly combinatorial: route choices, task assign- ments, and charging slots interact, while task sequences and vehicle states are largely known to the fleet operator. 4.1.2. Uncoordinated Multi-EV: Shared-Capacity Allocation The uncoordinated regime corresponds to the station operator actor in Table 1. A single operator allocates limited station power among indepen- dently arriving vehicles, each with an arrival time, departure deadline, and 29 energy requirement. When several vehicles charge simultaneously, their pro- files become coupled by the shared station-level power limit, so allocating more power to one vehicle reduces what is available to others. This setting falls within deterministic EV charging scheduling on paral- lel chargers with time-window and power-limit constraints [29]. Objectives include minimizing electricity cost under time-of-use tariffs, reducing trans- former peak load [125], and improving fairness across heterogeneous charging needs [126]. Overall, the problem becomes a resource-constrained scheduling program whose size scales with the number of vehicles and time slots. 4.1.3. Assumptions and Limitations Despite this regime-specific difficulty, both regimes above, in their base- line form, share several restrictive assumptions that the remainder of Sec- tion 4 progressively relaxes: 1. Deterministic arrivals and departures: every vehicle’s dwell-time win- dow and energy requirement are known at the time of scheduling, per- mitting the entire horizon to be optimized in a single pass. 2. Unidirectional charging without upstream grid modeling: vehicles can only draw power, and network-side constraints beyond the station-level limit are ignored. 3. Offline scheduling: the schedule is computed once before execution, with no re-optimization as conditions evolve. 30 Section 4.2 relaxes assumptions (2) and partially (1) by introducing bidirec- tional power exchange and upstream grid constraints, while Section 4.3 re- laxes assumptions (1) and (3) by replacing deterministic inputs with stochas- tic arrivals, prices, and renewable generation, and by moving from offline to online, closed-loop scheduling. 4.1.4. Solution Approaches The two scheduling regimes require different solution strategies. Coordi- nated fleet scheduling inherits the NP-hardness of EVRP/VRP, so exact branch-and-price or branch-and-cut methods are limited to moderate in- stances. Larger fleets typically rely on ALNS and related metaheuristics, which scale well but sacrifice optimality guarantees [114, 121]; decomposi- tion methods that separate routing and charging offer a practical middle ground. For uncoordinated shared-capacity allocation, deadline-based online rules such as EDF and LLF are simple and effective at plug-in time [127, 128], while LP, MILP, or QP dispatch models optimize coupled multi-EV charging with fairness-aware objectives, such as proportional allocation by remaining energy need, but must be resolved at each decision epoch [29]. 4.2. Bidirectional Scheduling and Grid Integration Section 4.1 treated each vehicle as a passive, unidirectional load and ig- nored upstream network constraints. This subsection relaxes both assump- tions. Conceptually, this is also where the aggregator and DSO actors enter: bidirectional power flows and feeder constraints are precisely the dimensions 31 in which an aggregator monetizes flexibility and a DSO enforces network limits. 4.2.1. V2G and Bidirectional Power Allocation Vehicle-to-grid (V2G) expands EV scheduling from charge-only control to bidirectional charging and discharging, turning EVs from passive loads into controllable energy resources. Because discharge is constrained by bat- tery C-rate and inverter limits, the feasible action space is asymmetric. V2G can support peak shaving, frequency regulation, and energy arbitrage [129]; frequency regulation is especially attractive due to EV batteries’ fast re- sponse [130], while arbitrage exploits time-of-use price differences [131]. Bidirectional operation adds trade-offs around battery degradation, con- version losses, and mobility requirements. Since aging depends on charge rate and depth of discharge [132], repeated V2G cycling can accelerate capacity fade. Degradation-aware models therefore balance V2G revenue against bat- tery health while ensuring minimum departure energy [133]. Related vehicle- to-building settings use EVs to supply building loads under variable prices and demand charges, and have been studied with multi-agent reinforcement learning [134], MDP-based online search [135], and neuro-symbolic MPC- learning frameworks [136]. 4.2.2. Grid Constraints and Multi-Resource Coordination When station power exchange is large relative to local distribution ca- pacity, grid constraints such as transformer limits, feeder thermal limits, 32 and voltage bounds become binding. Including these constraints turns sta- tion scheduling into a joint scheduling-and-power-flow problem, often using linearized power-flow models for tractability [137]. This issue is especially important in depot charging, where concentrated transit-fleet loads require joint trip assignment, charge scheduling, and distribution-network coordina- tion [138]. At the network level, coordinating multiple stations on the same feeder can reduce grid stress by shifting charging loads across space and time [139, 140]. Co-located PV and stationary storage add another layer of coor- dination, requiring joint dispatch of EV charging/discharging, battery cy- cling, and curtailable solar generation under station- and grid-level con- straints [141]. 4.2.3. Solution Approaches Bidirectional scheduling and grid-coupled dispatch introduce nonlinear- ities from battery degradation models, power flow equations, and the ex- panded charge-discharge action space. When degradation is modeled as a linear or piecewise-linear function of throughput and depth of discharge, the V2G scheduling problem can be formulated as an LP or MILP solvable by commercial solvers for single-station instances [142]. For grid-coupled for- mulations, linearized branch-flow approximations reduce the power flow con- straints to linear or second-order cone form, enabling joint scheduling-and- power-flow optimization at moderate network scales [143]. Multi-resource co- 33 ordination problems involving EV fleets, stationary storage, and local renew- ables are typically formulated as MILPs with time-coupled state-of-charge constraints; for larger instances, rolling-horizon and distributed optimization methods partition the problem temporally or spatially [144]. 4.3. Real-World Complications: Uncertainty and Adaptive Scheduling In practice, the Scheduling layer faces short-run operational uncertainty (hours to days) that unfolds while decisions are being executed, driven by stochastic arrivals, real-time prices, and intermittent renewables [12]. Two paradigms address this challenge, distinguished by how uncertainty is re- vealed and whether closed-loop state feedback is available. 4.3.1. Anticipative Optimization under Uncertainty Anticipative scheduling commits to a schedule or contingency plan at one decision point while hedging against uncertain future demand, prices, or renewable generation. Stochastic programming. Two-stage stochastic programming first sets a day-ahead schedule, then uses recourse actions such as real-time power adjustments after uncertainty is revealed [145]. Applications include fast- charging stations with stochastic arrivals [146], fleet depots with uncertain trip schedules [147], and joint charging–energy trading under volatile whole- sale prices [148]. As in planning models, scenario richness improves fidelity but increases computational burden. 34 Robust and distributionally robust optimization. Robust optimization pro- tects against worst-case outcomes within uncertainty sets [149], while distri- butionally robust optimization hedges against worst-case distributions within ambiguity sets [150]. Scheduling uncertainty sets often cover arrival inter- vals, price bands, or forecast errors. Applications include robust day-ahead EV aggregator scheduling [151] and distributionally robust coordination un- der ambiguous renewable forecasts [152]. These methods improve feasibility guarantees but may sacrifice expected-case performance. 4.3.2. Sequential Decision-Making under Uncertainty The second paradigm abandons one-shot optimization in favor of sequen- tial decisions that respond to the evolving system state. The scheduling problem is formalized as a Markov decision process (MDP) with state s t encoding vehicle charge levels, queue occupancy, and grid signals; action a t collecting power allocation decisions; transition kernel capturing stochastic arrivals and price dynamics; and reward measuring operational objectives (cost minimization, service quality, grid stability) [153]. The optimal policy satisfies the Bellman optimality equation, but the state space grows com- binatorially with the number of vehicles, charger types, and time granular- ity, rendering exact dynamic programming intractable for realistic station sizes [154]. This intractability motivates the approximate solution methods discussed below. 35 4.3.3. Solution Approaches Anticipative scheduling uses methods similar to the Planning layer (Sec- tion 3.3), including SAA, column-and-constraint generation, and scenario re- duction for uncertain arrivals and prices. Sequential control often uses MPC, which repeatedly solves a finite-horizon problem, implements the first action, and re-optimizes as new information arrives; it is attractive because it needs no offline training, handles hard constraints, and is robust to forecast errors, with applications in workplace charging [13], PV-coupled fast charging [155], and fleet charging under time-varying prices [156]. When repeated optimiza- tion is too costly, or models are unavailable, DRL can learn policies from sim- ulation for real-time scheduling [157, 158] and dynamic pricing [159], though constraints are often penalty-based and generalization remains challenging. Multi-station settings extend these ideas through decentralized MARL [160] or hierarchical distributed decomposition [161]. Overall, this section treats infrastructure and user behavior as exogenous, while the next section models user response to prices, wait times, and station attributes. 5. The Behavior Layer: From Individual Choice to Dynamic Equi- librium EV charging choice is modeled within the random utility framework [162, 163], but two features distinguish it from generic facility choice. First, the feasible station set depends on the vehicle’s remaining energy; second, the charging context (en route versus destination) affects both the relevant al- 36 ternatives and the attributes that shape utility. 5.1. Basic Choice: Where to Charge The most basic behavioral question is straightforward: given a set of available charging stations, which one does a user choose? As in the broader transportation demand literature, EV charging choice is commonly modeled within a random utility framework [162, 163]. However, two features make EV charging different from generic facility choice. First, the set of stations that a driver can realistically consider depends on the vehicle’s remaining energy. Second, the charging context, especially the distinction between en- route and destination charging, affects both the relevant alternatives and the factors that shape utility. 5.1.1. The Discrete-Choice Benchmark A natural benchmark in this literature is the standard discrete-choice model, which most directly describes a private driver such as a commuter or long-distance traveler. In this benchmark setting, the feasible station set is treated as given, and the user’s decision depends on personal characteristics together with observable station attributes. The central idea is that each driver compares the available charging options and chooses the one with the highest perceived utility. In practice, the systematic part of utility is typically specified as a func- tion of variables such as distance, charging price, charger type, and posted waiting time, while the remaining term captures unobserved preference vari- 37 ation. Under the usual assumptions, this leads to the multinomial logit (MNL) model [162], which remains the canonical benchmark for charging- choice analysis. Empirical applications of MNL to EV charging station choice have identified travel distance, charging cost, and waiting time as the domi- nant attributes [164]. When stations can be grouped naturally, for example by area, operator, or charging technology, nested logit is often used to relax the independence-of-irrelevant-alternatives assumption by allowing correla- tion within groups [163]. In this section, MNL and nested logit serve as the baseline choice models; richer formulations such as mixed logit and latent- class models are discussed later in Section 5.3. 5.1.2. Endogenous Feasibility and Charging Context EV charging choice differs from generic facility choice because the feasible choice set depends on the vehicle state. A station is relevant only if the EV has enough remaining energy to reach it, so battery state directly shapes feasible alternatives [165]. In practice, reachability also depends on traffic, temperature, and driving style, making battery-based feasibility a first-order approximation. Charging context further affects choice-set construction and utility design. In en-route charging, drivers must recharge to complete an ongoing trip, so route geometry, remaining range, detour distance, charger speed, and waiting time are central. In destination charging, where vehicles charge while already parked at home, work, or commercial sites, range constraints are weaker and 38 the trade-off shifts toward cost, convenience, and dwell-time compatibility. These differences explain why empirical studies often model en-route and destination charging separately [166, 164]. 5.1.3. Assumptions and Limitations The basic choice models in this subsection share three simplifying as- sumptions: 1. Exogenous and independent attributes: station prices, wait times, and availability are fixed inputs, unaffected by the aggregate choices of other users, so that each user’s utility depends only on station attributes and personal characteristics; 2. Static: the decision is made once, with no temporal dynamics or re- peated interaction; 3. Utility-maximizing response: each user selects the alternative that max- imizes perceived random utility, given the attributes of available alter- natives. These assumptions reduce the problem to a standard discrete-choice model. The subsections that follow progressively relax them: first endogenizing con- gestion (Section 5.2), then enriching the resulting equilibrium along several dimensions (Section 5.3), and finally introducing temporal dynamics (Sec- tion 5.4). 39 5.2. Choice with Feedback: Static Equilibrium The equilibrium analysis below treats users as a homogeneous population. Once assumption (i) of Section 5.1 is relaxed, station attributes are no longer exogenous: the wait time at a station depends on how many users select it, which in turn depends on the wait times, creating a circular dependence that transforms the individual choice problem into a population-level consistency problem. 5.2.1. From Individual Choice to Population Fixed Point At the population level, each user’s station choice depends on congestion, while congestion is determined by the aggregate distribution of users across station-time-route alternatives. An equilibrium occurs when congestion- aware individual choices reproduce the same aggregate distribution. This fixed-point structure, introduced in Section 2, has been used to model congestion-coupled EV station choice in urban networks [17]. 5.2.2. Equilibrium Benchmarks Two classical benchmarks are commonly used. Wardrop user equilibrium (UE) assumes deterministic utility: all used stations or routes have equal generalized cost, and no unused alternative is cheaper [167]. It provides the deterministic fixed-point benchmark and can also be written as a variational inequality. Stochastic user equilibrium (SUE) instead allows random utility, yielding a logit-based fixed point in which choice probabilities depend on congestion-dependent utilities [168]. 40 This section adopts SUE as the main benchmark because stochastic per- ception fits discrete charging choice under imperfect information. Wardrop UE is recovered when perception noise vanishes, but in EV charging it is mainly a conceptual anchor because SOC-dependent reachability and range constraints make feasible paths state-dependent [169]. When users jointly choose routes and charging stations, the equilibrium extends traf- fic assignment to EV-specific and coupled transportation–power network equilibria [38, 170]. 5.2.3. Solution Approaches Computing the fixed-point equilibrium requires iterative methods whose choice depends on the equilibrium concept. For deterministic Wardrop UE, the equivalent convex optimization formulation [171] permits gradient-based and Frank-Wolfe algorithms, with convergence guaranteed under monotone cost mappings. For stochastic user equilibrium, the method of successive averages (MSA) is the most widely adopted approach: at each iteration, a new auxiliary flow is generated from the current congestion-dependent choice probabilities and averaged with the incumbent solution [172, 173]. MSA is simple to implement and converges under mild conditions, but convergence can be slow for large networks. More recently, path-based and projection- based variational inequality solvers have been adapted to EV-specific network equilibrium problems where SOC-dependent feasibility constraints alter the structure of the feasible set [169]. 41 5.3. Relaxing the Benchmark Assumptions The static equilibrium of Section 5.2 rests on three benchmark assump- tions: users are homogeneous, respond according to a single utility-consistent rule, and are individually negligible. This subsection relaxes each assump- tion along an independent dimension: user type, decision rule, and popula- tion structure. These are parallel relaxations, not a progression; each can be pursued independently of the others. 5.3.1. User Heterogeneity and Multi-Class Equilibrium Charging users differ in value of time, price sensitivity, range anxiety, and vehicle attributes [174, 175]. Multi-class equilibrium models capture this by dividing users into classes with distinct utility functions, while requiring each class’s choices to be consistent with the congestion jointly generated by all classes. This coupling increases state-space size and makes uniqueness harder to guarantee. When heterogeneity is unobserved, latent-class and mixed-logit models can infer population type distributions for use in these coupled equi- libria [20, 176]. 5.3.2. Bounded Rationality Benchmark equilibria assume users respond optimally to the modeled utility, but real drivers may satisfice, follow habits, or perceive costs imper- fectly. Bounded rationality captures these departures from utility-consistent response through several parallel mechanisms. 42 Relaxed optimization uses an indifference band, so users switch only when cost differences exceed a threshold [177]. Switching-friction models add habit or inertia costs for changing from familiar alternatives, with the dynamic for- mation of habits deferred to Section 5.4 [178]. Distorted value perception, including prospect-theory and reference-dependent utility, captures asym- metric reactions to gains and losses, such as aversion to unexpected price or wait-time increases [179]. Noisy-response models such as quantal response equilibrium replace exact best response with probabilistic choice whose er- ror declines with payoff differences [180]; in EV charging, QRE has been combined with level-k reasoning for congestion-aware station choice [181]. Each mechanism modifies the response map, potentially changing equi- librium existence, multiplicity, and welfare properties. 5.3.3. Finite-Player Strategic Interaction This relaxation changes the population structure. When a fleet oper- ator or aggregator controls many vehicles, each player’s charging strategy can noticeably affect congestion, so the equilibrium shifts from a nonatomic fixed point to a finite-player Nash equilibrium. Strategies usually span ve- hicle charging schedules over time, and when congestion admits a common potential function, the equilibrium can be characterized through potential minimization [182]. With asymmetric roles, such as aggregators interacting with individual users or market actors, Stackelberg models capture leader– follower behavior [23]. These relaxations enrich the static equilibrium object; 43 the next subsection adds temporal dynamics. 5.4. Dynamic Behavior and Learning The equilibrium concepts developed in Sections 5.2 and 5.3 characterize self-consistent population distributions but do not address how such distri- butions arise or evolve. This subsection introduces the temporal dimension, moving from discrete-time adaptation rules to continuous-time evolutionary dynamics, then to the mean field game equilibrium. The day-to-day learning and adaptation models below are most plausibly empirically grounded in the TNC driver class, where repeated interaction with the system and observable earnings make learning dynamics directly identifiable. 5.4.1. Day-to-Day Adaptation Day-to-day models describe how users revise charging choices over discrete periods, such as daily, based on experienced wait times, price changes, and station availability. Common update rules include best- response switching, proportional shifts toward better-performing alterna- tives, and reinforcement-based increases in the probability of choices with favorable outcomes [183]. A central question is whether these dynamics converge to the static equi- libria in Sections 5.2–5.3. Convergence depends on the update rule and congestion structure, with stability analysis linking dynamic adjustment to static equilibrium [184]. 44 5.4.2. Evolutionary Dynamics For revision rules such as imitation or payoff-proportional updating, day- to-day adaptation can be approximated by continuous-time ordinary differ- ential equations. Replicator dynamics are a canonical example: alternatives with above-average payoff gain population share over time [185]. This for- mulation is an analytically tractable limit of specific behavioral update rules, not a separate paradigm. Its main advantage is that ODE stability tools can analyze the conver- gence of the adjustment process, as shown in traffic choice dynamics [186]. EV charging applications remain limited, but the framework is suitable for studying how station-choice patterns evolve as EV adoption and congestion change [187]. 5.4.3. Mean Field Game Equilibrium Mean field games (MFGs) [39, 188] model dynamic equilibrium in large populations of interacting agents. They couple individual optimal control, given the population distribution, with the evolution of that distribution under aggregate decisions. The resulting fixed point describes a trajectory where no individual benefits from deviating. MFGs extend the large-population, utility-consistent benchmark of Sec- tion 5.2 to dynamic settings, but do not directly capture the bounded ra- tionality or finite-player structure in Section 5.3. In EV charging, states may include battery level, location, and time of day, while controls represent 45 charging decisions or rates [189]. Existence and uniqueness often rely on monotonicity of the cost coupling [39]. 5.4.4. Solution Approaches The dynamic behavior models vary in computational complexity. Day- to-day adaptation and evolutionary dynamics are typically solved by for- ward simulation or ODE integration, with convergence checked numerically or through Lyapunov arguments [190, 191]. MFG equilibria require coupled backward–forward systems, usually solved after state-time discretization as finite-dimensional equilibrium or complementarity problems. When analyt- ical equilibria are unavailable, agent-based simulation provides a flexible al- ternative for studying how charging patterns emerge, stabilize, or oscillate through repeated interaction [19]. ABS populates a synthetic environment with boundedly rational agents using decision rules from Sections 5.1–5.3, and simulates population-level outcomes [192, 193]. Thus, ABS is not an equilibrium concept but a computational framework for richer behavioral assumptions and environments. Overall, the Behavior layer moves from in- dividual choice to congestion-coupled equilibrium and dynamic adaptation, while treating planning and scheduling decisions as fixed inputs. The next section relaxes these assumptions to examine cross-layer interactions. 6. Pairwise Coupling: A Structured Gap Analysis This section reads the three pairwise-coupling literatures (P×S, S×B, P×B) through the trilemma of Section 2.4, asking for each pair what the 46 joint literature endogenizes, what surrogate it adopts for the exogenized third layer, and what that surrogate costs in fidelity. Across all three pairs, the dominant response is to replace the exogenized layer with a static aggregate, a recurring pattern we call the static-aggregate compromise. Table 3 provides an overview for this section. Table 3: Summary of pairwise coupling gaps in the PSB framework. Coupling Endogenized layers Typical omitted-layer surrogate What is lostSuitable objective P×S Planning and scheduling Behavior represented as fixed demand, fixed arrival rates, or aggregate utilization User heterogeneity, station choice response, adoption feedback, and equity effects Grid-aware infrastructure planning and capacity sizing S×B Scheduling and behavior Planning represented as fixed station locations, fixed charger capacities, or static network topology Long-run investment feedback, capacity expansion effects, and siting adaptation Dynamic pricing, demand response, and smart charging control P×B Planning and behavior Scheduling represented as fixed service rates, queueing functions, utilization factors, or average congestion Temporal grid interaction, peak-load dynamics, marginal emissions, and operational flexibility Access-oriented siting, equity analysis, and station-choice-aware planning 47 6.1. Planning × Scheduling As Section 2.3 showed, coupling planning and scheduling produces a bilevel program whose lower level must be replaced with a tractable surro- gate. Across the joint P+S literature, the surrogates fall into three families arranged along a fidelity spectrum [27]. Notably, the advanced scheduling methods reviewed in Section 4.3 (MPC, stochastic programming, deep rein- forcement learning) do not yet appear as the lower level in any mainstream joint P+S formulation. At the lowest fidelity, the scheduling layer is collapsed into an average utilization factor fed back into the planning objective [7, 194, 82]. This yields a closed-form objective solvable as a standard mixed-integer program, but erases all temporal structure: peak-shaving, time-of-use response, and every feature that distinguishes one scheduling policy from another reduces to a single scalar. At intermediate fidelity, a representative-day deterministic dispatch pro- gram is embedded as the lower level, typically a linear program over a small number of typical days [22, 195]. This preserves intra-day load shape and price structure while staying within convex tractability, but flattens price volatility, ramping constraints, and V2G arbitrage into a deterministic sched- ule. At the highest fidelity currently available, a full bilevel program is re- tained with the lower level restricted to a problem class that admits closed- form reformulation, such as a linear or convex quadratic dispatch [21], or a 48 tractable demand-allocation problem solved through cooperative game con- ditions [42]. These preserve lower-level optimality but cannot accommodate stochastic arrivals or learned policies. A small frontier pushes beyond this through stochastic bilevel formulations [37] and learning-based approxima- tions that replace analytical lower-level reformulation with simulation-trained operational policies or configuration policies [113]. The three surrogates differ in how much intra-day structure they retain, but share a common cost: each collapses the hour-scale scheduling horizon into the year-scale planning horizon, aggregating away temporal variation. This timescale collapse returns in sharper form in Section 6.3, where the gap between planning and behavioral response is even wider. 6.2. Scheduling × Behavior As Section 2.3 showed, closing the operator-user feedback loop transforms the scheduling problem into a multi-agent stochastic game whose complexity grows with the number of user types. The joint S×B literature responds by splitting into two camps, each closing only one direction of the loop while reducing the other to a static aggregate. On one side, dynamic-pricing and demand-response papers treat the op- erator’s decisions as endogenous and absorb user response into a parametric demand function, such as a logit choice probability or a linear price elastic- ity [34, 196, 197, 198]. This reduces the operator’s problem to a Stackelberg or single-level optimization, but compresses user heterogeneity and strategic in- 49 teraction into a single demand curve. Rather than compressing user response into a demand curve, negotiation-based frameworks offer drivers incentive- backed options for flexibility in departure time or requested state of charge, aligning operator and user objectives through mechanism design [199]. On the other side, equilibrium routing and station-choice papers endog- enize user response through a discrete-choice or user-equilibrium model and let operator scheduling enter only through a fixed congestion or queueing function [17, 200, 201]. The cost is symmetric: operator pricing and dis- patch are frozen, suppressing feedback from real-time adaptation into the user equilibrium. The two camps are mirror images: each endogenizes one direction of the S↔B interface and reduces the other to a static aggregate. A small frontier of mean-field game and multi-agent reinforcement learning work attempts to close both directions simultaneously [189, 202, 203], but it remains small and mostly simulation-based. Across both camps and the frontier, the common cost is that the full bidirectional feedback between operator adaptation and user response has not yet been captured in a single tractable formulation. 6.3. Planning × Behavior Of the three pairs, P×B has the smallest joint literature. The struc- tural obstacle is that embedding the behavioral fixed point inside a plan- ning bilevel program produces a mathematical program with equilibrium constraints (MPEC) whose nonconvexity admits no clean reformulation un- 50 der logit choice or discrete travel decisions [35, 26, 204]. Two compounding reasons beyond this mathematical difficulty explain why the joint literature remains especially thin. First, the time-scale gap between planning (years) and behavioral re- sponse (days) is on the order of 365:1, far wider than the 24:1 ratio in P×S, making representative-day aggregation insufficient for capturing behavioral evolution over a changing network. Second, no institutional actor spans both long-horizon deployment and individual driver behavior: the P↔B edge has no institutional actor with direct control over both layers, unlike P↔S (in- tegrated utility) or S↔B (fleet manager). The rare P×B papers accordingly represent behavior as a fixed demand surface [36]. Together, these reasons compound the mathematical obstacle and keep the joint literature thin. The preceding subsections document the specific forms the static- aggregate compromise takes across all three pairs. A consistent pattern emerges: the pairs whose coupling edge is spanned by an institutional actor (P×S by the integrated utility, S×B by the fleet manager) have generated learning-based or simulation-based alternatives to the static default, while P×B, which lacks such an actor, also lacks such alternatives. Section 7 uses these findings to ask a forward-looking question: given that some simplifi- cation is unavoidable, which layer should be simplified, and what does that choice cost? 51 7. Future Research Directions Section 2.4 showed that practical joint formulations at a realistic scale typically simplify at least one layer, and that each simplification forecloses a different set of objectives. This section outlines four directions for future re- search, spanning methodology, data, emerging technologies, and policy. The purpose of future work should not be to eliminate simplification entirely, but to make simplification measurable, objective-dependent, and institutionally transparent. Quantifying the cost of simplification. The central open question is not which surrogate to use but how much each one costs in decision quality. Answer- ing this requires controlled experiments in which the same planning or pol- icy problem is solved at multiple surrogate fidelity levels and the outcomes are compared on decision-relevant metrics: total system cost, realized emis- sions, equity of access, or grid constraint violations. Such experiments would reveal, for instance, whether a representative-day scheduling surrogate pre- serves enough temporal variation to rank siting alternatives correctly for emission objectives, or whether a parametric demand function produces sta- tion placements that perform well once real users interact with the network. Existing work on bilevel approximation [37] and mean-field behavior mod- els [39] provides candidate surrogates at different fidelity levels, but head- to-head comparisons on a common instance are absent. Establishing these comparisons would convert the trilemma from a qualitative diagnostic into a 52 quantitative design tool. Data infrastructure and empirical calibration. Surrogate comparison requires shared data, but the field currently lacks standardized benchmarks. Three types of data are needed in combination: charging-session records with tem- poral and spatial resolution (session start, duration, energy delivered, station identity), distribution feeder models with transformer ratings and voltage constraints, and disaggregate user-behavior data linking charging choices to trip purpose, income, and vehicle type. Some of these are becoming partially available through open charging-network APIs and utility smart-meter pro- grams, but they remain fragmented across operators and jurisdictions, often proprietary, and rarely linked to behavioral covariates. On the empirical side, the behavioral models reviewed in Section 5 offer rich theoretical structures for heterogeneity, bounded rationality, and dynamic learning, but calibration against observed charging behavior at the urban scale is rare. Bridging this gap would ground the surrogate comparison of the previous paragraph in empirically validated models rather than stylized assumptions. Emerging technologies and evolving coupling structures. The preceding two directions address how to use the current PSB framework more rigorously. The remaining two ask whether the framework itself will need to evolve. The PSB framework is built on the current paradigm of human-driven EVs charging at fixed stations. Several emerging technologies could change which coupling edges matter and which simplifications are appropriate. 53 Autonomous vehicles shift charging decisions from individual drivers to fleet-level optimization algorithms, potentially collapsing the Behavior layer into the Scheduling layer and dissolving the S×B boundary. Battery swap- ping replaces continuous charging with discrete swap events, changing the queueing and scheduling structure at the operational level. Dynamic wire- less charging embeds infrastructure in the road surface, transforming the Planning problem from station siting to corridor design. Vehicle-to-grid and vehicle-to-building bidirectional energy exchange strengthens the cou- pling between scheduling and grid or building operations, making the S×P interface more consequential. Each of these technologies does not simply add complexity to the existing framework; it restructures the trilemma by altering which layers are coupled and how tightly. Extending the PSB frame- work to accommodate these shifts is a necessary step toward keeping the diagnostic current as the technology landscape evolves. Policy and institutional design. The trilemma shapes not only modeling choices but also the policy decisions that determine who plans, operates, and regulates charging infrastructure. Consider two concrete policy scenarios. When a government decides whether to allow vertical integration between network planning and station operation, the trilemma clarifies the stakes: vertical integration internalizes the P↔S coupling, reducing the need for scheduling surrogates in planning models, but it concentrates market power and may weaken the competitive signals that drive operational efficiency. When a regulator mandates open data sharing between operators and mu- 54 nicipal planners, it enables the planner to use higher-fidelity scheduling and behavioral inputs without controlling those layers directly, partially relieving the trilemma through institutional design rather than algorithmic improve- ment. More broadly, different policy objectives demand different modeling fidelity: tariff regulation requires detailed S×B modeling, subsidy allocation for station construction requires P×B modeling, and grid-integration stan- dards require P×S modeling. The trilemma tells each policy-maker which layer their model simplifies and, therefore, which outcomes their policy is blind to. Developing this connection between computational tradeoffs and governance design is a direction that sits squarely within the scope of trans- port policy research. 8. Conclusion This survey introduced a three-layer Planning, Scheduling, and Behavior (PSB) framework that separates long-term infrastructure decisions, short- term operational control, and user charging response while making their coupling explicit. It also identifies a PSB trilemma: realistic fidelity across all three layers is difficult to achieve without losing tractability. Existing cross-layer studies therefore typically integrate two layers while simplifying the third through fixed assumptions or aggregate surrogates. The central implication is that simplification is often necessary but should be deliberate. Simplifying Planning preserves operational and behavioral de- tail but ignores investment adaptation; simplifying Scheduling enables siting 55 and equilibrium analysis but omits temporal grid dynamics; and simplifying Behavior supports grid-aware planning or dispatch but suppresses hetero- geneity, strategic response, and equity-relevant differences. The appropriate abstraction therefore depends on the decision-maker, objective, and institu- tional context. Important gaps remain. High-fidelity pairwise models are still limited, full three-layer formulations are rare, and empirical validation is weak because few benchmarks combine charging sessions, distribution-network data, and disaggregate behavioral information. User heterogeneity is often assumed rather than calibrated, and the decision-quality loss from simplified surro- gates is seldom measured. Thus, it remains unclear when simplified models are adequate for investment, pricing, emissions, or equity analysis. The PSB framework should therefore serve as a design tool: it helps re- searchers state which layers are modeled, which are simplified, which coupling edges are retained, and which objectives may be distorted. Future progress requires stronger layer-specific methods, benchmark datasets for surrogate evaluation, empirical calibration of charging behavior, and extensions to au- tonomous fleets, battery swapping, dynamic wireless charging, V2G/V2B integration, and institutional mechanisms for data sharing and coordinated decision rights. 56 Declaration of generative AI and AI-assisted technologies in the manuscript preparation process During the preparation of this work, the authors used ChatGPT to sup- port language editing, rephrasing, and clarity of presentation. After using this tool, the authors reviewed and edited the content as needed and take full responsibility for the content of the manuscript. Data availability No data were used for the research described in the article. References [1] P. Jaramillo, S. K. Ribeiro, P. Newman, S. Kobayashi, T. Boulou- chos, D. Bun, J. Katima, D. S. Lee, S. B. Nugroho, R. Slade, A. V. de Carvalho, L. Fulton, Transport, in: P. Shukla, J. Skea, R. Slade, A. A. Khourdajie, R. van Diemen, D. McCollum, M. Pathak, S. Some, P. Vyas, R. Fradera, M. Belkacemi, A. Hasija, G. Lisboa, S. Luz, J. Mal- ley (Eds.), Climate Change 2022: Mitigation of Climate Change. Con- tribution of Working Group I to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change, Cambridge Univer- sity Press, Cambridge, UK and New York, NY, USA, 2022, Ch. 10. doi:10.1017/9781009157926.012. [2] M. S. Koroma, A. Alwosheel, Aligning vehicle electrification with power sector transitions: life cycle insights across diverse grids, 57 npj Sustainable Mobility and Transport (2026). doi:10.1038/ s44333-025-00076-y. [3] International Energy Agency, Global EV outlook 2025: Expanding sales in diverse markets, IEA, Paris, licence: C BY 4.0 (2025). URL https://w.iea.org/reports/global-ev-outlook-2025 [4] Y.-W. Wang, C.-C. Lin, Locating multiple types of recharging stations for battery-powered electric vehicle transport, Transportation Re- search Part E: Logistics and Transportation Review 58 (2013) 76–87. doi:https://doi.org/10.1016/j.tre.2013.07.003. URL https://w.sciencedirect.com/science/article/pii/ S1366554513001403 [5] M. Kuby, S. Lim, The flow-refueling location problem for alternative- fuel vehicles, Socio-Economic Planning Sciences 39 (2) (2005) 125–145. doi:https://doi.org/10.1016/j.seps.2004.03.001. URL https://w.sciencedirect.com/science/article/pii/ S0038012104000175 [6] R. Church, C. ReVelle, The maximal covering location problem, Papers in Regional Science 32 (1) (1974) 101–118. doi:10.1007/BF01942293. [7] S. Ghanbari Motlagh, L. Li, A review on electric vehicle charg- ing station planning: Infrastructure placement, sizing, upgrades, and uncertainties, Journal of Energy Storage 141 (2026) 119325. 58 doi:https://doi.org/10.1016/j.est.2025.119325. URL https://w.sciencedirect.com/science/article/pii/ S2352152X25040381 [8] H.-Y. Mak, Y. Rong, M. Shen, Infrastructure planning for electric vehicles with battery swapping, Management Science 59 (03 2012). doi:10.2139/ssrn.2022651. [9] J. Zhang, Z. Wang, E. J. Miller, D. Cui, P. Liu, Z. Zhang, Z. Sun, Multi-period planning of locations and capacities of public charging stations, Journal of Energy Storage 72 (2023) 108565. doi:https://doi.org/10.1016/j.est.2023.108565. URL https://w.sciencedirect.com/science/article/pii/ S2352152X2301962X [10] J. Cui, Y. Cao, B. Wang, J. Wu, Multi-stage adaptive expansion of ev charging stations considering impacts from the transporta- tion network and power grid, Applied Energy 386 (2025) 125544. doi:https://doi.org/10.1016/j.apenergy.2025.125544. URL https://w.sciencedirect.com/science/article/pii/ S0306261925002740 [11] E. Sortomme, M. A. El-Sharkawi, Optimal charging strategies for unidi- rectional vehicle-to-grid, IEEE Transactions on Smart Grid 2 (1) (2011) 131–138. doi:10.1109/TSG.2010.2090910. URL http://dx.doi.org/10.1109/TSG.2010.2090910 59 [12] S. Ghanbari Motlagh, J. Oladigbolu, L. Li, A review on elec- tric vehicle charging station operation considering market dy- namics and grid interaction, Applied Energy 392 (2025) 126058. doi:https://doi.org/10.1016/j.apenergy.2025.126058. URL https://w.sciencedirect.com/science/article/pii/ S0306261925007883 [13] B. Hermans, S. Walker, J. Ludlage, L. Özkan, Model predictive control of vehicle charging stations in grid-connected microgrids: An implementation study, Applied Energy 368 (2024) 123210. doi:https://doi.org/10.1016/j.apenergy.2024.123210. URL https://w.sciencedirect.com/science/article/pii/ S0306261924005932 [14] D. Qiu, Y. Wang, W. Hua, G. Strbac, Reinforcement learning for electric vehicle applications in power systems:a critical review, Renewable and Sustainable Energy Reviews 173 (2023) 113052. doi:https://doi.org/10.1016/j.rser.2022.113052. URL https://w.sciencedirect.com/science/article/pii/ S1364032122009339 [15] L. Zhang, B. Shaffer, T. Brown, G. Scott Samuelsen, The optimization of dc fast charging deployment in california, Applied Energy 157 (2015) 111–122. doi:https://doi.org/10.1016/j.apenergy.2015.07.057. 60 URL https://w.sciencedirect.com/science/article/pii/ S0306261915008971 [16] X. Xi, R. Sioshansi, V. Marano, Simulation–optimization model for location of a public electric vehicle charging infrastructure, Trans- portation Research Part D: Transport and Environment 22 (2013) 60–69. doi:https://doi.org/10.1016/j.trd.2013.02.014. URL https://w.sciencedirect.com/science/article/pii/ S1361920913000345 [17] Y. Huang, K. M. Kockelman, Electric vehicle charging station loca- tions: Elastic demand, station congestion, and network equilibrium, Transportation Research Part D: Transport and Environment 78 (2020) 102179. doi:https://doi.org/10.1016/j.trd.2019.11.008. URL https://w.sciencedirect.com/science/article/pii/ S1361920919306406 [18] F. He, Y. Yin, S. Lawphongpanich, Network equilibrium mod- els with battery electric vehicles, Transportation Research Part B: Methodological 67 (2014) 306–319. doi:https: //doi.org/10.1016/j.trb.2014.05.010. URL https://w.sciencedirect.com/science/article/pii/ S0191261514000915 [19] Z. Yi, B. Chen, X. C. Liu, R. Wei, J. Chen, Z. Chen, An agent- based modeling approach for public charging demand estimation 61 and charging station location optimization at urban scale, Com- puters, Environment and Urban Systems 101 (2023) 101949. doi:https://doi.org/10.1016/j.compenvurbsys.2023.101949. URL https://w.sciencedirect.com/science/article/pii/ S0198971523000121 [20] Y. Yang, S. Yang, H. Moon, J. Woo, Analyzing heteroge- neous electric vehicle charging preferences for strategic time- of-use tariff design and infrastructure development: A la- tent class approach, Applied Energy 374 (2024) 124074. doi:https://doi.org/10.1016/j.apenergy.2024.124074. URL https://w.sciencedirect.com/science/article/pii/ S0306261924014570 [21] K. Prakash, M. Ali, M. N. I. Siddique, A. K. Karmaker, C. A. Macana, D. Dong, H. R. Pota, Bi-level planning and scheduling of electric vehicle charging stations for peak shaving and congestion management in low voltage distribution networks, Computers and Electrical Engineering 102 (2022) 108235. doi:10.1016/j.compeleceng.2022.108235. [22] M. Tostado-Véliz, A. Rezaee Jordehi, Y. Zhou, S. A. Mansouri, F. Jurado, Best-case-aware planning of photovoltaic-battery systems for multi-mode charging stations, Renewable Energy 225 (2024) 120300. doi:https://doi.org/10.1016/j.renene.2024.120300. 62 URL https://w.sciencedirect.com/science/article/pii/ S0960148124003653 [23] Y. Zheng, Y. Chen, Q. Yang, A stackelberg-based competition model for optimal participation of electric vehicle load aggrega- tors in demand response programs, Energy 315 (2025) 134414. doi:https://doi.org/10.1016/j.energy.2025.134414. URL https://w.sciencedirect.com/science/article/pii/ S0360544225000568 [24] R. Cheng, G. Liu, W. Liu, Q. Shi, Q. Chen, A personalized network-aware price design for electric vehicle charging using a sequential stackelberg game, Applied Energy 381 (2025) 125145. doi:https://doi.org/10.1016/j.apenergy.2024.125145. URL https://w.sciencedirect.com/science/article/pii/ S0306261924025297 [25] S. Bae, B. Kulcsár, S. Gros, Personalized dynamic pricing policy for electric vehicles: Reinforcement learning approach, Transporta- tion Research Part C: Emerging Technologies 161 (2024) 104540. doi:https://doi.org/10.1016/j.trc.2024.104540. URL https://w.sciencedirect.com/science/article/pii/ S0968090X24000615 [26] D. Qiao, G. Wang, M. Xu, Fast-charging station location problem: A two-phase approach with mathematical program 63 with equilibrium constraints considering charging choice be- haviour, Sustainable Cities and Society 96 (2023) 104678. doi:https://doi.org/10.1016/j.scs.2023.104678. URL https://w.sciencedirect.com/science/article/pii/ S2210670723002895 [27] T. Unterluggauer, J. Rich, P. B. Andersen, S. Hashemi, Electric vehicle charging infrastructure planning for integrated transportation and power distribution networks: A review, eTransportation 12 (2022) 100163. doi:https://doi.org/10.1016/j.etran.2022.100163. URL https://w.sciencedirect.com/science/article/pii/ S2590116822000091 [28] O. Sadeghian, A. Oshnoei, B. Mohammadi-ivatloo, V. Vahidi- nasab, A. Anvari-Moghaddam, A comprehensive review on elec- tric vehicles smart charging: Solutions, strategies, technologies, and challenges, Journal of Energy Storage 54 (2022) 105241. doi:https://doi.org/10.1016/j.est.2022.105241. URL https://w.sciencedirect.com/science/article/pii/ S2352152X22012403 [29] A. Dolgui, S. Kovalev, M. Y. Kovalyov, Scheduling electric vehicle regular charging tasks: A review of deterministic models, Euro- pean Journal of Operational Research 325 (2) (2025) 221–232. doi:https://doi.org/10.1016/j.ejor.2024.11.044. 64 URL https://w.sciencedirect.com/science/article/pii/ S0377221724009354 [30] S. LaMonaca, L. Ryan, The state of play in electric vehicle charging services – a review of infrastructure provision, players, and policies, Renewable and Sustainable Energy Reviews 154 (2022) 111733. doi:https://doi.org/10.1016/j.rser.2021.111733. URL https://w.sciencedirect.com/science/article/pii/ S1364032121010066 [31] H. Liu, D. Z. Wang, Locating multiple types of charging facili- ties for battery electric vehicles, Transportation Research Part B: Methodological 103 (2017) 30–55, green Urban Transportation. doi:https://doi.org/10.1016/j.trb.2017.01.005. URL https://w.sciencedirect.com/science/article/pii/ S0191261516305641 [32] C. Bian, H. Li, F. Wallin, A. Avelin, L. Lin, Z. Yu, Find- ing the optimal location for public charging stations – a gis-based milp approach, Energy Procedia 158 (2019) 6582– 6588, innovative Solutions for Energy Transitions. doi:https: //doi.org/10.1016/j.egypro.2019.01.071. URL https://w.sciencedirect.com/science/article/pii/ S1876610219300803 65 [33] Z.-Q. Luo, J.-S. Pang, D. Ralph, Mathematical Programs with Equi- librium Constraints, Cambridge University Press, 1996. [34] H. Obeid, A. T. Ozturk, W. Zeng, S. J. Moura, Learn- ing and optimizing charging behavior at pev charging sta- tions:Randomized pricing experiments, and joint power and price optimization, Applied Energy 351 (2023) 121862. doi:https://doi.org/10.1016/j.apenergy.2023.121862. URL https://w.sciencedirect.com/science/article/pii/ S0306261923012266 [35] G. Zhou, Q. Dong, Y. Zhao, H. Wang, L. Jian, Y. Jia, Bilevel optimization approach to fast charging station planning in electri- fied transportation networks, Applied Energy 350 (2023) 121718. doi:https://doi.org/10.1016/j.apenergy.2023.121718. URL https://w.sciencedirect.com/science/article/pii/ S0306261923010826 [36] S. Lamontagne, M. Carvalho, E. Frejinger, B. Gendron, M. F. Anjos, R. Atallah, Optimising electric vehicle charging station placement using advanced discrete choice models, INFORMS Journal on Computing 35 (5) (2023) 909–1213, published online May 25, 2023. doi:10.1287/ ijoc.2022.0185. [37] A. Najafi-Ghalelou, M. Khorasany, R. Razzaghi, Stochastic two-stage coordination of electric vehicles in distribution networks: A multi- 66 follower bi-level approach, Journal of Cleaner Production 414 (2023) 137610. doi:https://doi.org/10.1016/j.jclepro.2023.137610. URL https://w.sciencedirect.com/science/article/pii/ S0959652623017687 [38] B. Li, J. Li, Z. Liu, H. Su, Network equilibrium of a transportation-power distribution coupled system: A stackel- berg–nash game model, Renewable Energy 241 (2025) 122250. doi:https://doi.org/10.1016/j.renene.2024.122250. URL https://w.sciencedirect.com/science/article/pii/ S0960148124023188 [39] J.-M. Lasry, P.-L. Lions, Mean field games, Japanese Journal of Math- ematics 2 (1) (2007) 229–260. doi:10.1007/s11537-007-0657-8. [40] M. Shariatzadeh, M. A. Lopes, C. Henggeler Antunes, Electric vehicle users’ charging behavior: A review of influential factors, meth- ods and modeling approaches, Applied Energy 396 (2025) 126167. doi:https://doi.org/10.1016/j.apenergy.2025.126167. URL https://w.sciencedirect.com/science/article/pii/ S0306261925008979 [41] Z. Liang, T. Qian, M. Korkali, R. Glatt, Q. Hu, A vehicle-to-grid plan- ning framework incorporating electric vehicle user equilibrium and dis- tribution network flexibility enhancement, Applied Energy 376 (2024) 67 124231. doi:https://doi.org/10.1016/j.apenergy.2024.124231. URL https://w.sciencedirect.com/science/article/pii/ S0306261924016143 [42] Z. Sun, Y. He, H. Wu, A. Y. Wu, Bi-level planning of elec- tric vehicle charging stations considering charging demand: A nash bargaining game approach, Energy 332 (2025) 137137. doi:https://doi.org/10.1016/j.energy.2025.137137. URL https://w.sciencedirect.com/science/article/pii/ S0360544225027793 [43] M. A. Kazemi, M. Sedighizadeh, M. J. Mirzaei, O. Homaee, Optimal siting and sizing of distribution system operator owned EV parking lots, Applied Energy 179 (2016) 1176–1184. doi:10.1016/j.apenergy. 2016.06.125. [44] Z. Yi, J. Smart, A framework for integrated dispatching and charg- ing management of an autonomous electric vehicle ride-hailing fleet, Transportation Research Part D: Transport and Environment 95 (2021) 102822. doi:10.1016/j.trd.2021.102822. [45] J. Wang, H. D. Kaushik, R. A. Jacob, J. Zhang, Spatiotemporal plan- ning of electric vehicle charging infrastructure: Demand estimation and grid-aware optimization under uncertainty, iScience 28 (9) (2025) 113368. doi:https://doi.org/10.1016/j.isci.2025.113368. 68 URL https://w.sciencedirect.com/science/article/pii/ S2589004225016293 [46] C. Toregas, R. Swain, C. ReVelle, L. Bergman, The location of emer- gency service facilities, Operations Research 19 (6) (1971) 1363–1373. doi:10.1287/opre.19.6.1363. [47] K. Yazır, A. Karasan, İhsan Kaya, Optimization and prioritization of electric vehicle charging locations for intercity transport: A dual-phase approach, Engineering Applications of Artificial Intelligence 159 (2025) 111786. doi:https://doi.org/10.1016/j.engappai.2025.111786. URL https://w.sciencedirect.com/science/article/pii/ S0952197625017889 [48] S. L. Hakimi, Optimum locations of switching centers and the absolute centers and medians of a graph, Operations Research 12 (3) (1964) 450–459. doi:10.1287/opre.12.3.450. [49] F. J. Faustino, J. C. Lopes, J. D. Melo, T. Sousa, A. Padilha-Feltrin, J. A. Brito, C. O. Garcia, Identifying charging zones to allocate public charging stations for electric vehicles, Energy 283 (2023) 128436. doi:https://doi.org/10.1016/j.energy.2023.128436. URL https://w.sciencedirect.com/science/article/pii/ S0360544223018303 [50] H. Ljósheim, S. Jenkins, K. Searle, J. Wolff, Optimal placement of 69 electric vehicle slow-charging stations: A continuous facility location problem under uncertainty, Computers & Operations Research 185 (2026) 107289. doi:https://doi.org/10.1016/j.cor.2025.107289. URL https://w.sciencedirect.com/science/article/pii/ S0305054825003181 [51] M. J. Kłos, G. Sierpiński, Siting of electric vehicle charging sta- tions method addressing area potential and increasing their ac- cessibility, Journal of Transport Geography 109 (2023) 103601. doi:https://doi.org/10.1016/j.jtrangeo.2023.103601. URL https://w.sciencedirect.com/science/article/pii/ S096669232300073X [52] S. L. Hakimi, Optimum distribution of switching centers in a communi- cation network and some related graph theoretic problems, Operations Research 13 (3) (1965) 462–475. doi:10.1287/opre.13.3.462. [53] M. Kchaou-Boujelben, Charging station location problem: A comprehensive review on models and solution approaches, Trans- portation Research Part C: Emerging Technologies 132 (2021) 103376. doi:https://doi.org/10.1016/j.trc.2021.103376. URL https://w.sciencedirect.com/science/article/pii/ S0968090X21003776 [54] B. D. Chung, S. Park, C. Kwon, Equitable distribution of recharging stations for electric vehicles, Socio-Economic Planning Sciences 63 70 (2018) 1–11. doi:https://doi.org/10.1016/j.seps.2017.06.002. URL https://w.sciencedirect.com/science/article/pii/ S0038012116302579 [55] C. M. S. Lezcano, C. D. Harper, D. Nock, Siting for demand and equity: Optimizing level 2 electric vehicle charger place- ment, Journal of Transport Geography 128 (2025) 104369. doi:https://doi.org/10.1016/j.jtrangeo.2025.104369. URL https://w.sciencedirect.com/science/article/pii/ S0966692325002601 [56] M. Kuby, S. Lim, The flow-refueling location problem for alternative- fuel vehicles, Socio-Economic Planning Sciences 39 (2) (2005) 125–145. doi:10.1016/j.seps.2004.03.001. [57] M. Ghamami, A. Zockaie, Y. M. Nie, A general corri- dor model for designing plug-in electric vehicle charging in- frastructure to support intercity travel, Transportation Re- search Part C: Emerging Technologies 68 (2016) 389–402. doi:https://doi.org/10.1016/j.trc.2016.04.016. URL https://w.sciencedirect.com/science/article/pii/ S0968090X16300298 [58] F. Xie, Z. Lin, Integrated u.s. nationwide corridor charg- ing infrastructure planning for mass electrification of inter- city trips, Applied Energy 298 (2021) 117142. doi:https: 71 //doi.org/10.1016/j.apenergy.2021.117142. URL https://w.sciencedirect.com/science/article/pii/ S0306261921005821 [59] I. Capar, M. Kuby, V. J. Leon, Y.-J. Tsai, An arc cover–path cover formulation and strategic analysis of alternative-fuel station locations, European Journal of Operational Research 227 (1) (2013) 142–151. doi:10.1016/j.ejor.2012.11.033. [60] M. Miralinaghi, Y. Lou, B. B. Keskin, A. Zarrinmehr, R. Shabanpour, Refueling station location problem with traffic deviation considering route choice and demand uncertainty, International Journal of Hydro- gen Energy 42 (5) (2017) 3335–3351. doi:10.1016/j.ijhydene.2016. 12.137. [61] T. Dimatulac, H. Maoh, R. Carriveau, Electricity demand assessment and charging infrastructure planning for long-haul electric vehicle operations in ontario, canada, Transport Policy 178 (2026) 103970. doi:https://doi.org/10.1016/j.tranpol.2025.103970. URL https://w.sciencedirect.com/science/article/pii/ S0967070X2500513X [62] B. Borlaug, V. Caristo, F. Ouren, F. Yang, E. Wood, L. Rober- son, Economics of electric vehicle corridor fast charging in the united states, Advances in Applied Energy 21 (2026) 100257. 72 doi:https://doi.org/10.1016/j.adapen.2025.100257. URL https://w.sciencedirect.com/science/article/pii/ S2666792425000514 [63] D. Speth, P. Plötz, M. Wietschel, An optimal capacity-constrained fast charging network for battery electric trucks in germany, Trans- portation Research Part A: Policy and Practice 193 (2025) 104383. doi:https://doi.org/10.1016/j.tra.2025.104383. URL https://w.sciencedirect.com/science/article/pii/ S0965856425000114 [64] V. Anadón Martínez, A. Sumper, A. E. Saldaña-Gonzalez, V. Gadelha, Planning fast-charging stations along highways us- ing probability distribution functions and traffic data, Sustain- able Energy Technologies and Assessments 82 (2025) 104547. doi:https://doi.org/10.1016/j.seta.2025.104547. URL https://w.sciencedirect.com/science/article/pii/ S2213138825003789 [65] K. Nareshkumar, D. Das, Optimal location and sizing of elec- tric vehicles charging stations and renewable sources in a coupled transportation-power distribution network, Renew- able and Sustainable Energy Reviews 203 (2024) 114767. doi:https://doi.org/10.1016/j.rser.2024.114767. 73 URL https://w.sciencedirect.com/science/article/pii/ S1364032124004933 [66] Q. Ye, P. Bansal, B. Adey, Multi-period charging infras- tructure planning under uncertainty: Challenges and oppor- tunities, Sustainable Cities and Society 116 (2024) 105908. doi:https://doi.org/10.1016/j.scs.2024.105908. URL https://w.sciencedirect.com/science/article/pii/ S2210670724007327 [67] K. Jain, V. Vazirani, Approximation algorithms for metric facility location and k-median problems using the primal-dual schema and lagrangian relaxation, Journal of The ACM - JACM 48 (03 2001). doi:10.1145/375827.375845. [68] Y. Li, J. Luo, C.-Y. Chow, K.-L. Chan, Y. Ding, F. Zhang, Grow- ing the charging station network for electric vehicles with trajectory data analytics, in: 2015 IEEE 31st International Conference on Data Engineering, 2015, p. 1376–1387. doi:10.1109/ICDE.2015.7113384. [69] M. Song, L. Cheng, M. Du, C. Sun, J. Ma, H. Ge, Charging sta- tion location problem for maximizing the space-time-electricity accessibility:A lagrangian relaxation-based decomposition scheme, Expert Systems with Applications 222 (2023) 119801. doi:https://doi.org/10.1016/j.eswa.2023.119801. 74 URL https://w.sciencedirect.com/science/article/pii/ S0957417423003020 [70] L. Yin, Y. Zhang, Particle swarm optimization based on data driven for ev charging station siting, Energy 310 (2024) 133197. doi:https://doi.org/10.1016/j.energy.2024.133197. URL https://w.sciencedirect.com/science/article/pii/ S0360544224029724 [71] G. Zhou, Z. Zhu, S. Luo, Location optimization of elec- tric vehicle charging stations: Based on cost model and ge- netic algorithm, Energy 247 (2022) 123437. doi:https: //doi.org/10.1016/j.energy.2022.123437. URL https://w.sciencedirect.com/science/article/pii/ S0360544222003401 [72] I. Frade, A. Ribeiro, G. Gonçalves, A. Pais Antunes, Optimal location of charging stations for electric vehicles in a neighborhood in lisbon, portugal, TRR 2252 (12 2011). doi:10.3141/2252-12. [73] A. H. K. Babar, A. Yousaf, Optimization of charging infrastructure planning for plug-in electric vehicles based on a dynamic program- ming model, Transportation Planning and Technology 45 (1) (2022) 59–75. arXiv:https://doi.org/10.1080/03081060.2021.2017207, doi:10.1080/03081060.2021.2017207. URL https://doi.org/10.1080/03081060.2021.2017207 75 [74] Z. Bi, G. A. Keoleian, T. Ersal, Wireless charger deploy- ment for an electric bus network: A multi-objective life cycle optimization, Applied Energy 225 (2018) 1090–1101. doi:https://doi.org/10.1016/j.apenergy.2018.05.070. URL https://w.sciencedirect.com/science/article/pii/ S030626191830789X [75] H. Pourvaziri, H. Sarhadi, N. Azad, H. Afshari, M. Taghavi, Planning of electric vehicle charging stations: An integrated deep learning and queueing theory approach, Transportation Research Part E: Logistics and Transportation Review 186 (2024) 103568. doi:https://doi.org/10.1016/j.tre.2024.103568. URL https://w.sciencedirect.com/science/article/pii/ S1366554524001595 [76] D. Xiao, S. An, H. Cai, J. Wang, H. Cai, An optimization model for electric vehicle charging infrastructure planning considering queuing behavior with finite queue length, Journal of Energy Storage 29 (2020) 101317. doi:https://doi.org/10.1016/j.est.2020.101317. URL https://w.sciencedirect.com/science/article/pii/ S2352152X19309053 [77] O. B. Kınay, F. Gzara, S. A. Alumur, Charging station location and sizing for electric vehicles under congestion, Transportation Science 76 57 (6) (2023) 1433–1451. doi:10.1287/trsc.2021.0494. URL https://doi.org/10.1287/trsc.2021.0494 [78] W. Whitt, Approximations for the GI/G/m queue, Production and Operations Management 2 (2) (1993) 114–161. doi:10.1111/j. 1937-5956.1993.tb00094.x. [79] H. Zhang, S. J. Moura, Z. Hu, Y. Song, Pev fast-charging station sit- ing and sizing on coupled transportation and power networks, IEEE Transactions on Smart Grid 9 (4) (2018) 2595–2605. doi:10.1109/ TSG.2016.2614939. [80] I. S. Bayram, A. Tajer, M. Abdallah, K. Qaraqe, Capacity planning frameworks for electric vehicle charging stations with multiclass cus- tomers, IEEE Transactions on Smart Grid 6 (4) (2015) 1934–1943. doi:10.1109/TSG.2015.2406532. [81] H. Pourvaziri, M. Taghavi, H. Sarhadi, H. Afshari, N. Azad, Multi- objective planning of electric vehicles charging stations by integrating drivers’ preferences and fairness considerations: A case study in halifax, canada, Computers & Industrial Engineering 201 (2025) 110886. doi:https://doi.org/10.1016/j.cie.2025.110886. URL https://w.sciencedirect.com/science/article/pii/ S0360835225000312 [82] J. Yang, J. Dong, L. Hu, A data-driven optimization-based approach 77 for siting and sizing of electric taxi charging stations, Transporta- tion Research Part C: Emerging Technologies 77 (2017) 462–477. doi:https://doi.org/10.1016/j.trc.2017.02.014. URL https://w.sciencedirect.com/science/article/pii/ S0968090X17300542 [83] Y. Meng, Y. Sun, D. Xie, M. Xiao, C. Yin, L. Zhao, A two-stage optimization framework for ev charging station planning considering investment cost and service satisfaction, Applied Energy 402 (2025) 126911. doi:https://doi.org/10.1016/j.apenergy.2025.126911. URL https://w.sciencedirect.com/science/article/pii/ S0306261925016411 [84] Y. S. Liu, M. Tayarani, F. You, H. O. Gao, Bayesian optimiza- tion for battery electric vehicle charging station placement by agent-based demand simulation, Applied Energy 375 (2024) 123975. doi:https://doi.org/10.1016/j.apenergy.2024.123975. URL https://w.sciencedirect.com/science/article/pii/ S0306261924013588 [85] M. E. Baran, F. F. Wu, Network reconfiguration in distribution systems for loss reduction and load balancing, IEEE Transactions on Power Delivery 4 (2) (1989) 1401–1407. doi:10.1109/61.25627. [86] M. Farivar, S. H. Low, Branch flow model: Relaxations and 78 convexification—part I, IEEE Transactions on Power Systems 28 (3) (2013) 2554–2564. doi:10.1109/TPWRS.2013.2255317. [87] V. S. M. B. K., P. Chakraborty, M. Pal, Planning of fast charging infrastructure for electric vehicles in a distribution system and prediction of dynamic price, International Jour- nal of Electrical Power & Energy Systems 155 (2024) 109502. doi:https://doi.org/10.1016/j.ijepes.2023.109502. URL https://w.sciencedirect.com/science/article/pii/ S0142061523005598 [88] H. Zhang, S. J. Moura, Z. Hu, Y. Song, Pev fast-charging station sit- ing and sizing on coupled transportation and power networks, IEEE Transactions on Smart Grid 9 (4) (2018) 2595–2605. doi:10.1109/ TSG.2016.2614939. [89] S. A. MirHassani, A. Khaleghi, F. Hooshmand, A two-stage stochas- tic programming model to locate capacitated EV-charging stations in urban areas under demand uncertainty, EURO Journal on Transporta- tion and Logistics 9 (4) (2020) 100025. doi:10.1016/j.ejtl.2020. 100025. [90] Y. Zhou, G. P. Ong, Q. Meng, H. Cui, Electric bus charging facility planning with uncertainties: Model formulation and algorithm design, Transportation Research Part C: Emerging Technologies 150 (2023) 79 104108. doi:https://doi.org/10.1016/j.trc.2023.104108. URL https://w.sciencedirect.com/science/article/pii/ S0968090X23000979 [91] R. Xie, W. Wei, M. E. Khodayar, J. Wang, S. Mei, Planning fully re- newable powered charging stations on highways: A data-driven robust optimization approach, IEEE Transactions on Transportation Electri- fication 4 (3) (2018) 817–830. doi:10.1109/TTE.2018.2843125. [92] B. Zhou, G. Chen, T. Huang, Q. Song, Y. Yuan, Planning PEV fast- charging stations using data-driven distributionally robust optimiza- tion approach based on phi-divergence, IEEE Transactions on Trans- portation Electrification 6 (1) (2020) 170–180. doi:10.1109/TTE. 2020.2973366. [93] S. Wang, A. Chen, P. Wang, C. Zhuge, Predicting electric vehicle charg- ing demand using a heterogeneous spatio-temporal graph convolutional network, Transportation Research Part C: Emerging Technologies 153 (2023) 104205. doi:https://doi.org/10.1016/j.trc.2023.104205. URL https://w.sciencedirect.com/science/article/pii/ S0968090X23001948 [94] Q. Jiang, N. Zhang, B. Yueshuai He, C. Lee, J. Ma, Large-scale public charging demand prediction with a scenario- and activity-based approach, Transportation Research Part A: Policy and Practice 179 80 (2024) 103935. doi:https://doi.org/10.1016/j.tra.2023.103935. URL https://w.sciencedirect.com/science/article/pii/ S0965856423003555 [95] A. Orzechowski, L. Lugosch, H. Shu, R. Yang, W. Li, B. H. Meyer, A data-driven framework for medium-term electric vehicle charging demand forecasting, Energy and AI 14 (2023) 100267. doi:https://doi.org/10.1016/j.egyai.2023.100267. URL https://w.sciencedirect.com/science/article/pii/ S2666546823000393 [96] D. Bertsimas, V. Gupta, N. Kallus, Data-driven robust optimization, Mathematical Programming 167 (2) (2018) 235–292. doi:10.1007/ s10107-017-1125-8. [97] A. N. Elmachtoub, P. Grigas, Smart “predict, then optimize”, Manage- ment Science 68 (1) (2022) 9–26. doi:10.1287/mnsc.2020.3922. [98] C. Li, Z. Dong, G. Chen, B. Zhou, J. Zhang, X. Yu, Data-driven plan- ning of electric vehicle charging infrastructure: A case study of sydney, australia, IEEE Transactions on Smart Grid 12 (4) (2021) 3289–3304. doi:10.1109/TSG.2021.3054763. [99] B. He, B. Yang, Y. Han, Y. Zhou, Y. Hu, H. Shu, S. Su, J. Yang, Y. Huang, J. Li, L. Jiang, H. Li, Optimal evcs plan- ning via spatial-temporal distribution of charging demand fore- 81 casting and traffic-grid coupling, Energy 313 (2024) 133885. doi:https://doi.org/10.1016/j.energy.2024.133885. URL https://w.sciencedirect.com/science/article/pii/ S0360544224036636 [100] H. Zhang, S. J. Moura, Z. Hu, W. Qi, Y. Song, A second-order cone pro- gramming model for planning pev fast-charging stations, IEEE Trans- actions on Power Systems 33 (3) (2018) 2763–2777. doi:10.1109/ TPWRS.2017.2754940. [101] H. Zhang, S. J. Moura, Z. Hu, W. Qi, Y. Song, Joint pev charging net- work and distributed pv generation planning based on accelerated gen- eralized benders decomposition, IEEE Transactions on Transportation Electrification 4 (3) (2018) 789–803. doi:10.1109/TTE.2018.2847244. [102] H. Shi, H. Xiong, W. Gan, Y. Lin, C. Guo, Fully dis- tributed planning method for coordinated distribution and urban transportation networks considering three-phase un- balance mitigation, Applied Energy 377 (2025) 124449. doi:https://doi.org/10.1016/j.apenergy.2024.124449. URL https://w.sciencedirect.com/science/article/pii/ S0306261924018324 [103] S. Faridimehr, S. Venkatachalam, R. B. Chinnam, A stochastic pro- gramming approach for electric vehicle charging network design, IEEE 82 Transactions on Intelligent Transportation Systems 20 (5) (2019) 1870– 1882. doi:10.1109/TITS.2018.2841391. [104] M. Kabli, M. A. Quddus, S. G. Nurre, M. Marufuzza- man, J. M. Usher, A stochastic programming approach for electric vehicle charging station expansion plans, Interna- tional Journal of Production Economics 220 (2020) 107461. doi:https://doi.org/10.1016/j.ijpe.2019.07.034. URL https://w.sciencedirect.com/science/article/pii/ S0925527319302713 [105] S. Wang, Q. Tan, Q. Jiang, et al., Stochastic carbon-aware planning of renewable dgs and ev charging stations with demand flexibility in smart urban grids, Scientific Reports 15 (2025) 34433. doi:10.1038/ s41598-025-17504-1. URL https://doi.org/10.1038/s41598-025-17504-1 [106] B. Zhou, G. Chen, T. Huang, Q. Song, Y. Yuan, Planning pev fast- charging stations using data-driven distributionally robust optimiza- tion approach based on φ-divergence, IEEE Transactions on Trans- portation Electrification 6 (1) (2020) 170–180. doi:10.1109/TTE. 2020.2971825. [107] A. Zhang, J. E. Kang, C. Kwon, Incorporating demand dy- namics in multi-period capacitated fast-charging location plan- ning for electric vehicles, Transportation Research Part B: 83 Methodological 103 (2017) 5–29, green Urban Transportation. doi:https://doi.org/10.1016/j.trb.2017.04.016. URL https://w.sciencedirect.com/science/article/pii/ S0191261516304349 [108] S. Li, Y. Huang, S. J. Mason, A multi-period optimization model for the deployment of public electric vehicle charging stations on network, Transportation Research Part C: Emerging Technologies 65 (2016) 128–143. doi:https://doi.org/10.1016/j.trc.2016.01.008. URL https://w.sciencedirect.com/science/article/pii/ S0968090X16000267 [109] M. F. Anjos, B. Gendron, M. Joyce-Moniz, Increasing elec- tric vehicle adoption through the optimal deployment of fast- charging stations for local and long-distance travel, Euro- pean Journal of Operational Research 285 (1) (2020) 263–278. doi:https://doi.org/10.1016/j.ejor.2020.01.055. URL https://w.sciencedirect.com/science/article/pii/ S0377221720300928 [110] J. R. Birge, F. Louveaux, Introduction to Stochastic Programming, 2nd Edition, Springer, 2011. [111] A. A. Kadri, R. Perrouault, M. K. Boujelben, C. Gicquel, A multi- stage stochastic integer programming approach for locating electric 84 vehicle charging stations, Computers & Operations Research 117 (2020) 104888. doi:https://doi.org/10.1016/j.cor.2020.104888. URL https://w.sciencedirect.com/science/article/pii/ S0305054820300058 [112] M. A. Quddus, M. Kabli, M. Marufuzzaman, Modeling electric vehicle charging station expansion with an integration of renew- able energy and vehicle-to-grid sources, Transportation Research Part E: Logistics and Transportation Review 128 (2019) 251–279. doi:https://doi.org/10.1016/j.tre.2019.06.006. URL https://w.sciencedirect.com/science/article/pii/ S136655451830485X [113] K. Schroer, R. Ahadi, W. Ketter, T. Y. Lee, Data-driven planning of large-scale electric vehicle charging hubs using deep reinforcement learning, Transportation Research Part C: Emerging Technologies 177 (2025) 105126. doi:https://doi.org/10.1016/j.trc.2025.105126. URL https://w.sciencedirect.com/science/article/pii/ S0968090X25001305 [114] M. Schneider, A. Stenger, D. Goeke, The electric vehicle-routing prob- lem with time windows and recharging stations, Transportation Science 48 (4) (2014) 500–520. doi:10.1287/trsc.2013.0490. [115] M. Keskin, B. Çatay, Partial recharge strategies for the elec- tric vehicle routing problem with time windows, Transporta- 85 tion Research Part C: Emerging Technologies 65 (2016) 111–127. doi:https://doi.org/10.1016/j.trc.2016.01.013. URL https://w.sciencedirect.com/science/article/pii/ S0968090X16000322 [116] W. Wang, J. Zhao, Partial linear recharging strategy for the electric fleet size and mix vehicle routing problem with time windows and recharging stations, European Journal of Operational Research 308 (2) (2023) 929–948. doi:https: //doi.org/10.1016/j.ejor.2022.12.011. URL https://w.sciencedirect.com/science/article/pii/ S0377221722009389 [117] A. Montoya, C. Guéret, J. E. Mendoza, J. G. Villegas, The electric vehicle routing problem with nonlinear charging func- tion, Transportation Research Part B: Methodological 103 (2017) 87–110, green Urban Transportation. doi:https: //doi.org/10.1016/j.trb.2017.02.004. URL https://w.sciencedirect.com/science/article/pii/ S0191261516304556 [118] A. Froger, O. Jabali, J. E. Mendoza, G. Laporte, The electric vehicle routing problem with capacitated charging stations, Transportation Science 56 (2) (2022) 460–482. doi:10.1287/trsc.2021.1111. [119] G. Hiermann, R. F. Hartl, J. Puchinger, T. Vidal, Routing a 86 mix of conventional, plug-in hybrid, and electric vehicles, Euro- pean Journal of Operational Research 272 (1) (2019) 235–248. doi:https://doi.org/10.1016/j.ejor.2018.06.025. URL https://w.sciencedirect.com/science/article/pii/ S0377221718305605 [120] W. Wang, Y. Adulyasak, J.-F. Cordeau, G. He, The heterogeneous- fleet electric vehicle routing problem with nonlinear charging functions, Transportation Research Part C: Emerging Technologies 170 (2025) 104932. doi:https://doi.org/10.1016/j.trc.2024.104932. URL https://w.sciencedirect.com/science/article/pii/ S0968090X24004534 [121] I. Kucukoglu, R. Dewil, D. Cattrysse, The electric vehicle routing prob- lem and its variations: A literature review, Computers & Industrial Engineering 161 (2021) 107650. doi:10.1016/j.cie.2021.107650. [122] Y. Zhou, Q. Meng, G. P. Ong, H. Wang, Electric bus charging schedul- ing on a bus network, Transportation Research Part C: Emerging Tech- nologies 161 (2024) 104553. doi:10.1016/j.trc.2024.104553. [123] M. Duan, F. Liao, G. Qi, W. Guan, Integrated optimization of elec- tric bus scheduling and charging planning incorporating flexible charg- ing and timetable shifting strategies, Transportation Research Part C: Emerging Technologies 152 (2023) 104175. doi:10.1016/j.trc.2023. 104175. 87 [124] S. S. G. Perumal, R. M. Lusby, J. Larsen, Electric bus planning & scheduling: A review of related problems and methodologies, European Journal of Operational Research 301 (2) (2022) 395–413. doi:10.1016/ j.ejor.2021.10.058. [125] W. Wu, Y. Lin, R. Liu, Y. Li, Y. Zhang, C. Ma, Online ev charge scheduling based on time-of-use pricing and peak load minimization: Properties and efficient algorithms, IEEE Transactions on Intelligent Transportation Systems 23 (1) (2022) 572–586. doi:10.1109/TITS. 2020.3014088. [126] D. Dreucci, Y. Yu, G. R. Chandra Mouli, A. Shekhar, P. Bauer, Cen- tralised distribution grid congestion management through ev charging control considering fairness and priority, Applied Energy 384 (2025) 125417. doi:https://doi.org/10.1016/j.apenergy.2025.125417. URL https://w.sciencedirect.com/science/article/pii/ S0306261925001473 [127] Y. Xu, F. Pan, L. Tong, Dynamic scheduling for charging electric vehi- cles: A priority rule, IEEE Transactions on Automatic Control 61 (12) (2016) 4094–4099. doi:10.1109/TAC.2016.2541305. [128] N. Chen, C. Kurniawan, Y. Nakahira, L. Chen, S. H. Low, Smoothed least-laxity-first algorithm for electric vehicle charging: Online decision and performance analysis with resource augmentation, IEEE Transac- 88 tions on Smart Grid 13 (3) (2022) 2209–2217. doi:10.1109/TSG.2021. 3138615. [129] S. P. Rao, T. S. Olusegun, P. Ranganathan, U. Kose, N. Gov- eas, Vehicle-to-grid technology: Opportunities, challenges, and future prospects for sustainable transportation, Journal of Energy Storage 110 (2025) 114927. doi:https://doi.org/10.1016/j.est.2024.114927. URL https://w.sciencedirect.com/science/article/pii/ S2352152X24045134 [130] H. Liu, K. Huang, N. Wang, J. Qi, Q. Wu, S. Ma, C. Li, Optimal dispatch for participation of electric vehicles in fre- quency regulation based on area control error and area regulation requirement, Applied Energy 240 (2019) 46–55. doi:https://doi.org/10.1016/j.apenergy.2019.02.044. URL https://w.sciencedirect.com/science/article/pii/ S0306261919303411 [131] M. Mohammadian, K. Baker, C. Crozier, Spatial arbitrage through bidirectional electric vehicle charging with deliv- ery fleets, Applied Energy 381 (2025) 125036. doi:https: //doi.org/10.1016/j.apenergy.2024.125036. URL https://w.sciencedirect.com/science/article/pii/ S0306261924024206 [132] S. Bashash, S. J. Moura, J. C. Forman, H. K. Fathy, Plug-in hybrid 89 electric vehicle charge pattern optimization for energy cost and battery longevity, Journal of Power Sources 196 (1) (2011) 541–549. doi:https://doi.org/10.1016/j.jpowsour.2010.07.001. URL https://w.sciencedirect.com/science/article/pii/ S0378775310011390 [133] C.-F. Lu, G.-P. Liu, Y. Yu, J. Cui, A coordinated model predictive control-based approach for vehicle-to-grid scheduling considering range anxiety and battery degradation, IEEE Transactions on Transporta- tion Electrification 11 (2) (2025) 5688–5699. doi:10.1109/TTE.2024. 3488075. [134] F. Liu, R. Sen, J. P. Talusan, A. Pettet, A. Kandel, Y. Suzue, A. Mukhopadhyay, A. Dubey, Reinforcement learning-based approach for vehicle-to-building charging with heterogeneous agents and long term rewards, in: Proceedings of the 24th International Conference on Autonomous Agents and Multiagent Systems, AAMAS ’25, Inter- national Foundation for Autonomous Agents and Multiagent Systems, Richland, SC, 2025, p. 1345–1353. [135] R. Sen, Y. Zhang, F. Liu, J. P. Talusan, A. Pettet, Y. Suzue, A. Mukhopadhyay, A. Dubey, Online decision-making under un- certainty for vehicle-to-building systems, in: Proceedings of the ACM/IEEE 16th International Conference on Cyber-Physical Sys- tems (with CPS-IoT Week 2025), ICCPS ’25, ACM, 2025, p. 1–12. 90 doi:10.1145/3716550.3722024. URL http://dx.doi.org/10.1145/3716550.3722024 [136] R. Sen, F. Liu, J. P. Talusan, A. Pettet, Y. Suzue, A. Mukhopad- hyay, A. Dubey, P-v2b: A neuro-symbolic framework for leveraging user persistence in vehicle-to-building charging, in: Proceedings of the HSCC/ICCPS 2026: 29th ACM International Conference on Hybrid Systems: Computation and Control and 17th ACM/IEEE Interna- tional Conference on Cyber-Physical Systems, 2026. [137] F. Ahmed, S. Ahmad, M. T. Rahman, M. R. Hazari, R. Faiz, T. Ahmed, M. Karimi, A holistic review of electric vehicle charging impacts on power distribution networks: Technical challenges, smart mitigation strategies and future directions, Applied Energy 402 (2026) 126961. doi:https://doi.org/10.1016/j.apenergy.2025.126961. URL https://w.sciencedirect.com/science/article/pii/ S0306261925016915 [138] R. Sen, A. Sivagnanam, A. Laszka, A. Mukhopadhyay, A. Dubey, Grid-aware charging and operational optimization for mixed-fleet pub- lic transit, in: 2024 IEEE 27th International Conference on Intel- ligent Transportation Systems (ITSC), 2024, p. 4172–4179. doi: 10.1109/ITSC58415.2024.10920218. [139] H. Saber, H. Ranjbar, E. Hajipour, M. Shahidehpour, Two-stage coor- dination scheme for multiple ev charging stations connected to an ex- 91 clusive dc feeder considering grid-tie converter limitation, IEEE Trans- actions on Transportation Electrification 10 (4) (2024) 8213–8223. doi:10.1109/TTE.2024.3357217. [140] X. Chen, X. Wang, M. Shahidehpour, L. Affolabi, Z. Lu, K. Li, Dis- tributed peer-to-peer coordination of hierarchical three-phase energy transactions among electric vehicle charging stations in constrained power distribution and urban transportation networks, IEEE Trans- actions on Transportation Electrification 10 (2) (2024) 4407–4420. doi:10.1109/TTE.2023.3315398. [141] X. Hu, H. Li, C. Xie, Optimal charging scheduling of an electric bus fleet with photovoltaic-storage-charging stations, Applied Energy 390 (2025) 125714. doi:https://doi.org/10.1016/j.apenergy.2025. 125714. URL https://w.sciencedirect.com/science/article/pii/ S0306261925004441 [142] R. Khezri, D. Steen, E. Wikner, L. A. Tuan, Optimal v2g schedul- ing of an ev with calendar and cycle aging of battery: An milp ap- proach, IEEE Transactions on Transportation Electrification 10 (4) (2024) 10497–10507. doi:10.1109/TTE.2024.3384293. [143] J. Zhang, M. Cui, B. Li, H. Fang, Y. He, Fast solving method based on linearized equations of branch power flow for coordinated charging of 92 evs (evcc), IEEE Transactions on Vehicular Technology 68 (5) (2019) 4404–4418. doi:10.1109/TVT.2019.2904464. [144] W. u. Rehman, J. W. Kimball, R. Bo, Multilayered energy manage- ment framework for extreme fast charging stations considering de- mand charges, battery degradation, and forecast uncertainties, IEEE Transactions on Transportation Electrification 10 (1) (2024) 760–776. doi:10.1109/TTE.2023.3275110. [145] F. Wu, R. Sioshansi, A two-stage stochastic optimization model for scheduling electric vehicle charging loads to relieve distribution-system constraints, Transportation Research Part B: Methodological 102 (2017) 55–82. doi:https://doi.org/10.1016/j.trb.2017.05.002. URL https://w.sciencedirect.com/science/article/pii/ S0191261516305343 [146] K. Seddig, P. Jochem, W. Fichtner, Two-stage stochastic optimization for cost-minimal charging of electric vehicles at public charging stations with photovoltaics, Applied Energy 242 (2019) 769–781. doi:https://doi.org/10.1016/j.apenergy.2019.03.036. URL https://w.sciencedirect.com/science/article/pii/ S0306261919304301 [147] S. Rafique, M. Nizami, U. Irshad, M. Hossain, S. Mukhopad- hyay, A two-stage multi-objective stochastic optimization strat- 93 egy to minimize cost for electric bus depot operators, Jour- nal of Cleaner Production 332 (2022) 129856. doi:https: //doi.org/10.1016/j.jclepro.2021.129856. URL https://w.sciencedirect.com/science/article/pii/ S0959652621040282 [148] Y. Zheng, H. Yu, Z. Shao, L. Jian, Day-ahead bidding strategy for electric vehicle aggregator enabling multiple agent modes in uncertain electricity markets, Applied Energy 280 (2020) 115977. doi:https://doi.org/10.1016/j.apenergy.2020.115977. URL https://w.sciencedirect.com/science/article/pii/ S0306261920314276 [149] B. Kandpal, P. Pareek, A. Verma, A robust day-ahead scheduling strategy for ev charging stations in unbalanced distribution grid, Energy 249 (2022) 123737. doi:https: //doi.org/10.1016/j.energy.2022.123737. URL https://w.sciencedirect.com/science/article/pii/ S0360544222006405 [150] J. Li, Y. Xu, J. Zhang, C. Gao, H. Sun, Distributed ev scheduling in distribution networks with reserve market participation under am- biguous probability distribution, Applied Energy 383 (2025) 125269. doi:https://doi.org/10.1016/j.apenergy.2024.125269. 94 URL https://w.sciencedirect.com/science/article/pii/ S0306261924026539 [151] Á. Porras, R. Fernández-Blanco, J. M. Morales, S. Pineda, An efficient robust approach to the day-ahead operation of an aggregator of electric vehicles, IEEE Transactions on Smart Grid 11 (6) (2020) 4960–4970. doi:10.1109/TSG.2020.3004268. [152] M. Bagheri Tookanlou, S. A. Pourmousavi, M. Marzband, A three-layer joint distributionally robust chance-constrained framework for optimal day-ahead scheduling of e-mobility ecosystem, Applied Energy 331 (2023) 120402. doi:https: //doi.org/10.1016/j.apenergy.2022.120402. URL https://w.sciencedirect.com/science/article/pii/ S0306261922016592 [153] N. Sadeghianpourhamami, J. Deleu, C. Develder, Definition and eval- uation of model-free coordination of electrical vehicle charging with re- inforcement learning, IEEE Transactions on Smart Grid 11 (1) (2020) 203–214. doi:10.1109/TSG.2019.2920320. [154] J. Jin, Y. Xu, Z. Yang, Optimal deadline scheduling for electric vehicle charging with energy storage and random supply, Automatica 119 (2020) 109096. doi:https://doi.org/10.1016/j.automatica. 2020.109096. 95 URL https://w.sciencedirect.com/science/article/pii/ S0005109820302946 [155] F. Wamalwa, T. Raji, R. Wafula, C. Ekoh, On adaptive mpc development and application to a dc fast charger (dcfc) station with an embedded microgrid, Applied Energy 398 (2025) 126428. doi:https://doi.org/10.1016/j.apenergy.2025.126428. URL https://w.sciencedirect.com/science/article/pii/ S0306261925011584 [156] B. Škugor, L. Grden, J. Deur, Hierarchical model predic- tive control-based electric vehicle fleet charging manage- ment, Energy Conversion and Management 342 (2025) 120119. doi:https://doi.org/10.1016/j.enconman.2025.120119. URL https://w.sciencedirect.com/science/article/pii/ S0196890425006430 [157] Z. Wan, H. Li, H. He, D. Prokhorov, Model-free real-time ev charging scheduling based on deep reinforcement learning, IEEE Transactions on Smart Grid 10 (5) (2019) 5246–5257. doi:10.1109/TSG.2018. 2879572. [158] H. Li, G. Li, T. T. Lie, X. Li, K. Wang, B. Han, J. Xu, Constrained large-scale real-time ev scheduling based on re- current deep reinforcement learning, International Journal 96 of Electrical Power & Energy Systems 144 (2023) 108603. doi:https://doi.org/10.1016/j.ijepes.2022.108603. URL https://w.sciencedirect.com/science/article/pii/ S0142061522005993 [159] L. Cui, Q. Wang, H. Qu, M. Wang, Y. Wu, L. Ge, Dy- namic pricing for fast charging stations with deep reinforce- ment learning, Applied Energy 346 (2023) 121334. doi:https: //doi.org/10.1016/j.apenergy.2023.121334. URL https://w.sciencedirect.com/science/article/pii/ S0306261923006980 [160] Y. Yang, R. Tang, Cooperative multi-agent reinforcement learn- ing for grid-aware ev charging management with cross-site redirection, Energy and AI 23 (2026) 100664. doi:https: //doi.org/10.1016/j.egyai.2025.100664. URL https://w.sciencedirect.com/science/article/pii/ S266654682500196X [161] B. Khaki, C. Chu, R. Gadh, Hierarchical distributed framework for ev charging scheduling using exchange problem, Applied Energy 241 (2019) 461–471. doi:https://doi.org/10.1016/j.apenergy.2019. 03.008. URL https://w.sciencedirect.com/science/article/pii/ S0306261919304027 97 [162] D. McFadden, Conditional logit analysis of qualitative choice behavior, in: P. Zarembka (Ed.), Frontiers in Econometrics, Academic Press, New York, 1973, p. 105–142. [163] M. E. Ben-Akiva, S. R. Lerman, Discrete Choice Analysis: Theory and Application to Travel Demand, MIT Press, Cambridge, MA, 1985. [164] F. A. Bhat, G. Y. Tiwari, A. Verma, Preferences for pub- lic electric vehicle charging infrastructure locations: A dis- crete choice analysis, Transport Policy 149 (2024) 177–197. doi:https://doi.org/10.1016/j.tranpol.2024.02.004. URL https://w.sciencedirect.com/science/article/pii/ S0967070X24000398 [165] Y. Yang, E. Yao, Z. Yang, R. Zhang, Modeling the charg- ing and route choice behavior of bev drivers, Transportation Research Part C: Emerging Technologies 65 (2016) 190–204. doi:https://doi.org/10.1016/j.trc.2015.09.008. URL https://w.sciencedirect.com/science/article/pii/ S0968090X15003319 [166] L. Sica, A. Carboni, F. Deflorio, C. Botta, Electric vehicle recharg- ing options in urban areas: Discrete choice modeling to estimate user preference, Sustainable Cities and Society 121 (2025) 106175. doi:https://doi.org/10.1016/j.scs.2025.106175. 98 URL https://w.sciencedirect.com/science/article/pii/ S2210670725000538 [167] J. G. WARDROP, Road paper. some theoretical aspects of road traffic research., Proceedings of the Institution of Civil Engineers 1 (3) (1952) 325–362. arXiv:https://w.emerald.com/jpric/article-pdf/1/ 3/325/2704995/ipeds_1952_11259.pdf, doi:10.1680/ipeds.1952. 11259. URL https://doi.org/10.1680/ipeds.1952.11259 [168] K. Liu, Y. Liu, Stochastic user equilibrium based spatial- temporal distribution prediction of electric vehicle charg- ing load, Applied Energy 339 (2023) 120943. doi:https: //doi.org/10.1016/j.apenergy.2023.120943. URL https://w.sciencedirect.com/science/article/pii/ S0306261923003070 [169] Z. Xu, Y. Peng, G. Li, A. Chen, X. Liu, Range-constrained traffic assignment for electric vehicles under heterogeneous range anxiety, Transportation Research Part C: Emerging Technologies 158 (2024) 104419. doi:https://doi.org/10.1016/j.trc.2023.104419. URL https://w.sciencedirect.com/science/article/pii/ S0968090X23004096 [170] B. Zeng, X. Yu, Z. Zhang, Z. Pu, Optimizing charging station locations for mixed traffic of electric and gasoline vehicles: a 99 stochastic user equilibrium approach, Transportation Research Part E: Logistics and Transportation Review 204 (2025) 104438. doi:https://doi.org/10.1016/j.tre.2025.104438. URL https://w.sciencedirect.com/science/article/pii/ S136655452500479X [171] M. J. Beckmann, C. B. McGuire, C. B. Winsten, Studies in the Eco- nomics of Transportation, Yale University Press, New Haven, CT, 1956. [172] Y. Sheffi, W. B. Powell, An algorithm for the equilibrium assignment problem with random link times, Networks 12 (2) (1982) 191–207. [173] W. B. Powell, Y. Sheffi, The convergence of equilibrium algorithms with predetermined step sizes, Transportation Science 16 (1) (1982) 45–55. doi:10.1287/trsc.16.1.45. [174] M. Thorhauge, J. Rich, S. Mabit, Charging behaviour and range anxiety in long-distance ev travel: an adaptive choice design study, TransportationPublisher Copyright:© The Author(s) 2024. (2026). doi:10.1007/s11116-024-10561-x. [175] A. Pellegrini, M. Diana, J. Matthew Rose, A latent-based segmentation framework for the investigation of charging behaviour of electric vehicle users, Transportation Research Part C: Emerging Technologies 165 (2024) 104722. doi:https://doi.org/10.1016/j.trc.2024.104722. 100 URL https://w.sciencedirect.com/science/article/pii/ S0968090X24002432 [176] X. Zhou, Y. Ji, Will electric vehicle drivers use smart charging? considering both intentions and preferences, Transportation Research Record 2679 (8) (2025) 409–428. arXiv:https://doi.org/10.1177/ 03611981251333715, doi:10.1177/03611981251333715. URL https://doi.org/10.1177/03611981251333715 [177] C.-L. Zhao, H.-J. Huang, Experiment of boundedly rational route choice behavior and the model under satisficing rule, Transporta- tion Research Part C: Emerging Technologies 68 (2016) 22–37. doi:https://doi.org/10.1016/j.trc.2016.03.006. URL https://w.sciencedirect.com/science/article/pii/ S0968090X16000760 [178] M. Thorhauge, J. Swait, E. Cherchi, The habit-driven life: Accounting for inertia in departure time choices for commuting trips, Trans- portation Research Part A: Policy and Practice 133 (2020) 272–289. doi:https://doi.org/10.1016/j.tra.2020.01.023. URL https://w.sciencedirect.com/science/article/pii/ S0965856419307797 [179] K. Gao, L. Sun, Y. Yang, F. Meng, X. Qu, Cumulative prospect the- ory coupled with multi-attribute decision making for modeling travel 101 behavior, Transportation Research Part A: Policy and Practice 148 (2021) 1–21. doi:https://doi.org/10.1016/j.tra.2021.03.025. URL https://w.sciencedirect.com/science/article/pii/ S0965856421000884 [180] C.-L. Zhao, H.-J. Huang, Modeling bounded rationality in congestion games with the quantal response equilibrium, Procedia - Social and Behavioral Sciences 138 (2014) 641–648, the 9th International Conference on Traffic and Transportation Studies (ICTTS 2014). doi:https://doi.org/10.1016/j.sbspro.2014.07.253. URL https://w.sciencedirect.com/science/article/pii/ S1877042814041731 [181] Y. Xiong, B. An, S. Kraus, Electric vehicle charging strategy study and the application on charging station placement, Autonomous Agents and Multi-Agent Systems 35 (1) (2021) 3. [182] D. Paccagnan, B. Gentile, F. Parise, M. Kamgarpour, J. Lygeros, Nash and wardrop equilibria in aggregative games with coupling constraints, IEEE Transactions on Automatic Control 64 (4) (2019) 1373–1388. doi:10.1109/TAC.2018.2849946. [183] H. Ye, On stochastic-user-equilibrium-based day-to-day dynamics, Transportation Science 56 (1) (2022) 103–117. doi:10.1287/trsc. 2021.1088. 102 [184] J. Bie, H. K. Lo, Stability and attraction domains of traffic equilibria in a day-to-day dynamical system formulation, Trans- portation Research Part B: Methodological 44 (1) (2010) 90–107. doi:https://doi.org/10.1016/j.trb.2009.06.007. URL https://w.sciencedirect.com/science/article/pii/ S0191261509000770 [185] G. Wang, S. Ma, N. Jia, A combined framework for modeling the evolution of traveler route choice under risk, Transporta- tion Research Part C: Emerging Technologies 35 (2013) 156–179. doi:https://doi.org/10.1016/j.trc.2013.06.019. URL https://w.sciencedirect.com/science/article/pii/ S0968090X13001526 [186] K. Satsukawa, K. Wada, T. Iryo, Stability analysis of a depar- ture time choice problem with atomic vehicle models, Trans- portation Research Part B: Methodological 189 (2024) 103039, transportation Research Part B: Methodological - ISTTT25. doi:https://doi.org/10.1016/j.trb.2024.103039. URL https://w.sciencedirect.com/science/article/pii/ S0191261524001632 [187] X. Huang, Y. Lin, M. K. Lim, F. Zhou, F. Liu, Electric ve- hicle charging station diffusion: An agent-based evolutionary game model in complex networks, Energy 257 (2022) 124700. 103 doi:https://doi.org/10.1016/j.energy.2022.124700. URL https://w.sciencedirect.com/science/article/pii/ S0360544222016036 [188] M. Huang, R. P. Malhamé, P. E. Caines, Large population stochastic dynamic games: Closed-loop McKean–Vlasov systems and the Nash certainty equivalence principle, Communications in Information and Systems 6 (3) (2006) 221–252. doi:10.4310/CIS.2006.v6.n3.a5. [189] R. Lin, Z. Xu, X. Huang, J. Gao, H. Chen, T. Shen, Optimal schedul- ing management of the parking lot and decentralized charging of electric vehicles based on mean field game, Applied Energy 328 (2022) 120198. doi:https://doi.org/10.1016/j.apenergy.2022.120198. URL https://w.sciencedirect.com/science/article/pii/ S0306261922014556 [190] M. J. Smith, The stability of a dynamic model of traffic assignment—an application of a method of lyapunov, Transportation Science 18 (3) (1984) 245–252. doi:10.1287/trsc.18.3.245. [191] M. J. Smith, D. P. Watling, A route-swapping dynamical system and lyapunov function for stochastic user equilibrium, Trans- portation Research Part B: Methodological 85 (2016) 132–141. doi:https://doi.org/10.1016/j.trb.2015.12.015. URL https://w.sciencedirect.com/science/article/pii/ S0191261515002829 104 [192] R. A. Waraich, M. D. Galus, C. Dobler, M. Balmer, G. Andersson, K. W. Axhausen, Plug-in hybrid electric vehicles and smart grids: Investigations based on a microsimulation, Transportation Research Part C: Emerging Technologies 28 (2013) 74–86, euro Transportation: selected paper from the EWGT Meeting, Padova, September 2009. doi:https://doi.org/10.1016/j.trc.2012.10.011. URL https://w.sciencedirect.com/science/article/pii/ S0968090X12001337 [193] R. Wolbertus, R. van den Hoed, M. Kroesen, C. Chorus, Charging infrastructure roll-out strategies for large scale introduction of electric vehicles in urban areas: An agent-based simulation study, Trans- portation Research Part A: Policy and Practice 148 (2021) 262–285. doi:https://doi.org/10.1016/j.tra.2021.04.010. URL https://w.sciencedirect.com/science/article/pii/ S0965856421001087 [194] Z. Yi, X. C. Liu, R. Wei, Electric vehicle demand estimation and charging station allocation using urban informatics, Transportation Research Part D: Transport and Environment 106 (2022) 103264. doi:https://doi.org/10.1016/j.trd.2022.103264. URL https://w.sciencedirect.com/science/article/pii/ S1361920922000931 [195] H. Zhang, D. Jin, B. Han, F. Xue, S. Lu, L. Jiang, Joint opti- 105 mization of fleet sizing, charging station planning, and operation for autonomous electric vehicle fleets in urban transportation net- works, Sustainable Energy, Grids and Networks 44 (2025) 101946. doi:https://doi.org/10.1016/j.segan.2025.101946. URL https://w.sciencedirect.com/science/article/pii/ S2352467725003285 [196] X. Zhou, Y. Cai, Q. Meng, Y. Ji, A user behavior and re- inforcement learning based dynamic pricing method for idle connection time reduction at charging stations, Transporta- tion Research Part A: Policy and Practice 192 (2025) 104358. doi:https://doi.org/10.1016/j.tra.2024.104358. URL https://w.sciencedirect.com/science/article/pii/ S0965856424004063 [197] A. Kazemtarghi, A. Mallik, Y. Chen, Dynamic pricing strat- egy for electric vehicle charging stations to distribute the congestion and maximize the revenue, International Jour- nal of Electrical Power & Energy Systems 158 (2024) 109946. doi:https://doi.org/10.1016/j.ijepes.2024.109946. URL https://w.sciencedirect.com/science/article/pii/ S0142061524001674 [198] Y. Xiong, J. Gan, B. An, C. Miao, Y. C. Soh, Optimal pricing for efficient electric vehicle charging station management, in: Proceedings 106 of the 2016 International Conference on Autonomous Agents & Multi- agent Systems, 2016, p. 749–757. [199] R. Sen, F. Liu, J. P. Talusan, A. Pettet, Y. Suzue, M. Bailey, A. Mukhopadhyay, A. Dubey, Consent: A negotiation framework for leveraging user flexibility in vehicle-to-building charging under uncer- tainty, arXiv preprint arXiv:2601.01581 (2026). arXiv:2601.01581, doi:10.48550/arXiv.2601.01581. URL https://doi.org/10.48550/arXiv.2601.01581 [200] Z. Chen, Y. Deng, C. Xie, C. Guan, T. Pan, Network equilibrium of battery electric vehicles considering drivers’ resting behavior, Transportation Research Part C: Emerging Technologies 155 (2023) 104305. doi:https://doi.org/10.1016/j.trc.2023.104305. URL https://w.sciencedirect.com/science/article/pii/ S0968090X23002942 [201] W. Yang, A user-choice model for locating congested fast charging stations, Transportation Research Part E: Logistics and Transportation Review 110 (2018) 189–213. doi:https: //doi.org/10.1016/j.tre.2017.11.009. URL https://w.sciencedirect.com/science/article/pii/ S1366554517305100 [202] L. Liu, Z. Huang, J. Xu, Multi-Agent Deep Reinforcement Learning Based Scheduling Approach for Mobile Charging in Internet of Elec- 107 tric Vehicles , IEEE Transactions on Mobile Computing 23 (10) (2024) 10130–10145. doi:10.1109/TMC.2024.3373410. URL https://doi.ieeecomputersociety.org/10.1109/TMC.2024. 3373410 [203] D. Qiu, Y. Wang, M. Sun, G. Strbac, Multi-service provision for electric vehicles in power-transportation networks towards a low-carbon transition: A hierarchical and hybrid multi-agent rein- forcement learning approach, Applied Energy 313 (2022) 118790. doi:https://doi.org/10.1016/j.apenergy.2022.118790. URL https://w.sciencedirect.com/science/article/pii/ S0306261922002379 [204] Y. Xiong, J. Gan, B. An, C. Miao, A. L. C. Bazzan, Optimal elec- tric vehicle fast charging station placement based on game theoretical framework, IEEE Transactions on Intelligent Transportation Systems 19 (8) (2018) 2493–2504. doi:10.1109/TITS.2017.2754382. 108