Table 2 Summary of the dynamic EV-DRT literature

Al-Kanj et al. [50]

EAV-DARP

Linear, CP = A

Dynamic vehicle dispatch, repositioning, and recharge. Uses a hierarchical aggregation approach for value-function approximation under approximate dynamic programming

Bongiovanni [47]

EAV-DARP

Linear, partial recharge, CP = A

A two-stage heuristic approach. Uses a greedy insertion algorithm to insert new feasible requests and then re-optimize the decisions based on large neighborhood search heuristics. New recharging/idling decisions are checked at the end of vehicle routes

Shi et al. [51]

EAV-DARP

Linear, CP = A

Reinforcement learning approach to optimize vehicle routing and charging decisions under a spatio-temporal discretization framework

Kullman et al., [54]

EV-DARP

Linear, full recharge, CP = B

Deep reinforcement learning to learn optimal routing and charging decisions under uncertainty

Ma [17]

EV-DARP

Linear, partial recharge, CP = C

Two-stage optimization approach. Determines when and how much energy to charge in the first stage and then where to charge in the second stage, based on charging station occupancy information. Mixed-integer optimization formulation

Yu et al. [55]

EAV-DARP

Linear, CP = B

Approximate dynamic programming under a Markov decision process

Zalesak and Samaranayake [18]

EV-DARP

Concave, CP = C

Two-stage optimization. Assigns new requests under the current charging schedule in the first stage, then optimizes the charging schedule (when and where to charge) given assigned requests. Mixed-integer optimization formulation. Charging priority depends on the sorted SOC of vehicles