On the impact of charger queues and traffic dynamics on electric dial-a-ride services

Most studies of electric on-demand services assume static operating conditions, neglecting the exogenous dynamics of transportation systems and their interactions with charging infrastructure. Conversely, the few studies that consider exogenous charging demand and explicitly model queue dynamics over time do not consider optimization of an on-demand passenger service. This paper introduces a framework for optimizing real-time electric dial-a-ride services that accounts for both network and charging station congestion, including exogenous impacts from private (non-fleet) vehicles. The service is modeled as a dynamic electric dial-a-ride problem operating within a rolling horizon framework and solved using Gurobi. Three models have been presented for predicting queue times at charging stations: a static model based on M/M/ n queuing theory, a quasi-dynamic model that directly combines these static predictions with periodic observations of charging station queues, and a dynamic model that explicitly models the evolution of queues over time by integrating queuing theory with periodic queue observations. Network congestion is incorporated through accumulation-based network macroscopic fundamental diagrams to estimate average regional speeds. Results demonstrate that incorporating queue prediction models improves on-demand service quality by enabling fleet vehicles to avoid charging stations with long queues. The dynamic queue model enables the best performance, both in terms of solution computation and the on-demand service objectives.