Coordinated charging station search in stochastic environments: A multiagent approach

Abstract Range and charge anxiety remain essential barriers to a faster electric vehicle (EV) market diffusion. To this end, quickly reliably finding suitable charging stations may foster an EV uptake by mitigating drivers' anxieties. Here, existing commercial services help drivers find available based on real‐time availability data but struggle with inaccuracy, for example, due conventional vehicles blocking the access public stations. In context, recent works have studied stochastic search methods account uncertainty in order minimize driver's detour until reaching station. So far, both practical theoretical approaches ignore driver coordination enabled requests centralization or sharing of data, observations stations' visit intentions between drivers. Against background, we study coordinated algorithms, which reduce station conflicts improve experience. We model multiagent problem as finite‐horizon Markov decision process introduce online solution framework applicable static dynamic policies. contrast policies, policies information updates during policy planning execution. present hierarchical implementation single‐agent heuristic decentralized making rollout algorithm centralized making. Extensive numerical studies show that compared uncoordinated setting, setting decreases system cost 26%, is nearly good 28% decrease achieved setting. Even long horizons, our reduces 25% while increasing each reliability.