Abstract:Range and charge anxiety remain essential barriers to a faster electric vehicle market diffusion. To this end, quickly and reliably finding suitable charging stations may foster an electric vehicle uptake by mitigating drivers' anxieties. Here, existing commercial services help drivers to find available stations based on real-time availability data but struggle with data inaccuracy, e.g., due to conventional vehicles blocking the access to public charging stations. In this context, recent works have studied stochastic search methods to account for availability uncertainty in order to minimize a driver's detour until reaching an available charging station. So far, both practical and theoretical approaches ignore driver coordination enabled by charging requests centralization or sharing of data, e.g., sharing observations of charging stations' availability or visit intentions between drivers. Against this background, we study coordinated stochastic search algorithms, which help to reduce station visit conflicts and improve the drivers' charging experience. We model a multi-agent stochastic charging station search problem as a finite-horizon Markov decision process and introduce an online solution framework applicable to static and dynamic policies. In contrast to static policies, dynamic policies account for information updates during policy planning and execution. We present a hierarchical implementation of a single-agent heuristic for decentralized decision making and a rollout algorithm for centralized decision making. Extensive numerical studies show that compared to an uncoordinated setting, a decentralized setting with visit-intentions sharing decreases the system cost by 26%, which is nearly as good as the 28% cost decrease achieved in a centralized setting, and saves up to 23% of a driver's search time while increasing her search reliability.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is how to effectively coordinate electric vehicle drivers through a multi - agent system to find suitable charging stations in an uncertain environment, so as to reduce the increase in search time and cost caused by the unavailability of charging stations. Specifically, the paper focuses on how to optimize the charging station search process of multiple electric vehicle drivers by sharing information (such as the planned - to - visit charging stations and the observed occupancy of charging stations), thereby improving the efficiency of the overall system and the charging experience of drivers.
### Background and Motivation of the Paper
With the rapid growth of the electric vehicle market, "range anxiety" (caused by the unreliability and insufficiency of public charging infrastructure) has become one of the main obstacles to the popularization of private electric vehicles. Although existing commercial services can help drivers find available charging stations based on real - time available data, these services are often affected by inaccurate data, such as the "ICEing" phenomenon caused by traditional cars occupying public charging stations. In addition, uncoordinated searches by multiple drivers simultaneously may lead to conflicts, that is, multiple drivers go to the same charging station, thereby increasing search time and cost.
### Research Objectives
The objective of the paper is to solve the above - mentioned problems by introducing coordinated random search algorithms, which can consider information sharing and possible request centralization, thereby reducing charging station access conflicts and improving the charging experience of drivers. Specifically, the authors studied the multi - agent random charging station search problem within a finite - time horizon and modeled it as a finite - time - horizon Markov Decision Process (MDP), and proposed an online solution framework applicable to static and dynamic policies.
### Main Contributions
1. **Formalize the multi - agent decision - making problem**: Define the planning problem as a single - decision - maker MDP, and show that through additional policy constraints, this MDP can be represented as a set of single - agent MDPs, thereby supporting decentralized and static planning.
2. **Propose multiple online algorithms**: For different scenarios, propose multiple online algorithms under static and dynamic planning, including centralized and decentralized planning settings, and different levels of information sharing.
3. **Numerical study**: Conducted extensive numerical studies based on actual cases, and analyzed the improvement effects of different coordination strategies on system performance, driver search time and reliability.
### Results and Impacts
The research shows that, compared with the non - coordinated setting, sharing access intentions in the decentralized setting can reduce system costs by 26%, which is almost the same as 28% in the centralized setting, while also saving up to 23% of the driver's search time and improving search reliability. These results indicate that by coordinating the charging station searches of multiple agents, the charging experience of electric vehicle users and the efficiency of the overall system can be significantly improved.