Abstract: Electric vehicle(EV) mobile charging stations(MCS) face challenges including demand prediction errors, insufficient multi-objective coordination, and lack of multi-agent interaction mechanisms, necessitating multidimensional collaborative optimization. An IGDT-based dual-strategy DQN optimization method for MCS scheduling is proposed. First, a multi-objective optimization framework is constructed to achieve coordinated balance with the goals of minimizing total system costs, minimizing the average user waiting time, and improving the satisfaction rate of charging demand. Second, by combining IGDT to quantify uncertainty parameters in charging demand, a dynamic scheduling strategy is designed. Under high-uncertainty scenarios(such as holiday demand spikes), a robustness model is activated to prioritize to ensure uninterrupted basic services in high-demand areas. Under low-uncertainty periods(such as weekday commutes periods), an opportunistic model is applied to optimize costs through route planning and off-peak charging, accommodating different scenario-based decision preferences. Finally, an improved dual-strategy deep reinforcement learning DQN algorithm is used to solve the model, and its effectiveness is verified through simulation experiments. Case analysis shows that, compared to the traditional actor-critic(AC) algorithm, the proposed DQN model can effectively reduce MCS operating costs and reasonably meet EV users' charging demands by dynamically selecting robust or opportunistic decision-making.

Key words: electric vehicle, mobile charging station, information gap decision theory(IGDT), reinforcement learning, optimal scheduling