ChargeUp: Trust-Aware EV Charging Scheduling via Fuzzy-Stackelberg Game Theory and Credibility-Shaped Reinforcement Learning

Abstract

The fast development of electric vehicles (EVs) is placing pressure on charging stations that have a heterogeneous demand, unpredictable arrivals and non-classical strategic behaviour that fail to be captured by traditional queueing rules. Specifically, priority-based schedulers can be abused by users who misreport either the state-of-charge or the urgency, harming both operational efficacy and fairness. In this paper, we introduce ChargeUp, a 7-equation scheduling system that integrates four components: (i) a 27-rule Mamdani fuzzy priority engine that includes a trust-modulated cooperation and aging-based starvation prevention, (ii) a load-aware Stackelberg-inspired heuristic pricing layer, (iii) a Bayesian two-dimensional trust model that is updated asynchronously, and (iv) credibility-shaped reinforcement learning for charger assignment. We also evaluate the system using a Price-of-Anarchy-inspired empirical cost ratio, Gini/Lorenz analysis, and a need-based fairness index to capture tradeoffs that are not visible from mean wait time alone. In a 500-run Monte Carlo study against eight baselines, ChargeUp reduces mean wait to 54.7 min, a 38.0% improvement over FIFO ($p < 0.001$, Cohen's $d = -1.14$), while statistically outperforming SJF on wait time and outperforming several baselines on fairness- and revenue-oriented metrics. These results indicate that an integrated trust-aware scheduler can improve simulated charging performance under adversarial behavior, while also highlighting the need for field-calibrated validation beyond the present synthetic environment.