Dynamic Dispatch of Heterogeneous Autonomous Electric Fleets under First-In–First-Out Traffic and Charging Constraints

Abstract

Dynamic routing is operationally fragile when autonomous electric vehicles differ in capacity and energy use, traffic changes mid-arc, and charging delays redispatch. We formulate a bi-objective problem minimizing operating cost and customer response without time windows. The model combines FIFO-consistent piecewise traffic, load- and speed-sensitive energy, autonomy auxiliary power, SOC reserve, charging loss and duration, and three vehicle classes. Real-world light-duty data and VECTO heavy-duty simulations calibrate four speeds (18.12–40.60 km/h) and three energy intensities (0.146–1.365 kWh/km). We introduce causal charge-aware regret insertion with exchange (CCARI-EX), combining adaptive batching, heterogeneous insertion, relocation, and cross-class exchange. A route-set mixed-integer model proves supported optima for 63 scalarized solves over 21 dispatch epochs. A frozen-seed holdout contains 480 request streams and 4,800 policy executions across four sizes, four release processes, and 30 seeds per cell. An independently implemented validator recorded no rejected dispatch. CCARI-EX (0.75) averaged USD 405.18 and 88.96 min response, improving cost-greedy by USD 6.53 and 4.71 min (paired Holm-adjusted p < .001); it was nondominated in 69.0% of streams. The exact formulation validation confirmed consistency between the route-set model and heuristic evaluation on tractable dispatch epochs. A 12-run SUMO transfer reproduced ordered traffic regimes but exposed severe congestion censoring. The results support charging-continuous replanning while delimiting what aggregate traffic calibration can establish.