Abstract: To address the cooperative delivery routing problem in urban last-mile distribution arising from the coupled effects of limited electric vehicle range, customer time windows, drone no-fly zones, and vehicle-drone launch and recovery synchronization, an electric vehicle-multi-drone cooperative delivery optimization problem considering charging constraints and no-fly zones is investigated. A mixed-integer programming model is formulated with the objective of minimizing the system makespan by jointly incorporating vehicle routing, drone task assignment, charging-station visits, customer time windows, no-fly-zone avoidance, and vehicle-drone synchronization. Based on the adaptive large neighborhood search framework, an improved ALNS algorithm is developed by integrating a charging-aware repair strategy, an arc-based drone task generation and priority assignment mechanism, and local search operators. Numerical experiments are conducted using instances derived from the Solomon C101 benchmark. The results show that, for the small-scale N6, N7, and N8 instances, the improved ALNS obtains the same makespan as complete enumeration. For the 15-, 20-, 25-, and 30-customer instances, the average makespan is reduced by 1.26%, 11.44%, 8.45%, and 1.76%, respectively, compared with Basic-ALNS, and the improvements for the 15-, 20-, and 25-customer instances are statistically significant at the 0.05 level. For the 30-customer stress instance, the improved ALNS achieves a feasibility rate of 60%, compared with 30% for Basic-ALNS. Ablation experiments further demonstrate that the charging-aware repair, drone task priority assignment, and local search strategies effectively improve route feasibility and solution quality. The results indicate that the proposed model and improved algorithm can effectively handle the coupled constraints of charging, time windows, no-fly zones, and vehicle-drone synchronization, providing an effective approach for electric vehicle–drone cooperative last-mile delivery optimization.