Event

Beihang Team Proposes Rider-Drone Relay Delivery Solution Framework MoE-Pointer, Published at KDD 2026

1 reports · 1 sources

Professor Wang Jingyuan's team from the School of Computer Science at Beihang University, in collaboration with the School of Management at Shanghai University, defined the "Dynamic Multi-Stage Pickup and Delivery Problem" (DM-PDP) for the rider-drone relay delivery scenario and proposed the solution framework MoE-Pointer, which uses a sequence-to-sequence generation framework to achieve end-to-end reinforcement learning joint optimization of a three-tier delivery chain, simultaneously handling order allocation and route planning. The research has been published at the data mining conference KDD 2026. According to the paper, experiments were conducted on synthetic datasets and real urban datasets from Hangzhou, Shanghai, and Chongqing, comparing multiple classes of baseline methods including rolling-horizon OR solvers, metaheuristic algorithms, rule-based heuristics, and reinforcement learning; ablation experiments verified the necessity of each module, and the method maintained robust performance when tested under constraints such as station congestion, drone endurance limits, and travel time fluctuations.