Heterogeneous Alignment-Based Spatio-Temporal Graph Reinforcement Learning for Dynamic Multi-UAVs Task Assignment
Haojie Zhu, Mou Chen, Tongle Zhou, Zengliang Han · 2025
The dynamic task assignment for heterogeneous Unmanned Aerial Vehicles (UAVs) has emerged as a pivotal challenge in mission-critical applications such as disaster response and urban logistics. A novel Heterogeneous Alignment-based Spatio-Temporal Graph Reinforcement Learning (HASTG-RL) framework is proposed to address this issue. First, a Dynamic Spatio-Temporal Graph (DSTG) is constructed to update environment states continuously. Second, a Transformer-based heterogeneous alignment mechanism is developed to resolve the UAV heterogeneity. Moreover, independent critic networks incorporating multi-objective optimization are designed to simultaneously evaluate task completion, energy consumption, and flight distance. The simulation results show that the proposed method demonstrates superior performance in task completion rate and better flight distance compared to state-of-the-art baselines in dynamic scenarios.