Design of Multi-UAV Task Allocation Algorithm Based on Deep Reinforcement Learning
作者
Yuxin Cui,Yan Gao,Dequn Zhao
出处
期刊:2020 International Conference on Computer Engineering and Application (ICCEA) [Institute of Electrical and Electronics Engineers] 日期:2025-04-25卷期号:: 440-443被引量:4
标识
DOI:10.1109/iccea65460.2025.11103142
摘要
The paradigm of multi-UAV task execution is rapidly emerging. Solving the task allocation strategy for UAV swarms is an NP-hard problem, traditionally addressed through heuristics and manual rules. This paper proposes MTDRL (Multi-UAV Task Distribution with Deep Reinforcement Learning), a method that enhances computational efficiency and adaptability. Firstly, the problem is formulated as a Markov Decision Process (MDP) to define its core structure. Then, we design a policy network GFA-Net (GCN Fusion Encoder-Decoder with Multi-Head Attention) to encodes UAV states and task node features while capturing interdependencies through a multi-head attention mechanism. Finally, we employ a rollout-based REINFORCE algorithm to train DRL model, which can optimize the allocation strategy and improving training efficiency. Experiments demonstrate that the proposed method effectively solves task instances in three scenario scales under flight distance constraints, achieving improvements in both speed and efficiency.