Event
Proposes multi-agent reinforcement learning method for UAV obstacle-avoidance path planning
1 reports · 1 sources
Qizhen Wu, Kexin Liu, Lei Chen, and Jinhu Lü propose a multi-agent reinforcement learning method improved based on model predictive control ideas, using centralized training and decentralized execution, for path planning and obstacle avoidance of multiple UAVs in stochastic environments. Each agent communicates only with a centralized planner and makes decentralized decisions online in stochastic environments; each agent plans feasible paths through an extended observation that fuses neighboring agent information based on a distance-weighted mean-field method. The study also employs multi-step value convergence to improve training efficiency and reduce the expensive interactions required for convergence. The authors state that comparative, ablation, and real-machine experiments verified the effectiveness and generalization performance of the method.
Report timeline
- Multi-Agent Reinforcement Learning-Based UAV Path Planning for Obstacle Avoidance in Stochastic Environments
This study proposes a centralized training with decentralized execution method based on multi-agent reinforcement learning for UAV obstacle-avoidance path planning in stochastic environments, improved from the model predictive control (MPC) idea. Each agent communicates only with the centralized planner, plans feasible paths through extended observations based on a distance-weighted mean field approach, and uses multi-step value convergence to improve training efficiency. Comparative, ablation, and real-machine experiments verified the method's effectiveness and generalization performance.
arXiv cs.RO ·