Fixed-wing unmanned aerial vehicle distributed formation and obstacle avoidance method based on diffusion reinforcement learning
Patent Information
- Application Number
- CN202511608251.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-11-05
- Publication Date
- 2026-03-03
AI Technical Summary
Existing technologies struggle to effectively achieve distributed formation and obstacle avoidance of fixed-wing UAV swarms in complex and unknown environments, especially in situations with dense obstacles. Traditional methods lack adaptability and are prone to oscillating behavior or collisions.
A deep reinforcement learning approach based on a generative-diffusion model is adopted for multiple UAVs. Through a diffusion policy network and a graph value network, control commands are generated using local observations to achieve distributed formation and obstacle avoidance of UAVs. This method comprises a diffusion model policy network (actuator) and a graph value network (evaluator), operating under a centralized training and distributed execution paradigm. A graph attention network is used to process dynamic neighbor information to generate diverse actions.
In unknown and chaotic environments, fixed-wing UAV swarms can maintain tight formation and navigate safely, significantly improving flight stability and reducing collision rates, which is superior to existing methods.
Smart Images

Figure CN121596887A_ABST
Abstract
Citation Information
Cited By
Flight cabin remote real machine control method based on multi-source perception fusion
CN122363341A
Flight cabin remote real machine control method based on multi-source perception fusion
CN122363341B