A cooperative coverage path planning method for multiple USVs with dynamic collision avoidance using heuristic fusion and deep reinforcement learning
2
조회
0
다운로드
0
인용
0
참고문헌
키워드
초록
To address the inefficiency and lack of real-time dynamic obstacle avoidance in regional coverage algorithms under complex maritime conditions, this paper proposes a cooperative path planning method for multiple Unmanned Surface Vehicles (USVs) in environments with static and dynamic obstacles. This method employs electronic charts to model the environment, enhancing adaptability to actual sea areas and navigational safety through obstacle expansion and grid processing. Building upon this foundation, it integrates the following three core modules: (1) the Divide Areas Algorithm for Optimal Multi-USVs Coverage Path Planning (DASP) for task assignment, (2) an adaptive heuristic fusion A* algorithm for complete static coverage, and (3) an improved Dynamic Window Approach-Artificial potential field framework with Deep Q-learning optimized weights for dynamic obstacle avoidance. Simulation results demonstrate the proposed method’s effectiveness in achieving full area coverage and avoiding all obstacles. It reduced the total path lengths by 12.7 %, 7.0 %, and 6.8 % across the three comparison experiments, respectively, while effectively mitigating collision risks and path deviation issues caused by improper parameter selection in complex environments.
Abstract
To address the inefficiency and lack of real-time dynamic obstacle avoidance in regional coverage algorithms under complex maritime conditions, this paper proposes a cooperative path planning method for multiple Unmanned Surface Vehicles (USVs) in environments with static and dynamic obstacles. This method employs electronic charts to model the environment, enhancing adaptability to actual sea areas and navigational safety through obstacle expansion and grid processing. Building upon this foundation, it integrates the following three core modules: (1) the Divide Areas Algorithm for Optimal Multi-USVs Coverage Path Planning (DASP) for task assignment, (2) an adaptive heuristic fusion A* algorithm for complete static coverage, and (3) an improved Dynamic Window Approach-Artificial potential field framework with Deep Q-learning optimized weights for dynamic obstacle avoidance. Simulation results demonstrate the proposed method’s effectiveness in achieving full area coverage and avoiding all obstacles. It reduced the total path lengths by 12.7 %, 7.0 %, and 6.8 % across the three comparison experiments, respectively, while effectively mitigating collision risks and path deviation issues caused by improper parameter selection in complex environments.