Energy-constrained collaborative path planning for heterogeneous amphibious unmanned surface vehicles in obstacle-cluttered environments

Abstract

Amphibious unmanned surface vehicles (AUSVs) have gained attention due to their ability to operate across both water surfaces and aerial environments. However, path planning for multiple AUSVs in obstacle-cluttered environments, considering energy consumption, safety, and collaboration, remains a significant challenge. This paper addresses the collaborative path planning problem for multiple AUSVs, focusing on optimizing path length, threat avoidance, path smoothness, energy consumption, and collaboration efficiency in complex environments. A novel hyper-heuristic evolutionary algorithm combined with proximal policy optimization (HEA-PPO) is proposed. This approach utilizes reinforcement learning to dynamically select suitable evolutionary operators, reducing human intervention and enhancing search efficiency. An adaptive waypoint encoding strategy is also introduced to improve path smoothness in cluttered environments. Experimental results show that the HEA-PPO algorithm outperforms existing algorithms regarding path quality, convergence speed, and robustness. The proposed method optimizes safety, energy consumption, and collaboration time across multiple scenarios. The HEA-PPO algorithm significantly improves the robustness and efficiency of collaborative path planning for AUSVs, demonstrating its potential in complex, obstacle-cluttered environments. Integrating reinforcement learning with evolutionary algorithms provides a promising approach for future path-planning applications.