Efficient collision-free path planning for autonomous underwater vehicles in dynamic environments with a hybrid optimization algorithm

Abstract

This paper presents an efficient path-planner based on a hybrid optimization algorithm for autonomous underwater vehicles (AUVs) operating in cluttered and uncertain environments. The algorithm integrates particle swarm optimization (PSO) algorithm with Legendre pseudospectral method (LPM), which is named as hybrid PSO-LPM algorithm. PSO is first employed as an initialization generator with its strong global searching ability and robustness to random initial values. Then, the searching algorithm is switched to LPM with the initialization obtained by PSO algorithm to accelerate the following searching process. The flatness property of AUV is also utilized to reduce the computational cost for planning, making the optimization algorithm valid for local re-planning to efficiently solve the collision avoidance problem. Simulation results show that the hybrid PSO-LPM algorithm is able to find a better trajectory than standard PSO algorithm and with the re-planning scheme it also succeeds in real-time collision avoidance from both static obstacles and moving obstacles with varying levels of position uncertainty. Finally, 100-run Monte Carlo simulations are carried out to check robustness of the proposed re-planner. The results demonstrate that the hybrid optimization algorithm is robust to random initializations and it is effective and efficient for collision-free path planning.