Online AUV Path Replanning Using Quantum-Behaved Particle Swarm Optimization with Selective Differential Evolution

• Published in: Computer modeling in engineering & sciences 2020-01, Vol.125 (1), p.33-50
• Main Authors: Lim, Hui Sheng, Chin, Christopher K. H., Chai, Shuhong, Bose, Neil
• Format: Article
• Language: English
• Subjects: Autonomous Underwater Vehicle; Autonomous underwater vehicles; Computer simulation; Evolutionary algorithms; Evolutionary computation; Marine environment; Monte Carlo Methods; Monte Carlo simulation; Ocean currents; Optimization; Particle Swarm Optimization; Path Planning; Robustness (mathematics); Sonar Detection; Stochastic processes; Travel time
• ISSN: 1526-1492; 1526-1506; 1526-1506
• DOI: 10.32604/cmes.2020.011648

This paper presents an online AUV (autonomous underwater vehicle) path planner that employs path replanning approach and the SDEQPSO (selective differential evolution-hybridized quantum-behaved particle swarm optimization) algorithm to optimize an AUV mission conducted in an unknown, dynamic and cluttered ocean environment. The proposed path replanner considered the effect of ocean currents in path optimization to generate a Pareto-optimal path that guides the AUV to its target within minimum time. The optimization was based on the onboard sensor data measured from the environment, which consists of a priori unknown dynamic obstacles and spatiotemporal currents. Different sensor arrangements for the forward-looking sonar and horizontal acoustic Doppler current profiler (H-ADCP) were considered in 2D and 3D simulations. Based on the simulation results, the SDEQPSO path replanner was found to be capable of generating a time-optimal path that offered up to 13% reduction in travel time compared to the situation where the vehicle simply followed a path with the shortest distance. The proposed replanning technique also showed consistently better performance over a reactive path planner in terms of solution quality, stability, and computational efficiency. Robustness of the replanner was verified under stochastic process using the Monte Carlo method. The generated path fulfilled the vehicle's safety and physical constraints, while intelligently exploiting ocean currents to improve the vehicle's efficiency.