A Q-Learning-Based Multi-Objective PSO Algorithm for Robot Path Planning on Rough Terrain

Path planning is a key challenge in the field of mobile robotics. However, the standard particle swarm optimization (PSO) has some significant limitations when dealing with path planning problems, including a tendency to fall into local optimal solutions, lack of diversity and low accuracy. These limitations directly affect the quality and efficiency of path planning, resulting in low exploration capability, increased resource consumption, and increased risk of collision and path deviation. Therefore, a novel hybrid path planning algorithm is proposed to solve these problems. The algorithm combines Q-learning (QL) with multi-objective particle swarm optimization (MOPSO), called MOQLPSO. In this algorithm, QL is introduced into MOPSO, which is used to dynamically adjust the inertia weight and learning factors of MOPSO online to improve the convergence ability of the algorithm to the Pareto frontier. The simulation results show that compared with the standard MOPSO, the proposed MOQLPSO algorithm has significant advantages in search accuracy, population diversity and optimization effect. Specifically, MOQLPSO reduces path length by 7.96%, while reducing terrain roughness by 22.60%, respectively.