Dynamic path planning for certain mobile robots in the 3D rough terrain: Fusion of the Q-learning enhanced MOPSO and improved DWA

Targeting the standard particle swarm optimization (PSO) approach that tends to converge to the local optimum and struggles to solve the path planning problem in dynamic environments, a novel hybrid path planning algorithm is proposed for mobile robots operating in intricate 3D terrain, which is a fusion algorithm called MOQLPSO-ADWA, combining the Q-learning (QL) based multi-objective particle swarm optimization (MOPSO) and the adaptive dynamic window approach (ADWA). Firstly, QL is introduced into MOPSO to update the inertia weights and acceleration parameters of the MOPSO online to improve the ability of the current particles to converge to the Pareto front. Secondly, the crossover operator is introduced to increase the explorability and diversity of the population during the iteration process. Thirdly, the obstacle avoidance strategy in DWA is modified where the weighting coefficients of the dynamic window evaluation function of the DWA algorithm are dynamically designed based on global optimization, and adjusted according to the information of targets and obstacles, to improve the adaptive ability of mobile robots to different environments. After that, the planned path points of the MOQLPSO algorithm are extracted as the temporary target points of the ADWA algorithm, and the fusion of MOQLPSO and ADWA algorithms is realized on the basis of the global optimum. The fusion algorithm is confirmed to be able to successfully avoid obstacles and achieve a significant improvement in planning efficiency and path safety through the comparison of simulation results, which is more in line with the motion characteristics of mobile robots. Both the problem and the obtained results exhibit significant nonlinear characteristics, which play a crucial role in the complexity of the search space and the diversity of solutions.