DRL-MORIME: A multi-objective path planning algorithm for mobile robots with adaptive operator selection and archive optimization

Abstract This study focuses on a multi-objective path planning problem that simultaneously minimizes path length, ensures trajectory smoothness, and enhances safety. Based on the characteristics of the problem, a deep reinforcement learning-based multi-objective RIME (DRL-MORIME) optimization algorithm is proposed. Regarding the limitations of the original MORIME, such as premature convergence, low search efficiency, and proneness to local optima, three improvements are made. During the search process, the DRL framework adaptively selects search operators, effectively balancing global exploration and local exploitation. For archive management, a novel hybrid selection mechanism is proposed to improve the diversity and distribution of the Pareto front. To improve path smoothness, a post-processing strategy reduces redundant nodes in the path. Comprehensive experiments and ablation studies on the ZDT, DTLZ, and WFG benchmark suites validate the superior performance of the DRL-MORIME algorithm. Path planning on maps of varying complexity achieves average improvements of 13.10% in path length, 69.58% reduction in total turning angle, and 13.96% reduction in collision risk. These results confirm the effectiveness of the proposed algorithm for multi-objective path planning of mobile robots.