Optimizing a human-robot collaborative selective maintenance problem considering task failure via a genetic engineering with experience replay algorithm

Selective maintenance strategy has been introduced to scientifically optimize maintenance planning and enhance the effectiveness and feasibility of maintenance solutions, thereby identifying the optimal maintenance program that ensures the rational allocation of resources. This paper considers a selective human-robot collaborative maintenance model based on expected profit and worker fatigue accumulation, taking into account the risk of maintenance task failure. Additionally, a Genetic Engineering with Experience Replay Algorithm (GEERA) is developed, which seamlessly integrates the capabilities of the experience replay mechanism to retain dominant individuals and locally identify the optimal solution, along with the global exploration and optimization capabilities of the Genetic Engineering Algorithm (GEA) to explore and optimize globally. Moreover, the rationality of the model and the practical effectiveness of the algorithm are validated through an actual maintenance case study. Based on 20 independent runs, GEERA increases NPS by approximately 16.88% and reduces IGD by 14.35% compared with the best-performing benchmark, while also achieving better HV and Spread values. These results demonstrate its effectiveness in providing high-quality maintenance alternatives for balancing maintenance benefits and worker fatigue.