Multi-source Transfer-assisted Evolutionary Optimization for Dynamic Human-Robot Collaborative Disassembly Planning

Traditional manual disassembly can no longer meet the needs of modern enterprises, so more industrial robots are designed to assist or even replace humans to complete the disassembly. In actual disassembly environments, there are many uncertainties that affect the whole disassembly process. In this paper, a dynamic optimization problem of human and robot collaborative disassembly line balancing is studied, which is based on the uncertainty of human ability. Moreover, the dynamic optimization problem needs to follow the Pareto optimal solution set quickly and accurately, and transfer learning has been proved to be a suitable method. Therefore, this paper proposes a multi-source transfer-assisted evolutionary dynamic optimization algorithm. The algorithm reuses knowledge from multiple historical environments to accelerate the generation of initial population and improve the convergence speed of the solution. In the end, based on several sets of problem instances with different scales and environmental similarities, the effectiveness of the algorithm is validated through comparison of several competitors.