Transfer Learning-Assisted Evolutionary Dynamic Optimisation for Dynamic Human-Robot Collaborative Disassembly Line Balancing

Abstract

In a human-robot collaborative disassembly line, multiple people and robots collaboratively perform disassembly operations at each workstation. Due to dynamic factors, such as end-of-life product quality and human capabilities, the line balancing problem for the human-robot collaborative disassembly line is a dynamic optimisation problem. Therefore, this paper investigates this problem in detail and commits to finding the evolutionary dynamic optimisation. First, a task-based dynamic disassembly process model is proposed. The model can characterise all feasible task sequences of disassembly operations and the dynamic characteristics of tasks affected by uncertain product quality and human capabilities. Second, a multiobjective optimisation model and a feature-based transfer learning-assisted evolutionary dynamic optimisation algorithm for the dynamic human-robot collaborative disassembly line balancing problem are developed. Third, the proposed algorithm uses the balanced distribution adaptation method to transfer the knowledge of the optimal solutions between related problems in time series to track and respond to changes in the dynamic disassembly environment. Then, it obtains the optimal solution sets in a time-varying environment in time. Finally, based on a set of problem instances generated in this study, the proposed algorithm and several competitors are compared and analysed in terms of performance indicators, such as the mean inverted generational distance and the mean hypervolume, verifying the effectiveness of the proposed algorithm on dynamic human-robot collaborative disassembly line balancing.