Metaheuristic Optimization of a Multi-Type Human–Robot Collaborative Disassembly Line for End-of-Life Electric Vehicle Batteries

Abstract

This thesis focuses on the optimization of a multi-type human–robot collaborative disassembly line for end-of-life electric vehicle batteries. The studied problem combines several interdependent decisions, including task sequencing, station opening, task-to-station assignment, execution mode selection, resource compatibility, precedence constraints, cycle-time limitations and uncertainty in human operation times. The objective of this work is to propose a scalable metaheuristic approach capable of solving larger instances for which exact optimization becomes computationally difficult. For this purpose, an Adaptive Large Neighbourhood Search approach is adapted to the structure of the studied problem and denoted HRC-ALNS. Two versions are developed. The deterministic version, called D-HRC-ALNS, considers fixed operation times, while the stochastic version, called S-HRC-ALNS, integrates scenario-dependent human operation times and recourse costs related to station overloads. The proposed method is based on a specific solution representation composed of a global operation sequence, execution modes and active OR precedence choices. A decoder is then used to reconstruct station assignments, determine resource configurations, compute station loads and evaluate the objective function. Mixed destruction and hybrid repair operators are designed to guide the search toward better solutions. In the stochastic version, additional information such as human-time variability, recourse and worst-scenario loads is also exploited. Computational experiments were conducted on instances involving one to five battery types. For the small instances, the results were compared with exact MILP solutions. The obtained gaps were very small, confirming the quality of the proposed metaheuristic. For larger instances, the HRC-ALNS was able to generate feasible and interpretable solutions within reasonable computational time. The results show that the proposed approach represents a flexible and efficient solution method for complex human–robot collaborative battery disassembly line problems.