An integrated multi-objective approach to collaborative disassembly line balancing and planning with TAOG-based modelling

With the rise of Industry 5.0 and growing environmental regulations, efficient disassembly systems are more critical than ever. Disassembly enables the recovery of valuable components from End-of-Life (EoL) products and supports sustainable production strategies. This paper addresses the collaborative disassembly line balancing problem, jointly considering disassembly planning and task allocation in a human-robot environment. A bi-objective MILP model is proposed to minimise total line cost and balance workload across stations, incorporating complex precedence constraints extracted from a Transformed AND/OR Graph (TAOG) representing multiple disassembly alternatives. The model accounts for both interactive and parallel human-robot collaboration. The ϵ-constraint method is employed to generate optimal Pareto fronts for small and medium-sized instances, providing a benchmark to validate the approximate approach. For large-scale instances, we develop a two-phase Greedy Local Search-NSGA-II (GLS-NSGA-II) algorithm, adapted to the structure of the disassembly process and human-robot interaction dynamics. Numerical experiments on three real-world case studies demonstrate the effectiveness of the proposed framework and confirm the ability of GLS-NSGA-II to achieve optimal or near-optimal Pareto solutions, even for complex problems. The results are also analysed from a managerial perspective, providing decision guidelines that highlight trade-offs between cost and workload balancing across different human–robot collaboration modes.