A bi-objective discrete flower pollination algorithm for planning the collaborative disassembly of retired power batteries by humans and robots

Abstract

Human-robot collaboration (HRC) for the disassembly of retired power batteries is attracting attention due to the complementary advantages of humans and robots. To optimize workforce allocation and enhance scheme flexibility, a disassembly line balancing and sequencing problem in HRC (DLBSP_HRC) is formulated, aiming to minimize the total cost and disassembly time by considering variations in skill levels, workforce sizes, and salary grades. Since DLBSP_HRC is an NP-hard problem, a novel modified discrete flower pollination algorithm with Q-learning (MDFPA_QL) is proposed. The algorithm integrates a driving strategy and a preference policy based on the unique characteristics of the problem and incorporates Q-learning to intelligently balance global and local searches. Subsequently, the disassembly of the Tesla Model S is used to validate the advantage of MDFPA_QL over four other advanced meta-heuristics. Furthermore, a knowledge-based selection mechanism is introduced, examining the relationship between delay penalties from large-scale tasks and the cost of employing multi-human-robot teams with various skills. Comparative analysis across different scenarios highlights the superiority of the multi-human-robot scheme over traditional methods.