Multi-fidelity optimization with Adaptive Optimal Sampling for Position-constrained Human-robot Collaborative Disassembly Sequence Planning and Line Balancing

Abstract

When disassembly large-occupied end-of-life (EOL) products, the operation position is constrained due to geometric constraints and distribution. Operators with multiple tasks need to switch between designated operation positions, and the moving time cannot be ignored in this case. This paper proposes a position-constrained human-robot collaborative disassembly sequence planning and line balancing (PC-HRC-DSPLB) problem. In PC-HRC-DSPLB, each task has an optional operation position set, and the opted task will be assigned to a selectable operation position so that it can be completed earlier. Considering the solving difficulty of PC-HRC-DSPLB, a multi-fidelity optimization with adaptive optimal sampling (MOAOS) is proposed to solve a great deal of PC-HRC-DSPLB instances. The group order and sampling probability are adjusted adaptively in MOAOS, and the effectiveness of MOAOS is proved by comparing with typical multi-fidelity optimization algorithms.