Component-aware topology-guided task release and collision-free bimanual scheduling for cooperative disassembly of retired EV battery packs
Retired EV battery-pack disassembly requires robots to handle fastening dependencies, occlusion, tool compatibility, and inter-arm collisions in a compact workspace. This paper presents a component-aware framework that separates topology-guided task release, frontier-level operation selection, and fixed-path bimanual temporal coordination. A component-removal task is released only after its fasteners and blockers have been cleared and a tool-compatible single-arm path is feasible. Released operations are assigned to the screwdriver- or gripper-side arm, and the proposed priority rule selects the next operation from the current arm-specific executable frontier. For each selected path pair, an A ∗ search in path-index space coordinates left-arm, right-arm, or simultaneous advancement along fixed pre-generated single-arm paths, with node collision checks and time-synchronized sampled transition validation. The method was evaluated on a fixed electric-bus battery pack after manual safety preprocessing and top-cover removal. Across five complete robotic-stage trials per mode, S3 achieved a mean completion time of 1348 ± 53 s compared with 1955 ± 55 s for S0, corresponding to a 31.05% reduction and a reconstructed scheduling-level motion-overlap ratio of 44.51%. In a separate controlled 24-operation half-pack simulation, Proposed Priority reduced the makespan from 329.048 s with SPT to 317.273 s (3.58%) and increased the motion-overlap ratio from 4.48% to 8.33%, while the executable-frontier construction, arm assignment, pre-generated paths, A ∗ coordinator, collision model, and motion settings were held fixed. The results distinguish the effects of system-level task availability, frontier-level task selection, and fixed-path temporal coordination. Robustness to execution uncertainty and continuous collision freedom between finite collision-checking samples remain outside the demonstrated scope.
Authors
- Junyang Wang (ORCID: https://orcid.org/0000-0002-3204-6607)
- Wu Yan (ORCID: https://orcid.org/0000-0002-5348-3574)
- Shaohua Kong (ORCID: https://orcid.org/0009-0008-8389-3196)
- Jiangming Zhang
- Zhaoyang Liao
- Zhongjun He (ORCID: https://orcid.org/0009-0003-0117-5347)
- Xubin Lin
- Li Jiang
- Xuefeng Zhou
- Rucheng Xiang
- Zhihao Xu
- Zhijian Fu
- Hongmin Wu
Institutions
- Guangdong Academy of Sciences (CN)
- Guangdong Institute of Intelligent Manufacturing (CN)
- Wuyi University (CN)
- University of Hong Kong (HK)
- Wuyi University (CN)
Publication Details
- Journal
- Robotics and Computer-Integrated Manufacturing
- Published
- 2026-09-30
- DOI
- https://doi.org/10.1016/j.rcim.2026.103439
- Primary Topic
- Robot Manipulation and Learning
- Type
- article
- Field-Weighted Citation Impact
- 0.00