Collaborative Disassembly for Multiple Users in Virtual Reality

Disassembly Sequence Planning (DSP) is critical in product lifecycle management but computationally demanding, with traditional and machine learning approaches facing challenges in scalability and lacking the real-time responsiveness required for interactive collaboration in virtual reality (VR). We present Dynamic Parallel Disassembly and Tasking (DPDT), a novel framework tailored for immersive collaboration. DPDT integrates a spatial-temporal feature-based heuristic to accelerate search prioritization and employs a dynamic reactivation strategy to resolve inter-part dependencies during parallel execution. Furthermore, a proximity-aware task assignment algorithm optimizes the translation of these plans into efficient multi-user instructions. We conduct extensive evaluations on a large-scale dataset against multiple baselines, showing that DPDT significantly outperforms state-of-the-art baselines. Specifically, for complex assemblies, our method achieves a computational speedup of up to 4.7 times while improving planning reliability by approximately 9.3 percentage points. These algorithmic gains translate directly into operational efficiency: a 24-participant user study in VR confirms that DPDT significantly reduces task completion time, physical-effort proxies, and NASA-TLX workload, validating its potential for effective collaborative maintenance training.