Search papers, labs, and topics across Lattice.
This paper introduces a scale-invariant sampling strategy for multi-arm bandit rapidly-exploring random tree (MAB-RRT) planners to improve object extraction in disassembly tasks. The method uses a grow-shrink search to identify high-entropy sampling scales and then applies PCA to find effective extraction directions. Experiments across eight 3D scenarios show the proposed approach improves success rates by an order of magnitude compared to classical and modern sampling strategies in 7 out of 8 scenarios.
Scale-invariant sampling boosts object extraction success rates by 10x in complex disassembly scenarios, outperforming existing sampling strategies.
Object extraction tasks often occur in disassembly problems, where bolts, screws, or pins have to be removed from tight, narrow spaces. In such problems, the distance to the environment is often on the millimeter scale. Sampling-based planners can solve such problems and provide completeness guarantees. However, sampling becomes a bottleneck, since almost all motions will result in collisions with the environment. To overcome this problem, we propose a novel scale-invariant sampling strategy which explores the configuration space using a grow-shrink search to find useful, high-entropy sampling scales. Once a useful sampling scale has been found, our framework exploits this scale by using a principal components analysis (PCA) to find useful directions for object extraction. We embed this sampler into a multi-arm bandit rapidly-exploring random tree (MAB-RRT) planner and test it on eight challenging 3D object extraction scenarios, involving bolts, gears, rods, pins, and sockets. To evaluate our framework, we compare it with classical sampling strategies like uniform sampling, obstacle-based sampling, and narrow-passage sampling, and with modern strategies like mate vectors, physics-based planning, and disassembly breadth first search. Our experiments show that scale-invariant sampling improves success rate by one order of magnitude on 7 out of 8 scenarios. This demonstrates that scale-invariant sampling is an important concept for general purpose object extraction in disassembly tasks.
