Event
ContactBench 1.0 Contact-Rich Manipulation Benchmark Released
1 reports · 1 sources
Jingyi Xiang and other authors propose ContactBench 1.0, which decomposes contact-rich manipulation into eight basic skills and builds a benchmark with 36 reference tasks and more than 40 object assets; these assets can be reproduced on consumer-grade 3D printers and are paired with corresponding simulation models. The authors state that the benchmark will be open-sourced, providing physical and simulated assets as well as controlled task variants for evaluating the generalization ability of robotic systems. Recent purely visual and visual-tactile policy learning methods were evaluated on six benchmark tasks, and the results show that purely visual policies struggle to perceive or respond to contact events, while visual-tactile policies also did not bring significant improvement.
Report timeline
- ContactBench 1.0: A Benchmark for Everyday Contact-Rich Robotic Manipulation
The authors propose ContactBench 1.0, which uses 36 reference tasks and more than 40 object assets to break contact-rich manipulation down into eight essential skills; the assets can be reproduced on consumer-grade 3D printers and each comes with a matching simulation model. They evaluate recent vision-only and visual-tactile policy learning methods on six benchmark tasks, and the results show that vision-only policies struggle to perceive or respond to contact events, while visual-tactile policies do not bring noticeable improvement. The benchmark will be open-sourced, providing physical and simulated assets and controlled task variations for evaluating the generalization ability of robotic systems.
arXiv cs.RO ·