Bimanual Robotic Manipulation
Overview
Bimanual robotic manipulation studies how two arms perceive, plan, and act together, enabling tasks such as lifting large objects, folding cloth, assembling parts, and using tools. Common approaches include imitation learning from demonstrations, diffusion and vision-language-action policies, reinforcement learning, and model-based motion or force planning; tactile sensing and geometric representations help with contact-rich tasks. Researchers evaluate methods in simulation benchmarks such as RoboTwin and RLBench, as well as on real robots, using tasks that test coordination, generalization, and long-horizon execution. A central challenge is making policies reliable beyond their training scenes and motions, especially when actions are tightly coupled, objects deform, or simulation differs from the physical world.
Latest papers
189 papers28 papers in the last four weeks, down 15% on the four weeks before. 0.2% of all new papers.