Robot Manipulation Skill Learning

This work proposes PRIMP: PRobabilistically-Informed Motion Primitives, a learning-from-demonstration (LfD) method using probability densities on the workspaces of robot manipulators. The model learns the probability distribution of the end effector trajectories in the 6D workspace. It is able to adapt to new situations such as novel via points with uncertainty. Workspace-STOMP, a new version of the existing STOMP motion planner, is also introduced, which can be used as a post-process to improve the performance of PRIMP and any other reachability-based LfD method. The combination of PRIMP and Workspace-STOMP can further help the robot avoid novel obstacles that are unseen during the demonstration process. PRIMP runs more than 5 times faster than existing state-of-the-art methods while generalizing trajectories more than twice as close to both the demonstrations and novel desired poses. The methods are demonstrated in household tasks with unseen obstacles, e.g., pouring water, scooping particles, opening and closing doors, etc.

[Paper]

[1] Ruan, S., Liu, W., Wang, X., Meng, X. and Chirikjian, G.S., 2024. PRIMP: PRobabilistically-Informed Motion Primitives for Efficient Affordance Learning from Demonstration. IEEE Transactions on Robotics. doi: 10.1109/TRO.2024.3390052

https://ieeexplore.ieee.org/document/10502164