The goal here is to complement the “teleoperation” mode of control (decoding joint angles from EMG sensors on the arm) with an active compliance algorithm which uses tactile sensors to maximize the contact in surface with the object, while controlling for contact forces.
EMG decoding: Katie Zhuang
Active compliance: Nicolas Sommer
The index and pinky fingers are mapped to the two fingers of the robotic hand (the third finger of the robot is broken in this video).
In the video:
I) Teleoperation.
II) Grasping with teleoperation only.
III) Grasping with teleoperation and compliant control.
This video demos an early prototype of the system.
Navigation between obstacles using tactile sensing, based on demonstrations:
Comparison with and without the learned behaviour:
Page in progress, more videos will be added soonIn the meantime, more videos can be found on my YouTube page: YouTube