These sketches present our current progress in compliant control of humanoid robots at Willow Garage. The idea is to combine torque-based compliant behaviors for Cartesian and joint space tasks. This approach enables to respond quickly to contact, be safe and use effectively the dynamics of the mechanism. The focus of this research is more on the skill than on the single low level control process. We provide infrastructure to bridge the gap between the skill developer (who cares about the mathematics of movement and robot dynamics) and the skill user (who cares about motion / high level planning and perception).
This video shows various experiments of the control of UT Austin’s humanoid Meka robot. We demonstrate prioritized torque control strategies, addressing hand position tasks and dynamically compensated posture optimization. In particular, the main contribution of the experiments is the ability to optimize posture performance by dynamically compensating the postural mass matrix with the task’s null space matrix.
Collaborative implementation of compliant control with Stanford Univ. and Willow Garage. Three experiments are shown: (a) gravity effort compensation to validate the robot’s dynamic and kinematic model, (b) dynamically weighted posture control to demonstrate the validity of the mass/inertia matrix, and (c) operational space compliant control in the vertical and Sagittal directions to demostrate the ability to respond to the environment using the tool.
A series of experiments to validate the capabilities of a high force compliant fluidic actuator in terms of tracking bandwidth, stiffness regulation, gravity compensation, and torque control.