Torque-based multi-task and balancing control for humanoid robots

Abstract

The humanoid robots Justin and Toro at DLR are equipped with joint torque sensors, which allows to implement compliant motion control via highly sensitive impedance behaviors. The control framework presented in this work has been developed using a passivity-based argumentation and avoids feedback cancellation of the natural dynamics. In this overview also the extension to multiple task hierarchies using prioritized null-space projections will be discussed. The approach has been applied to the whole-body control of the mobile manipulator Justin, with position and torque controlled subsystems for locomotion and manipulation. Additionally, a generalization to the bipedal humanoid robot Toro is shown. Thereby, the distribution of the control forces to the available contact points has to be taken into account. The presented control approaches are well suited for tasks in which the robot has to physically interact with humans or the environment.