Torque based stabilization control for torque sensorless humanoid robots

Abstract

This paper proposes a stabilization control for joint torque controlled humanoid robots without joint torque sensors. The proposed controller's role is to compute all joint torques from the desired positions/postures of floating end effectors, the desired Center Of Gravity (COG), the desired root link posture and the reference Zero Moment Point (ZMP). The controller uses a very simple strategy based on the Single Degree Of Freedom (SDOF) model and statics. First, it calculates the wrenches at the COG or end effectors to keep their positions/postures. Second, it calculates the contact end effector wrenches by solving a constrained optimization problem. Finally, it computes joint torques from all end effector wrenches. Furthermore, we propose a method by which we control joint torques without joint torque sensors by compensating inertia torques and frictions. We implement the stabilization controller and the joint torque controller in a high-power biped robot. We confirm the effectiveness of the proposed controller through balancing and walking experiments on rough terrain.