Implementation and analysis of dynamic stability for bipedal robotic motion

Abstract

This work presents the design and simulation of a stable balance and locomotion approach for a bipedal robot. The torque response of a falling body is modelled and a low-pass filter was designed and implemented for the angular position of actuators within the robot's legs. A torque control method is also described, akin to using proportional and derivative control of the angular position of the actuators. Finally, a Zero Moment Point based capture step is described and implemented within simulation. With torque control alone, the result is a stable bipedal recovery from disturbances along the saggital plane of up to 11.25N of force, from a standing pose. In comparison, the previous implementation without dynamic stability leads to the robot falling after a minor disturbance of 2N. When capture step is included in the approach, the robot can recover from disturbances of up to 45N. The codebase is open-source and provides a humanoid robot simulation platform for research teams working in this area.