Human Equilibrium Control Principles Implemented into a Biped Humanoid Robot

Abstract

We describe how human equilibrium control principles, which were derived from neurophysiological experiments, can be implemented in a biped humanoid robot (‘PostuRob’). Stance control in humans uses sensorbased feedback which involves mainly three sensors that measure: (1) body motion in space (vestibular system), (2) body motion with respect to the feet (ankle angle proprioception), and (3) torque between body and feet (ankle torque proprioception). The sensor signals are not used directly for feedback, but instead are used to internally reconstruct the physical stimuli in the outside world by means of sensor fusions. These reconstructions yield internal estimates of the external force fields such as gravity, the external contact forces (e.g. push), and support surface motion. The estimates are fed into an ankle angle proprioceptive feedback loop. By way of the signal fusions the robot’s control adjusts to the external stimulus situations, a fact that allows for low loop gain and robust control. This control differs from that of the biped robots described in the literature. They tend to use a global stability measure such as the COP (centre of pressure) or ZMP (zero moment point) and this control does not adjust to the external stimulus situation. Our PostuRob is used in a biorobotics approach to better understand stance control deficits of neurological patients and the effects of therapy and rehabilitation.