Adaptive Impedance Control of Legged Robot

This paper presents an efficient compliance control scheme for legs of walking machines interacting with an uncertain environment and proposes an impact control based on neural system. The proposed method combines the technique of the indirect MRAC (Model Reference Adaptive Control) with the properties of self-learning neural nets. Present research is focused on the optimization of motion of mechanical systems with constraints on the contact with the environment in order to cover given contact quality requirements (minimal tendency to oscillations, contact force limitation, contact stability).