Goal-oriented biped walking based on force interaction control

This paper addresses the problem of modelling and control of a biped robot by combining Cartesian-based position and force control algorithms. The walking cycle is divided in two phases: single support, in which one leg is in contact with the ground and the other leg swings forward, and double support, in which the forward leg absorbs the impact and gradually accepts the robot's weight. The contact of the foot with the constrained surface is modelled through linear and nonlinear spring-damper systems. The proposed control approach is based on simple motion goals taking into account the reaction forces between the feet and the ground. The control algorithm is tested through several experiments and its effectiveness and robustness is discussed.