Novel inverse dynamics control strategy with different phases for the quadruped robot

Abstract

Aiming to reduce the computation and implement compliant control, this paper proposes a novel inverse dynamics control strategy based on the floating-base rigid body system. The control strategy assumes that each leg of the quadruped robot organizes itself into an independent autonomous system, a serial robot. Based on this assumption, the kinematics and the dynamics models of the quadruped robot have been created. The dynamical model supposes two different models according to the leg's state. In the stance phase the serial robot affixes its base frame to the shank and iterates the rigid body dynamics algorithms from the knee joint to the body. When the serial robot is in the swing phase, the dynamics algorithm is propagated from the hip joint to the shank, whose computing direction is just the reverse against the direction of which the serial robot is in the stance phase. The quadruped system doesn't need the fixed base to the system and avoids calculating the virtual joints of 6-DOF. Therefore, the algorithm proposed in this paper makes real-time computation of the quadruped robot dynamics possible. In order to evaluate the efficiency of the inverse dynamical control strategy, experiments are accomplished based on a practical quadruped robot. The experiments, which were done on a rubber mat and on asphalt, demonstrated that the quadruped robot is able to walk adaptively.