A fast and smooth walking pattern generator of biped robot using Jacobian inverse kinematics

In order to solve inverse kinematics of a multi-DOF (degree of freedom) mechanism, many methods have been proposed with the Jacobian linearization method. When solving inverse kinematics problems with this method, long computation time is needed because the Jacobian matrix should be updated in order to solve the configuration for each different end-effector trajectory knot. In this study, two smooth trajectories are generated as target positions, one for swing leg’s ankle, and the other for center of mass. These generated knot points in the task space with appropriate distance to each other are used to solve inverse kinematics by the proposed modified Jacobian method—Fixed leg Jacobian. It can guarantee that only one iteration is needed to solve the configuration when it is away from singularity with a small position error (0.0712% of leg length). We propose an algorithm that can generate the gait in real time including singularity avoidance and joint limit avoidance. Simulations have been carried out. The results showed that the proposed method can generate a smooth gait for robot walking on real time implementation.