Modified Jacobian Transpose Control of a Quadruped Robot

摘要

Quadruped robots have many advantages over other robots. Their excellent potentials to traverse different terrains are one of main advantages which distinguish these robots from wheeled mobile robot. In this article, the gait planning problem and also the design of an appropriate controller for a quadruped robot will be investigated. To this end, an explicit dynamics model of an 18-DOF quadruped robot will be derived. Then, a constraint elimination method is proposed to cancel the contact forces from dynamics equations and also express the dynamics equations in terms of the independent variables of the motion. In the walking gait, these DOFs are the position and the orientation of main body and also the joint angles of the swing leg. A stable COG trajectory generation algorithm based on the ZMP approach will be proposed. The path of swing legs is planned to prevent any impact between the swing legs and the ground. To apply the generated path on the robot, Modified Transpose Jacobian (MTJ) controller is used. This controller is based on the feedback linearization approach. It uses the command controller in the previous time step to increase the controller performance. The simulation is conducted to evaluate the performance of the gait planning algorithm and also the controller. Obtained results verify the merits of the proposed algorithms.