Motion Planning for a Bounding Quadruped Robot Using iLQG Based MPC

Abstract

When tackling motion control problem of the quadruped bionic robot with traditional control method, it is more and more difficult to improve the control effect due to the uncertainty caused by the factors such as model mismatch, distortion, and interference. Therefore, the demand for utilizing advanced control method to realize the motion control for legged bionic robot is growing. Among various advanced control methods, the model prediction control (MPC) has low requirements on model accuracy, and the system has good robustness and stability, which provides a good prospect for solving the balance problem of dynamic robots. In this paper model predictive control is used to realize the bounding gait motion control for a quadruped bionic robot. The MPC planner automatically generates the foot end position and contact force trajectory of the robot, so that the robot can track the desired motion state. However, as the model of the quadruped bionic robot is nonlinear, the generated MPC algorithm may be difficult to meet the real-time requirements. In order to solve this problem, this paper uses the iLQG optimization algorithm in the design of a model prediction controller and adds heuristic-based reference input. The comparison between the simulation and the classical SQP algorithm shows that the MPC based on the iLQG optimization algorithm has good convergence and tracking effect for the quadruped robot running, and has fast real-time performance.