Trajectory Generation for Biped Robot Climbing Variable Height Stairs Using MPC

Abstract

This paper proposes a method for generating a trajectory for a biped walking robot to climb and descend stairs of varying height using model predictive control. The robot is an unstable system because its two legs are not fixed to the ground when it is walking. It is very important to produce a proper trajectory to ensure stable robot behavior. But, biped robots have difficulty coping with uncertainties such as stairs and uneven terrain. Thus, an appropriate trajectory should be created that reflects the current state to ensure stability. In this aper, model predictive control was used as a method to solve this problem in real-time, and by using model predictive control, the biped robot creates a trajectory for climbing and descending stairs with varying heights in real-time. The motion of the robot climbing and descending stairs is nonlinear. So, nonlinear model redictive control is generally used. However, the nonlinear model predictive control method takes a lot of computation time and may cause problems such as a local minimum. To overcome the shortcomings of these nonlinear odel predictive controls, this paper used a linear model, considering the height of the stairs, for a model predictive control. The proposed method is verified by simulations.