Sliding Mode Control of Underactuated Five-Link Biped Robot for Climbing Stairs Based on Real Human Data

The aim of this paper is to develop a robust control algorithm for a five-link biped robot with one degree of underactuation to climb stairs in a natural manner. The biped consists of a five-link planar robot with two actuators at the hip and another two at the knee joints. A robust sliding mode controller is proposed to track gait trajectories while climbing stairs. The control signals are computed using first-order sliding surfaces as a linear combination of the tracking errors and their time derivatives of the four actuated joints and those of the underactuated ankle joint of the stance leg. In addition, the controller parameters are optimized using Genetic Algorithm. Since human sensorimotor controls are done in an optimal way following the principle of optimality, gait trajectories data of the human subject are captured by Kinect sensor during climbing stairs to be used as reference trajectories. For the subject, the hip and swing limb ankle trajectories are captured and filtered. Desired joint trajectories are obtained as a function of those ankle and hip trajectories using inverse kinematics. The results prove that the biped accurately follows the desired trajectories during climbing stairs.