A Position-Domain Adaptive Control Method for Underactuated Bipedal Walking on a Compliant Ground

Abstract

Motivated by the potential use of humanoid-robot in real environment, a position-domain adaptive control strategy is developed to stabilize the underactuated bipedal walking on a compliant ground. First, the robot-ground system is modeled as a rigid kinematical chain coupled with a spring-damper system. Then by observing the simulation result of walking on compliant ground, we find the improvement of walking stability can be realized by controlling the initial velocity of robot’s center-of-mass (COM) of each walking cycle. In consideration of the highly-complicated impact of real road surface on direct velocity control, through the analysis on the relationship between the robot’s COM velocity and its foot vertical velocity, the robot’s state is parameterized by the normalized relative vertical position between both the feet, and the control of robot’s COM initial velocity of current cycle is realized by controlling the relative vertical velocity between the robot’s two feet during the previous impact phase. To realize it, an adaptive feedback linearization control strategy is developed in position-domain. Finally, the availability and adaptability of this method are validated through simulations: Specific to three initial gaits, on four compliant ground with different damping parameters, the underactuated bipedal walking is stabilized and the performance is improved.