Experimental Realization of Deadbeat Control on a Hybrid Model of Legged Locomotion

Abstract

In this thesis work, I validate the deadbeat controller on a one-legged hopper by experiment, which includes the construction of the test bed, controller build-up, adjustment of dynamics for the controller and experimental validation for the controller. The mechanical test bed is designed and built for the one dimensional hopping experiment with changing terrain to validate the deadbeat controller. The deadbeat controller is generated by parameter optimization on the Poincare Map, which is provided by [9] based on the analysis of running animals whose dynamics can be partially or fully controlled by the encoded control laws in the feedforward trajectories of the appendages. The controller is able to achieve deadbeat stability of the apex by controlling the dynamic with only sensing of apex heights. Based on dynamics of the hopper, the method of decreasing energy dissipation is applied that is to attach springs on the leg. By exploring the hopper, the return map of terrain disturbance and control input is measured in physical experiment. Finally the controller is validated on the experiment of hopping on the changing terrain with small terrain feedback sensing.