| Artificial legged locomotion may one day be superior to wheeled or tracked vehicles, especially for very rough terrain. Applications include unmanned ocean floor, planet, and volcano explorations. Also, the development of artificial legged locomotion may help in further understanding the control structure of animal locomotor systems. It has been shown that the rhythmic motions in animals are generated and controlled by networks of coupled nonlinear neuronal oscillators. This thesis was inspired by these observations.;During the past few years researchers have considered various locomotory control models with marginal success, warranting the need for more research at the most basic levels. As an initial step in the analysis of a quadruped robot, a simple biologically-inspired leg model with two degrees of freedom was considered. Poincare return maps were used to analyze three schemes for hopping height control: reflex control which involves energy pumping of the ankle joint, periodic forcing of the joint, and an adaptive periodic forcing scheme. Analytical approximations to the maps based on perturbation methods were derived. For all three approaches, explicit formulas for the hopping height and conditions for stability were obtained. In terms of stability, hopping height, and bandwidth requirements, the adaptive periodic forcing approach was shown to combine the best features of the other two approaches. A novel, experimental robot leg was designed and constructed, and experimental data supporting the analysis was obtained. Finally, a control scheme to potentially extend these results to forward hopping, and eventually to four legged robotic locomotion, was considered.;Both experimental and theoretical results demonstrated the advantages of incorporating an oscillator in the system to control the hopping height of a legged robot. In addition, the results obtained showed the usefulness of Poincare return map analysis, together with perturbation methods, of hopping gaits. |
