An investigation of walker/terrain interaction

Abstract

There will be an increased need for walking robots to locomote over unstructured terrain. Walker/terrain interaction underlies all of the stability and reliability issues associated with walking on such terrain. In this work modeling, simulation, analysis, and experiments are used to investigate and characterize this interaction in the context of reliable, autonomous walking on natural terrain. The important effects of natural terrain that affect walker/terrain interaction are the effects of ground compliance and supports that might fail due to slope failures, slipping off the edges of rocks, etc. The interaction is characterized and controlled through nominal control, reactive control, force redistributions, and stability measures. Nominal control strategies, which arc normally used to control the walker, are developed in the context of walking on natural terrain. A model for how vertical forces redistribute under a set of compliant feet due to body motion has been developed to analyze the planned motions. New stability measures that take into account the effects of compliant terrain have been developed for planning and monitoring planned motions. Reactive control algorithms have been developed to respond to anomalous events such as support failures, unexpected foot forces, and low stability. Walker/terrain issues that were investigated are combined to form a viable walking prescription, where the state of the walker is continuously monitored, and used to characterize the nature of the walker/terrain interaction. If the interaction is favorable, planned machine motions may be executed by using nominal control. If the interaction is unfavorable, feet are repositioned to place the walker in a more favorable stance. It may be necessary to accept some poor footholds in order for a walker to progress; in these instances the force redistribution models may be used to ensure the stability of subsequent motions. If anomalous conditions arise that may possibly affect the stability of the walker, reactive control is employed to respond to such events. The characterization of walker/terrain interaction and the resulting walking prescription will lead to improved ability of autonomous robots to locomote reliably on general, unstructured terrain.