Biologically Motivated Control of Walking Machines

Abstract

Walking robots represent a field of increasing research activities in the lastyears. Especially the ability of walking machines to adapt to unstructuredterrain and the resulting requirements to the control architecture are empha-sized by the researchers. These efforts can be separated into two differentapproaches, one being the classical engineering approach using and refiningthe classical methods of feed-back control structures and dynamic modellingto control the robot, e.g. [1] and [2]. The other way is to adopt as much frombiological paragons for locomotion as possible regarding both mechanical de-sign and control architecture, [3] and [4].Biomechanical research of the last years has identified several key elementsbeing used in nature for adapting locomotion. These range from the geomet-rical structure of legs [5] and dynamic properties of muscles ( [6]) to neuralnetworks used for walking by insects [7] and [8]. The results of this researchimply a lot of benefits in case of a transfer of these principles to legged robots.Due to the high complexity of real walking machines and the impracticality ofmimicking especially nature’s actuators and sensors, up to now only some ofthe ideas have been transferred to the control architectures and the design ofreal robots.In this paper we give an introduction to the results we gained in the areof biologically inspired walking machines over the last years