A neuro-mechanical model of legged locomotion: single leg control

Abstract.

Most models of legged locomotion have concentrated on properties of either the mechanical or the neural system. Here a combined neuro-mechanical model of stepping in a single leg is presented, as the first step in the process of modeling and building a fast and dynamically stable quadruped. It is based on general principles of legged animals with special reference to vertebrates. The mechanical leg was first studied separately in order to take advantage of its inherent mechanical properties and avoid over-control during stepping generation. As a part of the design strategy it uses elastic actuators to increase shock tolerance and energy efficiency. The neural controller consists of a neural phase generator (NPG), a system of fast feedback pathways, and a single control neuron representing descending drive from higher centers in the brain. Sensory information directly influences the movements through the fast feedback pathways, but also entrains the NPG. The NPG has its own description of the state of the leg, which then enables it to set the feedback pathways so that only actions appropriate for the particular stage of the step cycle are undertaken. This preprogramming benefits from the NPG's ability to filter out any inconsistencies or gaps in the afferent input. In this way the model unites the use of central pattern generators and peripheral feedback systems for the generation of stepping movements. The neuro-mechanical system produced stable stepping patterns over a large velocity range and was adaptable to different body weights and landing from varying heights.