WalkNet — a Decentralized Architecture for the Control of Walking Behaviour Based on Insect Studies

Abstract

A network model for controlling a six-legged, insect-like walking systemis described, which is based as far as possible on data obtained frombiological experiments. The network contains internal recurrentconnections, but important recurrent connections utilize the loopthrough the environment. This approach leads to a modular structure,WalkNet, consisting of several subnets. One subnet controls the threejoints of a leg during its swing which, is arguably the simplestpossible solution. The task for the stance subnet appears more difficultbecause the movements of a larger and varying number of joints have tobe controlled such that each leg contributes efficiently to support andpropulsion and legs do not work at cross purposes, i.e. do not produceinteraction forces. This task appears to require some kind of ``motorintelligence{''}. We show that an extremely decentralized, simplecontroller, based on a combination of negative and positive feedback atthe joint level, copes with all these problems by exploiting thephysical properties of the system.