Adaptive hexapod gait control using anytime learning with fitness biasing

Abstract

Adaptive learning systems that generate con-trol programs for robots with varying capabili-ties is of importance in the implementation of autonomous robots. Learning done continu-ously with the best possible control program running the robot (anytime learning) can achieve the adaptability desired when imple-mented using some form of evolutionary com-putation. The difficulty with this method is that autonomous robots often lack the compu-tational power required to run evolutionary computation along with their control program. In addition, anytime learning usually requires input from internal sensors, which are not often available in small autonomous robots, to make adjustments for capability changes. In this pa-per, we propose an anytime learning system that employs off-line learning, using evolution-ary computation, with the control program be-ing downloaded to the on-line controller. The off-line learning does not require internal sen-sors but uses global observation (external overhead camera) to make the required ad-justments to guide the evolutionary computa-tion. The results of periodic tests, done on the actual robot, are used to bias the fitnesses cal-culated by the evolutionary computation, which uses a model of the robot. Experiments reported in this paper use a simulation of the actual robot (a more accurate model), while construction of the actual learning system is in progress. 1