A Scalable Approach to Evolvable Hardware

Abstract

Evolvable Hardware (EHW) has been proposed as a new method for designingsystems for complex real-world applications. However, so far, onlyrelatively simple systems have been shown to be evolvable. In thispaper, it is proposed that concepts from biology should be appliedto EHW techniques to make EHW more applicable to solving complexproblems. One such concept has led to the increased complexity schemepresented, where a system is evolved by evolving smaller sub-systems.Experiments with two different tasks illustrate that inclusion ofthis scheme substantially reduces the number of generations requiredfor evolution. Further, for the prosthesis control task, the bestperformance is obtained by the novel approach. The best circuit evolvedperforms better than the best trained neural network.