Finding Pareto solutions of design parameter of run-ning robot leg using soft computing

Abstract

In this thesis, using running leg model, co-design method will be suggested. This method will show hardware optimization considering software parameter. Using nondominated sort genetic algorithm II, K-means clustering, and pre-knowledge, optimization was performed. NSGA II is nonlinear global search method to find global minimum. Initializing population, evaluation, selection, crossover, and mutation are basic principles to avoid local minimum for multi objective function optimization problems. K-means clustering is method to extract important feature or compress data. Pre-knowledge is used to suggest evaluation equation using intuitive method about stability and performance. System validation was performed to validate suggested optimization process and find important design parameter. Using calculated design parameter, Conforming validation of design parameters was also performed, and hypothesis supported considering hardware and software simultaneously, and optimizing robot leg hardware could help controller in semi parallel design process. Though used method and model were simple and restrict, it showed support for importance of co-design using design parameter of running robot leg. ⓒ 2015 DGIST