Environment recognition system based on multiple classification analyses for mobile robot

Recently, various mechanisms have been developed combining linkage mechanisms and wheels, especially, the combination of passive linkage mechanisms and small wheels is one of main research trends, because standard wheel type mobile mechanisms have difficulties on rough terrain movements. In our research, a 6-wheeled mobile robot employing a passive linkage mechanism has been developed to enhance maneuverability and achieved climbing capability over a 0.20[m] height of bump. We designed a controller using neural network for high energy efficiency. In this paper, we propose an environment recognition system for the wheel type mobile robot which consists of multiple classification analyses. We evaluate the recognition performance by comparing Principle Component Analyses (PCA), k-means and Self-Organizing Map (SOM).