Brain-inspired modular controller with fuzzy module selection

Human capability in dealing with variant dynamic conditions and performing successive different tasks in usually non-stationary environments, has got the neuroscientists' attention to discover the mechanism of human neural system which performs motion control. One of these studies, suggests a modular structure, MOSAIC, to learn and control variant dynamic conditions by switching multiple modules. In this paper, we present an easy understandable fuzzy mechanism to select appropriate module which is an important problem about the switching, according to the dynamic condition. Although there have been some extensions of original MOSAIC, the previously proposed approaches either result in unstable module switching or exploit biologically unjustifiable probabilistic methods. In addition to the fact that fuzzy approach is greatly compatible with human brain behavior, the simulation results show that fuzzy module selection outperforms the probabilistic approach used in original MOSAIC.