A learning procedure for a fuzzy system: application to obstacle avoidance

The goal of this work is to propose a learning procedure for fuzzy systems. Fuzzy systems are able to treat uncertain and imprecise informations. They have a capability to express knowledge in the form of linguistic rules. Their drawbacks are caused mainly by the diiculty of deening accurate membership functions and lack of a systematic procedure for the transformation of the expert knowledge into the rule base. Neural networks have the ability to learn but both knowledge extraction and knowledge representation are diicult. First, a neuro-fuzzy architecture is proposed. A learning procedure based on the stochas-tic approximation method is described. The methodology is the supervised learning method developed in the eld of neural networks. In order to discuss the validity of the proposed method, three numerical examples are presented and it is shown that the proposed neuro-fuzzy system have the ability to learn. It is applied to the obstacle avoidance problem of a mobile robot. As an experimental platform, the Khepera mobile robot is used. To conclude, the application of a neuro-fuzzy controller for Khepera is discussed.