Learning in a neuro-fuzzy navigator for robotic manipulators

Presents a fuzzy-based navigation system for robotic manipulators. The fuzzy rules combine a repelling influence, related to the distance between the manipulator and nearby obstacles, with an attracting influence produced by the angular difference between the actual and final manipulator configurations, in order to generate the actuating motor commands. The use of fuzzy logic leads to a transparent system that can be tuned by hand or by a learning algorithm. The proposed learning algorithm can be adapted to the particular requirements of a given manipulator, as well as to the environment it operates in. The navigation method has been successfully applied to robot arms in different environments, giving encouraging results.