This article presents a neural based autonomous system applied to mobile robot navigation. A coordination repertory balances instinctive behaviors (target seeking and collision avoidance), generated by two neural structures. In the same hierarchical level of the coordination repertory, there is other neural repertory, a neural-fuzzy network, that performs two main functions: a self-evaluation of the autonomous system and storage of the most recent landscapes captured by sensors. The autonomous

more »

... ystem learns to navigate the robot, based on the conditioning theory, while interacts with the environment. Computer simulations show that the autonomous system is able to learn a general navigation strategy that provides good trajectories (without collisions and taking targets) for environment configurations even different from that considered during learning period. Furthermore, we consider a simpler version of the autonomous system (without the neural fuzzy network repertory) in comparison experiments. The experiment results show that the autonomous system escapes from "u" configurations, but the simpler version does not.