A new approach to design fuzzy controllers for mobile robots navigation

This article presents a new method to design, in two levels, fuzzy controller for reactive navigation of a mobile robot in a structured unknown environnment. At the rst level, adjacent sensors are grouped in areas and are used to deene local behaviors. These local behaviors are then gathered at the second level in order to deene a global behavior. Two experiments with diierent local behaviors and diierent mechanism of integration are presented on our Nomad200 mobile robot. 1 Motivations A key issue in research in mobile robotics is to design and to build autonomous robots able to act and move in a real environment, to achieve tasks like delivery , surveillance or cleaning. Mobile robots could also have a major role in exploration and intervention in a hostile environment (like nuclear plant, for instance). But, a mobile robot moving in a real environment has some problems to solve : The environment is vast and dynamic. Obstacles (or people) can move, appear or disappear. The set of all possible situations can not be computed. The mobile robot should be equipped with sensors (cameras, sonar or infrared) to have information about its local environment and act with this information. Due to the type of the ground and the slippage of the wheels, the actions are not completely reliable. Moreover, the values of the sensors have diierent reactions depending of the humidity of air, the temperature, the shape and the type of the objects included in the environment. As the robot moves in its environment and the environment changes, it has to react in a nite (often short) time to a new event. The goal of research in navigation is to nd a model which given the goal to reach and the data of the sensors , delivers the best action to apply to the mobile robot. As these data are noisy and the actions are not completely reliable, fuzzy control seems to be an appropriate way to control a mobile robot. Fuzzy control 4] is an application of fuzzy logic to the control of dynamic systems. 10] deenes the fuzzy logic as an extension of classic logic for representation and reasoning about approximate data. Several mobile robotic applications of fuzzy logic have been realized. For instance, 6]]8] use fuzzy logic to control a mobile robot in an unknown structured indoor environment. Unfortunately, fuzzy controller are sometimes diicult to design, because of : …