Sensor-based control versus neural network technology for the control of fast mobile robots behaviors in unstructured environments

Abstract

The authors address the problem of reactive behaviors for collision avoidance of fast mobile robots moving in unstructured and dynamic environments. Two approaches have been developed, simulated and implemented on fast indoor and outdoor vehicles. The first approach is based on the original description of a virtual deformable zone surrounding the robot. The deformations of this risk zone, due to the intrusion of proximity information, control the robot reactions. The second approach is based on neural network technology. The network inputs are the measured distances of obstacles in the path of the robot and its velocity. Its outputs are the controls of the robot acceleration, direction, and braking actuators.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>