ON AUTONOMOUS NAVIGATION IN A NATURAL-ENVIRONMENT

This paper presents a task-oriented approach that we developed for the perception system of a cross-country autonomous robot. We introduce an adaptive navigation method, well suited for the characteristics of complex natural environments. This approach is based on a multi-purpose perception system that manages different terrain representations. This paper focuses on two perception functionalities required during navigation, along with the corresponding representations. These functions deal with the navigation planning, and the robot self-localisation., they have been integrated within the control architecture loaded on board of our mobile robot ADAM. Finally, results of the EDEN experiment, always on the way at LAAS, are presented.