The goal of this paper is to present the implementation and the validation of a local navigation of a fast mobile robot using local grid representation of an unstructured environment. After a presentation of our fast mobile robot based on transputer devices, we present a local modeling of unstructured environment. This polar grid modeling fits with an original use of ultrasonic sensors. A fusion method based on fuzzy logic and an integration of robot motion complete this local modelization. In a third part, we present a local navigation that can take into account nonholonomy and dynamic constraints. Based on a time horizon and escape lines, the local navigator gives a pilot only admissible trajectories. The last part describes the implementation of the two parts together, and discusses results obtained on a real fast mobile robot: Snake II.
展开▼