A System for Fast Navigation of Autonomous Vehicles

Abstract

This report describes an autonomous mobile robot designed to navigate at high speeds over featureless terrain--terrain which is not navigable by relying on features that are found on paved roads and highways. To this end we have developed a paradigm that we call position based navigation that relies on explicit vehicle position information from inertial and satellite instruments, to navigate. Specifically we have tackled four main areas--path tracking, which guides the robot vehicle over a pre-specified path, obstacle detection, which is responsible for bringing the vehicle to a stop from high speed when an obstacle is detected, obstacle avoidance, which is responsible for steering the vehicle around detected obstacles so as to rejoin the specified path, and computing architecture, which integrates all the capabilities into one system. We discuss the algorithms and the devices that were implemented on NavLab, a navigation testbed at Carnegie Mellon. The most notable of our results is the 11m/s speed achieved by the vehicle.