Navigation Regimes for Off-Road Autonomy - Robotics Institute Carnegie Mellon University

Navigation Regimes for Off-Road Autonomy

PhD Thesis, Tech. Report, CMU-RI-TR-05-23, Robotics Institute, Carnegie Mellon University, May, 2005

Abstract

A new approach to high-speed, off-road navigation is presented. The interrelatedness of sensing horizon, prior map resolution, speed and efficiency is investigated over the space of off-road navigation. From this analysis, the space of off-road navigation is partitioned into three regimes (efficiency-limited, stop-limited and swerve-limited). Safeguarding through pre-planning and swerving emerges as an approach to achieve greater performance than possible with panic-stopping. The swerve-limited regime is characterized by high speeds that are achieved using plans derived in part from prior maps. Within the swerve-limited regime, sensing horizon is linearly proportional to robot speed. Thus an autonomous robot can operate at speeds beyond the point at which panic stopping is viable if it is safe to avoid obstacles by swerving. Navigating safely within this regime therefore explicitly requires some prior knowledge or assumption. The efficiency-limited regime is characterized by low speeds. Onboard sensing horizon is fixed given a specified prior map resolution and desired navigation efficiency. Typical applications include planetary exploration and mine-mapping For operation between the efficiency-limited and swerve-limited regimes, robot stopping distance limits speeds. In this stop-limited regime, sensing horizons increase quadratically as speed rises to ensure robot safety. Algorithms designed for navigation within the stop-limited regime generally require some cognizance of vehicle dynamics. Two implemented approaches to off-road navigation, one relevant for high-speed, long duration driving, and the other for planetary exploration, are contrasted and described in detail. These particular implementations exemplify the efficiency-limited and swerve-limited regimes and represent the state-of-the-art in planetary exploration and high-speed navigation.

BibTeX

@phdthesis{Urmson-2005-9177,
author = {Christopher Urmson},
title = {Navigation Regimes for Off-Road Autonomy},
year = {2005},
month = {May},
school = {Carnegie Mellon University},
address = {Pittsburgh, PA},
number = {CMU-RI-TR-05-23},
}