Page 23 - Autonomous Mobile Robots

Visual Guidance for Autonomous Vehicles 7

are given from stereo vision and image–ladar integration. The chapter ends
by returning to the road map in Section 1.4 and examining the potential role
of visual sensors in meeting the key challenges for autonomy in unstructured
settings: terrain classification and localization/mapping.

1.1.2 Classes of UGV
The motivation or driving force behind UGV research is for military application.
This fact is made clear by examining the sources of funding behind prominent
research projects. The DARPA Grand Challenge is an immediate example at
hand [2]. An examination of military requirements is a good starting point, in
an attempt to understand what a UGV is and how computer vision can play
a part in it, because the requirements are well defined. Another reason is that
as we shall see the scope and classification of UGVs from the U.S. military
is still quite broad and, therefore, encompasses many of the issues related to
autonomous vehicle technology. A third reason is that the requirements for
survivability in hostile environments are explicit, and therefore developers are
forced to face the toughest problems that will drive and test the efficacy of
visual perception research. These set the much needed benchmarks against
which we can assess performance and identify the most pressing problems.
The definitions of various UGVs and reviews of state-of-the-art are available in
the aforementioned road map [1]. This document is a valuable source for anyone
involved in autonomous vehicle research and development because the future
requirements and capability gaps are clearly set out. The report categorizes four
classes of vehicles with increasing autonomy and perception requirements:
Teleoperated Ground Vehicle (TGV). Sensors enable an operator to visualize
location and movement. No machine cognition is needed, but experience has
shown that remote driving is a difficult task and augmentation of views with
some ofthe functionality ofautomatic vision would help the operator. Fong[3]is
a good source for the reader interested in vehicle teleoperation and collaborative
control.
Semi-Autonomous Preceder–Follower (SAP/F). These devices are envis-
aged for logistics and equipment carrying. They require advanced navigation
capability to minimize operator interaction, for example, the ability to select a
traversable path in A-to-B mobility.
Platform-Centric AGV (PC-AGV). This is a system that has the autonomy
to complete a task. In addition to simple mobility, the system must include extra
terrain reasoning for survivability and self-defense.
Network-Centric AGV (NC-AGV). This refers to systems that operate as
nodes in tactical warfare. Their perception needs are similar to that of PC-AGVs
but with better cognition so that, for example, potential attackers can be
distinguished.

© 2006 by Taylor & Francis Group, LLC

FRANKL: “dk6033_c001” — 2006/3/31 — 16:42 — page7— #7