01-01-2013, 12:57 PM
CYBER-ATVs: DYNAMIC AND DISTRIBUTED RECONNAISSANCE AND SURVEILLLANCE USING ALL TERRAIN UGVs
DYNAMIC AND DISTRIBUTED RECONNAISSANCE AND SURVEILLLANCE.pdf (Size: 452.34 KB / Downloads: 39)
INTRODUCTION
Reconnaissance and surveillance are tasks that generally
require the simultaneous coverage of large areas. A system
based on mobile robotic technologies, in conjunction with a
distributed and collaborative multi-agent architecture, seems
to be a suitable candidate to perform this task. In the
Cyberscout project at Carnegie Mellon University we are
creating such a system. The focus of the project is to develop
mobile robotic technologies and unattended ground sensors
that will extend the sphere of awareness and mobility to
perform security operations.
In order to increase mobility and sensory coverage, we have
developed several robotic platforms with different sizes and
capabilities ranging from Microrobots (<5x5x5 cm3) to All
Terrain Vehicles (CyberATVs). The central idea is that a user
can task this heterogeneous group of platforms as a single
logical entity. Users task, control, and monitor the different
platforms through CyberRAVE (Real and Virtual
Environment), a software framework and graphical user
interface which enables rapid configuration and prototyping
of cooperating groups of real and virtual robots. This paper
describes our work on robotics All Terrain Vehicles or
CyberATVs. Further details about CyberRAVE and the
overall architecture can be found in [1].
Computer System
A PC/104 unit based on a 586-133MHz processor performs
low-level control of all the locomotion capabilities. Through
optoisolated analog/digital I/O channels, this processor is
able to receive each of the feedback signals and to control all
the actuators.
The high-level planing and perception is performed by 3
motherboards based on PentiumII-350MHz processors. These
motherboards are mounted on a custom-built housing, and
connected in a start configuration via Fast Ethernet
(100Mbps). General-purpose motherboards were preferred
instead of a parallel computer or DSP due to the higher
performance/cost ratio and easier programming.
SOFTWARE
The software of the ATVs is based on a distributed multiagent
architecture. Agents are independent processes that run
concurrently and can be started on an as-needed basis. The
architecture is totally decentralized, so that the interaction
between agents emerges from the environment or mutual
constraints. This scheme allows for an easy scalability of the
system through the incorporation of new agents to provide
new services. Figure 2 shows the main components of the
software architecture of the ATVs. In the figure, each of the
modules corresponds to a collection of agents.
OPERATION MODES
Figure 3 shows the current modes of operation of the
CyberATVs. Dark and light squares differentiate between
modes that we have completed, and modes that are still in
progress. Most of these modes have been already described in
previous publications [1][2][3][4]. Here we concentrate on
our current progress in autonomous navigation, giving a brief
description of the rest of the modes.
AUTONOMOUS NAVIGATION
At present we have implemented two modes of operation
with autonomous navigation: the GPS waypoint navigation
described above, and a Safe Wandering mode. In Safe
Wandering mode the ATV periodically choses a random
close GPS position as a new goal, and it tries to reach it. This
section describes the methods used to achieve these
autonomous navigation modes of operation.