Page 300 - Introduction to AI Robotics

Other Robots in the Hybrid Paradigm
7.7
7.7 Other Robots in the Hybrid Paradigm 283
One implicit criterion for evaluating the utility of a Paradigm and deriva-
tive architectures is its popularity. The Hybrid Paradigm describes many of
the robots being developed in laboratories today. It is clearly popular. A
large number of these robots fall into the Hybrid category by virtue of being
“neither fish nor fowl,” in this case being neither purely Reactive nor Hierar-
chical. Almost every robot has a set of functions that are equivalent to behav-
iors, though several have followed the lead of 3T in calling those functions
by other names (e.g., “skills”) to reduce the connotations of purely reflexive
behaviors. Even architectures that started out as Hierarchical, namely Albus’
1
Real-time Control Architecture and Simmons’ Task Control Architecture, 131
recast many of their functions into behavioral components.
One of the major influences on the mobile robot community has been the
DARPA UGV Demo II and Demo III projects, which forwarded the state of
the art in outdoor ground vehicle control and navigation. The tasks were
well-suited for a hybrid approach: the idea was to give a HMMWV military
jeep a map, a set of directions, and send it off. It would autonomously plan
a route, drive along roads, then go off roads as needed. In a military recon-
naissance mode, the HMMWV was expected to autonomously stay behind
hills, avoid trees and rocks, and covertly position itself so that Reconnais-
sance, Surveillance, and Target Acquisition (RSTA, pronounced “rist-tah”)
sensors could peek over and find enemy locations. Demo II even had multi-
ple HMMWVs traveling autonomously in formation.
The European Community ESPRIT agency has also significantly influenced
mobile robotics, sponsoring research in automating highway vehicles via ini-
tiatives in the 1990’s. The United States has shown some interest through
several iterations of “intelligent vehicle highway systems” programs. The
most notable of these automated vehicles were constructed by Dickmanns
and Graefe in Germany 45 and the CMU Navlab led by Dean Pomerleau. 116
A third influence on mobile robots has been NASA’s push for autono-
mous planetary rovers, which culminated in the Mars Pathfinder mission
with the Sojourner rover. Mapping a planet’s surface is intrinsically a delib-
erative function, as is planning paths for optimal exploration. Unfortunately,
from the Viking lander missions, Mars was known to be very rocky. The ap-
proaches quickly fell into camps. One, led by MIT, favored deploying many
small, legged robots, such as Genghis, that would climb over small rocks and
go around others. The “low-to-the-ground” viewpoint, which limits what
a robot—and therefore, the scientists on Earth—can sense, would be over-