Page 314 - Introduction to AI Robotics

297
8.2 Heterogeneity
Each robot was programmed with a sequence of simple reactive behav-
iors (renamed here for clarity), following the reactive layer of Arkin’s AuRA
architecture described in Ch. 4:
wander-for-goal This behavior was instantiated for two goals: trash and
trashcan. The motor schema was a random potential field, the perceptual
schema was color blob detection, where trash=“red” and trashcan=“blue.”

move-to-goal This behavior also had two different goals: trash and trashcan.
The motor schema was an attractive potential field, and the perceptual
schema for the trash and trashcan were the same as in the wander-for-
goal.

avoid-obstacle This behavior used the bump switch as the perceptual sche-
ma, and a repulsive field as the motor schema.
avoid-other-robots The three robots did not communicate with each other,
instead using only the repulsive field created by avoid-other-robots to re-
duce interference. The motor schema was a repulsive potential field (lin-
ear dropoff), while the perceptual schema detected “green.”

grab-trash The robot would move toward the trash until the perceptual sche-
ma reported that the IR beam on the gripper was broken; the motor schema
would close the gripper and back up the robot.

drop-trash When the robot reached the trashcan with trash in its gripper, the
motor schema would open the gripper and back up the robot, and turn 90
degrees.

8.2.2 Heterogeneous teams

A new trend in multi-agents is heterogeneous teams. A common heteroge-
neous team arrangement is to have one team member with more expensive
computer processing. That robot serves as the team leader and can direct the
other, less intelligent robots, or it can be used for special situations. The dan-
ger is that the specialist robot will fail or be destroyed, preventing the team
mission from being accomplished.
One interesting combination of vehicle types is autonomous air and ground
vehicles. Researchers as the University of Southern California under the di-
rection of George Bekey have been working on the coordination of teams
of ground robots searching an area based on feedback from an autonomous