Page 324 - Introduction to AI Robotics

307
8.6 Emergent Social Behavior

Figure 8.5 Robots cooperatively tracking an object under the ALLIANCE architec-
ture. (Photograph courtesy of Oak Ridge National Laboratories.)

The social behavior for avoiding robots was replaced with another heuristic:
the robots were repulsed from other robots, but as it moves away, it tries to
move in the same direction as a majority of other robots. (Each robot broad-
cast its heading over a radio transmitter to compensate for the inability to
recognize each other by vision or sonar, so that isn’t considered communica-
tion.) As a result, the robots exhibited a flocking behavior and went through
the door in single file! The need to go in the same direction created a ten-
dency to form a line, while repulsion caused the robots to essentially create
spaces for robots to merge into line. Together the two effects created a strong
need to go through the door single file, even though there was no such ex-
plicit direction. Not only were traffic jams reduced, but the overall task was
accomplished faster.

8.6.2 Motivation
In Mataric’s work, the robots reduced interference through simple social
rules with no communication, but the members of the team could not ac-
tively help out failed colleagues or change tasks dynamically. Lynne Parker
has attempted to address the larger issues of robustness and fault tolerance
with the ALLIANCE architecture, 114 an outgrowth of the Subsumption ar-
chitecture. The central idea is that members in the team can either observe or