Page 323 - Introduction to AI Robotics

306
8
Multi-agents

Figure 8.4 The Nerd Herd. (Photograph courtesy of USC Interaction Laboratory.)

time. The robots were placed randomly on the same side of the partition and
started moving at the same time.
IGNORANT In the first set of demonstrations, the robots functioned with ignorant coex-
COEXISTENCE istence. The robots coexisted in a team, but did not have any knowledge of
each other. A robot treated another robot as an obstacle. Each robot had the
equivalent of a move-to-goal and an avoid-obstacle behavior. Since robots
were treated as obstacles, once the robots gathered at the opening, they spent
most of their time avoiding each other. The team as a whole made slow
progress through the door to the goal location. Worse yet, the larger the
number of robots fielded, the larger the traffic jam, and the longer to get all
the team members through.
INFORMED In the second demonstration, informed coexistence, the robots were allowed
COEXISTENCE to recognize each other and given a simple social rule governing inter-robot
interactions. In addition to move-to-goal and avoid-obstacle, a third behav-
ior was created for avoiding robots. If a robot detected another robot, it
would stop and wait for time p. If the blocking robot was still in the way
after p, the robot would turn left and then resume moving to the goal. The
result of the new behavior was to reduce the traffic jams, and the group got
through the door in about the same time as a single agent going back and
forth through the opening 20 times.
INTELLIGENT The real surprise came in the third demonstration, intelligent coexistence.
COEXISTENCE