Page 318 - Introduction to AI Robotics

301
8.3 Control
Now consider a case where the daughters are not identical. Suppose that
one of the three micro-rovers has a different sensor suite and behaviors from
the other two. In that case c = 3, where p 1 = 1 , p 2 = 2 ,and p 3 = 1 .
4 4 4
Substituting into Eqn. 8.1 yields 1.5. Since 1:5 > 0:81 , the marsupial team
with the different daughter is more diverse that the marsupial team with all
identical daughters.

8.3 Control

CENTRALIZED Control of multi-agents can fall in a spectrum bounded by centralized control
CONTROL and distributed control regimes. In centralized control, the robots communi-
DISTRIBUTED CONTROL
cate with a central computer. The central computer distributes assignments,
goals, etc., to the remote robots. The robots are essentially semi-autonomous,
with the centralized computer playing the role of a teleoperator in a teleop-
erated system. In distributed control, each robot makes its own decisions
and acts independently. Of course, there is a range of regimes between fully
centralized and fully distributed; the robots can interact with a central con-
troller to receive new goals, then operate for the duration of the mission in a
distributed manner.
Examples of full and partial centralized control can be found by compar-
ing the RoboCup and MIROSOT robot soccer competitions. In those soccer
competition events, teams of robots are controlled remotely by a central com-
puter. In the small sized league of RoboCup and MIROSOT, teams of three,
very small self-contained robots (7.5cm x 7.5cm x 7.5cm) play on a 130cm
x 90cm arena with an orange golf ball serving as the miniature soccer ball.
Each robot had a unique pattern of bright colors to make it visible from the
overhead cameras, and the overhead camera is connected to a central pro-
cessor. The robots communicate with the central processor over a radio link.
In MIROSOT, the central processor commands each robot by supplying the
direction to move. In RoboCup, the central processor can give either explicit
directions or just locations of other robots and the ball, letting the robot’s on-
board behaviors generate the (one hopes) correct response. Fig. 8.3 shows a
view of the small-sized league from the 1998 RoboCup World Cup.
MIROSOT robots are more drone-like than their RoboCup counterparts,
since they are not required to carry any on-board sensing. They represent
the extreme of centralized control, where everything must go through a sin-
gle computer, much like the battle-droids in the Star Wars movie, The Phan-
tom Menace. RoboCup robots are required to have some type of on-board