Information Technology Reference
In-Depth Information
considered as the similar states. Though the description of the state is not precise,
it is a description of the strategic level that agent can be running in the same
strategic region actively. So, (S
A
,S
B
, S
G
) describes a particular state, which SA is
the regional code of offensive team member A, S
B
is the regional code of
offensive team member B, and , S
G
is the regional code of offensive team
member G. The regional code is calculated as follows: S = i * 8 +j. And the states
are preserved by triples of the three regional code.
A
B
Fig. 10.10. Robot soccer world cup training, 2 to 1
19
0
•
7
Fig. 10.11. Position Partition
The optional actions have {Shoot, Pass, Dribble}, described as follows.
Shoot: the strategy is obtained by learning through a strategy based on the
probability of shot.
Dribble: the strategy is always to reduce threat, and pass ball to regions with a
high probability to shoot the goal. In order to achieve this strategic objective, the
offensive region can be divided into a number of strategic areas. In each strategic
area, shot evaluation is recorded with the shooting success rate.
Pass: strategy is very simple, just passing ball between any two agents, and do
not need to choose the target agent. If the pass fails, then that the state of
adoption of this strategy is unsuccessful; through this training, impossible path of
passing ball can not be adopted.
