Information Technology Reference
In-Depth Information
(e.g. dribble) that a player should be able to execute. When applicable,
conditions
and
actions
can make use of
geometric regions
(e.g. circles) to specify locations in the
field such as in Coach Unilang. During a match, these rules are conveyed to players as
messages in order to adjust their behaviors.
SFL [60] extends CLang by abstracting its low-level concepts to represent team
behavior in a human-readable format that can be easily modifiable in real-time. This
language lacks the ability to specify a team's complete behavior with enough detail.
Coach Unilang [11] fixes the inability of CLang to fully specify a team's strategic
behavior by enabling the transmission of different types of strategic information (e.g.
instructions, statistics, formation, opponent's information) based on real soccer con-
cepts. Players can ignore these messages, interpret them as orders (must be used and
will replace existing knowledge) or as advices (can be used with a given trust level).
3.3
Player-to-Player Communication
The standard inter-player communication language CommLang [61], proposed as a
complement to CLang in the RoboCup 2D simulated soccer league, promotes player
interoperability and eases the creation of mixed-teams of players. This work specifies
how to compose and encode different types of information into messages suitable for
transmission in the existing limited bandwidth communication channel. However, it
does not address i) which types of information should be sent; ii) how often messages
should be sent; nor iii) how the information received should be used. Each piece of
information is represented by a uniquely identified message type. Current types of
messages convey the following information about the sender's beliefs: i) objects posi-
tions and velocities; ii) ball ownership; and iii) passing synchronization. A communi-
cation message can include one or more types of messages. The use of this language
has been shown to improve the confidence and accuracy of a player's world states.
The trend in the robotic soccer domain will be towards little or no communication be-
cause it might not always be available, it can be costly and it introduces an overhead and
delay that can degrade players performance. The combination of implicit coordination
with beliefs exchange has been shown to yield better performance with communication
loss than explicit coordination with intentions communication alone [55]. The exchange
of beliefs among teammates allows the achievement of more coherent and complete
beliefs about the world. These beliefs can be used by a player to predict agents utilities
(including his own) and adapt his actions to their predicted intentions to achieve the best
(joint) action. As state estimation accuracy reaches an acceptable upper bound it will
eventually replace explicit communication for achieving coordination.
3.4
Coaching
In real life soccer, natural hierarchical relations exist among different team members
which imply a leadership connotation (e.g. a coach instructs strategy to players). A
coach and trainer are privileged agents that can be used to advise players during on-
line games and offline work out (training) situations respectively. Coaching provides
a means to monitor and aid the creation of players with
adjustable autonomy
[62] and
is an effective method to help a team improve its performance [63].
