Information Technology Reference
In-Depth Information
virtual sensors for each robot and sends the measures of these sensors to the robots
periodically. The most important sensor is the vision sensor, but there's also a sense
body sensor (that informs the player of its stamina and own speed) and a hearing sen-
sor. Sensory data is in general subject to noise or to some other type of pre-processing
that precludes the agents from knowing the exact value of the measures.
The simulation advances in steps of 100ms, meaning that every 100ms the posi-
tions and velocities of every player and of the ball are updated by the simulator. Some
of the sensory data (like some modes of the vision sensor) is sent to the agents with a
different period than that of the simulation update.
Each agent controls only one virtual robot and must coordinate its efforts to make
its best contribution for the teams' goals. It is important to note that the knowledge of
the various agents about what is happening at a certain moment is not identical, due to
noise and restrictions on several sensors (like the angle of vision or the cut-off hearing
distance). Also the environment is very dynamic with the opposite team controlling
their robots to oppose the teams' goals.
The coach agent receives, from the simulator, the positions of all players in the
field and of the ball without noise. However, the coach cannot handle the ball and as
severe limitations on its communication with field agents, that make it impossible to
control the field robots using the coach information. The coach may have a very sig-
nificant impact on team performance by giving advice to the field players and using it
to perform high-level tasks like tactic analysis and selection or opponent modelling.
Visualization of the games is assured by an independent application that communi-
cates to the simulator to receive the players and ball positions. Fig. 1 show two possi-
ble visualizations of the games in the 2D simulation league. All the 3D features of the
3D viewer in Fig. 1 are not modelled in the simulator but inferred by the viewer for
better attraction.
Fig. 1.
2D Simulation League Traditional Viewer (left) and 2D Simulation League Viewer with
3D displaying capabilities ([43]) (right))
2.3
Simulation 3D League
The first version of the Simulation 3D league simulator [44] was made available to
the RoboCup community during January 2004. The proposal of the 3D simulator had
the following objectives:
