The main research topics to be addressed are:

Nonlinear hybrid feedback control

Systems coordination

Sensor fusion and navigation

Clearly these topics arise from the laboratory area of interest for Control Systems and in particular their application in mobile robotics.

The team previous experience in mobile robotics [PoRobot], and current participation advanced mobile robotics control projects [LSTS], provides an adequate background for the specific problem solving in this case of robotic football.


Localisation and Control
In the Robocup to achieve good team play capabilities, is required a good robot localisation and control. The self-localisation is mainly done by the fusion
of vision measures of world landmarks (goals, corner and ground marks), internal
sensors (odometric and magnetic compass) and external vision measures, using
Kalman filtering methods. We plan to test fusion algorithms based in covariance
intersection. Furthermore, each robot must have a world state with some
knowledge of position, attitude and his derivatives, with some uncertainty measure for all the game moving objects. This is accomplished with distributed sensor fusion, where the vision sensors play a key role in sensing. Also, a distributed dynamic camera allocation and managing is under study. In the presence of are communications problems, the desired information must be perceived individually by each player.In the last case, in spite of an obvious degradation of the world model, some team coordination must be accomplished by the perceived information of our robots. The motion control architecture approach is based in atomic parameterised hybrid feedback controller, also known as manoeuvre.
These controllers incorporate both continuous and event driven feedback. This approach involves the atomic parameterised hybrid controllers synthesis, where are defined and implemented a set of manoeuvres solving specific classes of motion problems which are classified according to the patterns of the associated constraints and objectives. Those manoeuvres are the resources to the coordination level. The set of manoeuvres that are being synthesised are:


Motion without ball:

Move to location avoiding obstacles
Block goal path
Approach ball with defined attitude
Move to maximise target vision information

Coordinated robot and kicker motion:

Smooth ball reception
Ball guidance
Interception

Strategy and coordination architecture
The team game evolution is coordinated in structural way, by defining the tactical function (goal keeper; defence; middle field; attack) for each robot .The robot adopts the corresponding tactical policy accordingly to its perception about the companions in the field. The player has one main place in the overall strategy (as consequence of its tactical position), but this can be reconfigured dynamically. Only one robot has the possibility to adopt the goal keeper position. The overall strategy solution results from the composition of the decisions taken in a distributed way by the operational robots in the field. The analysis of the vectors: game phase, ball possession and current topology formation in conjunction with the current robot game role and its perceived topological position determines the coordination level for each robot. An evaluation of the next action to be taken is made by the maximisation of hypothesis success. The coordination level is implemented by a modular and distributed controller synthesis trough the composition of discrete observers (corresponding to the analysis vectors) and a discrete controller parameterised by the adopted tactical functionality.