| Summary: | A Multi-Robot system can benefit from cooperative task assignments to achieve a common goal, as well as from merging information from all agents to form a better World Model. Highly accurate and responsive perception is still a fundamental step for effective cooperative robotics teams, especially in dynamic, stochastic and partially observable environments like the ones provided by the RoboCup soccer competitions. This research project was developed in the context of the CAMBADA RoboCup Middle-Size League robotics soccer team, from University of Aveiro, including the participation in local and international RoboCup competitions. While most of the literature presents simulated results, this project used the MSL competitions as a bench-marking and verification test-bed for the new solutions proposed in this document. The work aimed at developing new solutions for data fusion in a distributed approach in this team, as well as improving the existing high-level coordination strategy. In order to do so, the project starts by consolidating the robots local state estimation, providing an essential solid baseline for more advanced techniques - the main control cycle was revised, the robot pose and velocity state estimation was enhanced, and a new method for multi-object tracking has been proposed. A distributed sensing multi-object tracking module was built on top of that, providing a unified representation of the opponents on the soccer field, not only for the field players, but also for the coach agent, that is not able to use any sensors. Following, agent coordination techniques are discussed and a new method for leader election is proposed, tested and validated. With the enhanced world model, the leader election method and the matured modular software architecture in place, it was possible to develop a new set-play engine (CR7) that is user-friendly, robust and performant in competition environments. In a bottom-up approach, this project has incrementally touched a wide range of areas and has successfully met the initial objectives as well as helped to reinforce CAMBADA's position as one of the world-leading MSL teams.
|
|---|