Research On Multi-robot Cooperation In RoboCup MSL

This thesis , based on the problem of Multi-robot system cooperation in RoboCup MSL(Middle-Size League), is dedicated to the two problems: Multi-robot system cooperation mechanism and tactics decision-making.For the problem of multi-robot coordination, a hybrid coordination mechanism is proposed based upon the characteristic of RoboCup MSL environment. Compared with a single robot, cooperating with a number of robots to complete a particular task is an advantage of the Multi-robot system. Design appropriate coordination mechanism for multi-robot systems is very important to improve the overall performance. RoboCup MSL robots operate in dynamic, hostile and uncertain environment against unknown, competing robots. The robots have only local information from sensor and unreliable communications. These conditions require the coordination mechanism with the ability of strong adaptability and self-repair capacity after the failure of cooperation. Integrated with the advantages of centralized and distributed multi-robot system, a hybrid coordination mechanism was proposed with the characteristic of "global distribution and local concentration". Global distribution means that role assignment is distributed, so the problem of role assignment conflict that rooted in the disagreement of local information is solved, and cooperation based on social rules is realized. Local concentration means that in local area, attacker robot analysis competition state through tactics decision-making system, designate one teammate to complete tactics coordination. This coordination mechanism can improve overall performance of Multi-robot system with low time and communication cost.In tactics decision-making system aspect, a FNN(Fuzzy Neural Network) based on Rough Sets is designed to realize tactics decision-making for Multi-robot coordination mechanism. Rough Sets can reduce knowledge and analyze attribute weightiness for result; Fuzzy Theory facilitates the introduction of human knowledge; Artificial Neural Network has self-learning ability and fault-tolerant features. FNN based upon Rough Sets has integrated all the features hereinbefore, which makes the network structure with low space complexity and fast network training. This method has been used in tactics decision-making system. As a result, human knowledge was described objectively and the tactics decision-making system meets the request on the whole.The studies in this thesis realize steady role assignment and basic tactics decision-making for Multi-robot cooperation. For the further research on Multi-robot system cooperation, especially in RoboCup MSL, they have some great significance.