Socially Optimal Coordination For Multi-robot Systems Based On Game Theory

Multi-robot cooperation is a key technology of multi-robot systems. Multi-robot cooperation based on game theory has become a hot research topic at home and abroad.Game theory is applied to multi-robot cooperation tasks in this thesis, and the concept of socialization is applied to robot group at the same time. Individuals can compete freely under some mandatory requirements, so the group has the social attribute. There are three main research contents: Multi-robot target tracking based on extended cooperative game can coordinate movement of the multi-robot system; multi-robot task allocation based on Pareto optimality can achieve Pareto optimal state both among task sets and in every task set; target hunting based extended cooperative game can achieve non-formation hunting.This thesis firstly proposes target tracking game model based on extended cooperative and complete information game theory. The concept of variable order of game is introduced to the model and the free competition mechanism among individual robots is realized. Thus a multi-robot group with social attribute is being set up. Then an in-depth research is conducted for robot tracking strategies based on the model. The tracking strategies are made in two phases, which are selection alone and selection with game, to reduce the computational complexity of the whole system. The experimental results show that the proposed target tracking method can coordinate multi-robot system’s motion effectively. Secondly, a task allocation method for multi-robot system based on Pareto decision is proposed. The multi-robot group is optimized by three allocations. The initial allocation problem is solved by the nearest distribution method. Then Pareto decision distribution method is used to reallocate tasks among robot task sets to optimize all task sets and thus reduce the cost. Furthermore, sequential game method is used to reallocate tasks in every individual robot’s task set to optimize the task execution sequence. The experimental results show that the proposed task allocation method is effective, and the performance is better than several other algorithms. Finally, a multi-robot hunting model based on extended cooperation game is established, and obstacle avoidance model and two kinds of target searching strategies-– roaming search and regional search are proposed. An effective escape strategy for the target is also presented and the whole process of hunting is analyzed. In order to increase the flexibility of the multi-robot system, non-formation movement is adopted in the target hunting method proposed in this thesis. The simulation and experimental results verify that the method is feasible.