Distributed Task Autonomous Allocation And Cooperative Control

In recent years, with the rapid development of multi-Agent technology, Agent autonomy in growing, in order to compensate for Multi-Agent centralized structure drawback system, Distributed Multi-Agent architecture is an important direction of development, but also the distributed architecture Multi-Agent technology to solve complex problems of distributed artificial intelligence decision-making is a research hotspot.In general, the decision of environmental problems often have sudden, uncertainty and dynamic characteristics, which requires decisions based on dynamically changing environment. In this paper, the dynamic changes in the environment, multi-tasking distributed architecture Agent under control technologies expand and collaborative research, elaborated in a dynamic environment task allocation and cooperative control characteristics.This article will question the decision-making complex tasks into dynamic environment task assignment and task execution layer layer two levels. Among them, the task distribution layer using random game theory to establish a dynamic task allocation model, and the corresponding algorithm, by calculating the distribution plan, optimize the allocation of tasks; task execution layer Krakow use horses to establish cooperative control model of decision-making theory and given the appropriate algorithm. By calculating synergy rules, during the execution of the task to guide collaborative action.In the task allocation layer, using reinforcement learning algorithm for stochastic game theory model to solve the task, algorithms, each Agent task to select the optimal value according to Nash-Q. When the dynamic changes in the environment that can dynamically adjust the repeated task allocation, task allocation can improve the effectiveness of multi-Agent.In the task execution layer, using SHV-IP MAS distributed algorithm for solving Markov decision model. SHV-IP algorithm first state in the model space between the initial state and the target state, one can find the shortest Hamiltonian path through all of the other immediate reward with higher status, the shortest Hamiltonian path through collaborative guidance to find the optimal strategy By avoiding the search for all states to reduce the state-space model of collaborative, cooperative control from a lower difficulty.In this paper, the game randomly distributed task allocation model and Markov decision model presented separately simulation results show that the stochastic game task allocation model and distributed Markov decision model presented in this paper the dynamic environment with good adaptability. Finally, we will process the dynamic allocation and task execution tasks in a dynamic environment combined with cooperative control is used to solve the problem of dynamic decision-making environment. Simulation results show that the proposed method can effectively adapt to the dynamic environment decision-making, decision theory and methods for dynamic study has some reference.