Study On Distributed Real-time Simulational System And Planning Algorithm Of Multi-agent System

With the develop of agent technology, the fuse of multi agent system and control theory and computer technique has been the new focus for science study. The characteristics of the multi agent system are distributed and collateral and strong fault tolerant ability. The distributed system is an important embranchment of the distributed agent study. The value of the theory for study is that it makes the foregone close and isolate knowledge system distributed intelligent knowledge system and develops the intelligent centralization mode to non-absolute distributed wisdom mode.Now, the study on multi agent system can be carved up two levels: one is the study on the basic theory and the other is the study on the specific system frame and software realization. The main study of the distributed simulational system has two classes: one is the distribution and collating of the simulational mode and experimental tasks and the other is the building of the distributed simulational environment. This paper designs an emulational test bed based on the multi-robot material flow system of the storages and docks and using the LAN to communicate through the reference of the existing advanced method and the integration of correlative concept and technology. The system can simulate environments for many tasks and the user can append the control algorithms designing by themselves and can realize the dynamic shows of the process.The cooperating problem of the multi agent system is the nuclear issue of the distributed artificial intelligence and the cooperation mechanism of the multi agent system is used to solve the management and optimization of the agents and is the key realization of the distributed multi agent cooperation system. This paper aims to study the resource distributing problem of the multi agent cooperation system and the path planning after the completing of the task allocation.For the multi agent system with the task given the problem faced firstly is how to distribute effectively and reasonably the task to other agents. If the distribution for task is unreasonable, it affect the efficiency of the completion for tasks. So, we should follow some capability guide at designing the task allocation algorithm to satisfy the need for the system. Task allocation is to find a most reasonable allocation project in many method. It behaves to find a result satisfied the given restricted conditions and to make the result of object function maximal or minimal.The OD algorithm this paper advanced is in view of the optimization and uses the two-dimensional assignment mode so as to make every robot get the task. optimally. We can see through the test that the operation of the OD algorithm is complex so that the runtime of it is long and especially when the number of robot and task is big it can not get the result.To solve the problem above, we use the Hungary algorithm to realize the task allocation problem. Hungary algorithm is one of the classical algorithms and its application background is to solve the two-dimensional assignment problem. The characteristic of the algorithm is that it can both distribute the task optimally and ensure the runtime consumed by the algorithm is short. We can see that the runtime of the Hungary algorithm is shorter than the OD algorithm and it can solve the problem of the long runtime and distributes the task optimally.After task allocation, we need to make path planning for task distributed. To make the robots complete the tasks in the set time, the path planned ensures both to get to the aim and to be the shortest one. This paper advanced a path planning algorithm—SP algorithm based on the ideal of the depth-first search. The algorithm is used to find a shortest path viewing the length of path as the performance in the environment constructed using topological figure.After the algorithm was designed, we tested the algorithms in the simulational system designed and showed the performance of these algorithms.Though we discussed some problems in the multi agent system, bu...