The Multi-Agent System Of The Object Recognition And Location Estimation Information Fusion Process

Along with the rapid development of computer technique, communication, network and microelectric technique, and the complexity of the modern war improved day by day, there are a lot of problems in traditional information fusion system of the object recognition and location estimation, such as complexity about background and environment, diversification of multi-sensor information, requirements for higher information capacity and process speeds, requirements for " study " and " discern " ability in system etc. And the technology of information fusion doesn't form the intact theory frame up to now, especially the aspects of information fusion system function model, abstract level, system structural design and appraisal of performance etc, they all need carrying on the discussion from the systematic side. Just under this kind of background, this dissertation puts forward a new system analysis and design method based on Multi-Agent System technology.This dissertation aims at solving the existing problem of the information fusion system of the object recognition and location estimation and doing some connections and transform between Agent theory and application. The main research works are as follows:The information fusion system of the object recognition and location estimation is researched in the dissertation. The Multi-Agent System of the Object Recognition and Location Estimation Information Fusion Process is proposed, this system is called as MAS-ORLEIFP for short in the introduction after this dissertation. Any complicated function can all be resolved into a series of relatively simple, independent modules in the whole system. When the author designs the system, firstly, a series of atom Agents (i.e. basic Agents) abstractly are produced based on the series of relatively simple, independent modules.The cooperation and communication between multi-Agents are researched in the dissertation. And a new method of cooperation combined with the contract net and blackboard model is introduced. The new method improves the traditional contract net cooperation method, and remedies the deficiency that the traditional contract net cooperation method exists.The abilities of selves' learning and adaptive of Agents are enhanced by reinforcement learning of Agents and increasing the function of feedbacking in the dissertation. And it improves the system's fusion performance, intelligent and interaction with the environment.Finally programming realization of multi-Agent is researched into in the dissertation including the research on Agent developing instrument, procedure frame of the MAS-ORLEIFP system and course of programming realization. The author combines concrete basic Agent (Space-time Agent) and provides Agent programming realization course.