Design And Implementation Of Object Tracking Cooperation Method In Multi-Mobile Agent System

Cooperative object detection and tracking of robots team is a result of combining robot team cooperation and active object detection and tracking. A composite multi-mobile agent system is designed to meet the requires of real time cooperate tracking. It include the blackboard model in the server and the deterministic finite automaton (DFA) based composite agent model in the clients.Multiple agents cooperate with each other through the blackboard model. The construction and design of the blackboard model make the cooperate tracking come true. It provide the platform for sharing information, driving the different knowledge resources, predicting the object trajectories and guiding the agent groups movements. Another task distribution strategy combining Hungarian method with CNP (Contractor Network Protocol) is designed to make sure the real time tracking in effect.A DFA based composite agent model was proposed and it successfully applies the action status model with DFA, meanwhile making use of the state abstraction of DFA. So the model not only improves the performance of an Agent itself in dynamic cooperative object detection and tracking, but also improves the efficiency of group cooperation, in the perspective of sociology. This model is totally independent of action control unit, decision making and environment information, has good portability and extendibility.Compare the path data generated by the Agent instructed by the original algorithm and the one applied DFA, it can be found that the expected value of error and sample variant in the latter one is only half of the previous one. And it is used in the occlusion based object tracking experiment and region based multi-object tracking experiment successfully. The model is designed and improved for cooperative object detection and tracking, and can provide an effective approach to it.