Research And Implementation Of Agent In The Accident Escape

As the social and economic activities have become more frequent, and people lack of knowledge of disaster, resulting in huge casualties and property loss, therefore, virtual reality technology has been pervasively applied to disaster prevention for reducing the loss. This thesis mainly focuses on the agent technology based disaster escape, which is the one of core contents of virtual reality.Effect of simulated escape of agent primarily depends on some factors, such as mechanism of path navigation, performance of agent, approach of behavioral modeling, performance of manipulation algorithm and so on. Through analysis of the traditional path navigation algorithm, it is acknowledged that the classic Dijsktra algorithm lacking of heuristic information leads to inefficient search and the classic A*algorithm does not work without the location of target, so this thesis adopts stratified path navigation mechanism that uses different path search algorithm for the different particle size to improve the performance of path navigation mechanism. At the same time the system uses priority queue based on the index of node to reduce the time consumption of the searching algorithm. The effectiveness of these improvements and design has been confirmed through the test simulation.Suitability of exports for agent in the simulation environment depends on some factors, such as distance from agent to export, the number of agents round the export and so on, because of the concepts of factors are so ambiguous that the most appropriate export can not be straightforwardly found, therefore, this thesis improves the performance of agents through reasoning out the most appropriate export with fuzzy logic. In addition, this thesis models the escape process of agent through behavior trees, the agent’s complex behaviors of escape process can be clearly decomposed so that the difficulties of realizing agents are reduced.