中文摘要：

针对当前交通事件发生过程及状态自动识别的不足,围绕车联网道路交通信息所体现出的新特性,提出基于多Agent的车联网信息融合方法（VIFMA）。通过在多Agent间引入决策关联矩阵进行信息交互,将车联网道路交通信息融合建模为Agent决策问题,从而实现对交通事件的自动检测。仿真试验结果表明：VIFMA能较好地区分出交通事件发生过程中自由流、拥堵加剧和拥堵消散3类状态,揭示拥堵加剧状态与拥堵消散状态之间存在一定的＂粘黏＂;对比试验显示VIFMA具有更良好的容错性能和平稳特性。

英文摘要：

In order to make up for present shortcomings of automatic identification of traffic accidents process and status this paper, focused on new features reflected by road traffic information of vehicle network, and proposed a multi-agent based vehicle network information fusion algo- rithm （VIFMA） which casts information fusion problem into an agent decision problem by intro- ducing decision-making matrix into multi-agent systems to realize the automatic detection of traf- fic accidents. VIFMA makes use of multi-agent clustering method to classify sorting vector of in- dividual agent. The experimental results show that, VIFMA is able to distinguish three catego- ries, i.e. free flow, aggravation and dissipation of congestion, when some traffic incidents hap- pen. It also reveals there is certain ＂sticky＂ between the aggravation and dissipation of conges- tion. Comparing with the connection method, the experiments show VIFMA owns better fault- tolerant performance and stability characteristics. 5 tabs, 5 figs, 29 refs.