A Study On Fuzzy-Logic-Based Freeway Incident Detection Algorithm

Freeway construction and development have made significant social and economic benefits. However, as the growth of freeway mileage and more complex network structure, the traffic incidents have increased significantly, resulting in many problems, such as traffic safety, transport efficiency, environment pollution, and energy consumption. Based on analysis of the real-time traffic parameters, automatic incident detection can pass information to traffic managers. The aim of these is to reduce traffic congestion and the losses of traffic accidents. Automatic incident detection is one portion of the advanced traffic management systems, and it is especially important for the application of intelligent transportation systems.The categories of freeway incident and parameters of traffic flow have great impacts on traffic detection algorithm. In the dissertation, two types of traffic incidents are classified. The effects of recurring incident and non-recurring incident on traffic flow are studied. Besides that, the characters of traffic parameters are analyzed and a new method to select the detection algorithm parameters is proposed.Based on analysis of existent detection algorithms, a fuzzy-logic-based incident detection algorithm is proposed. The algorithm can not only deduce whether the incident happened, but also the state of the traffic flow. It is a multi-inputs and multi-outputs algorithm. The new algorithm has full consideration for complexity of the traffic flow, using membership to describe the specific traffic flow patterns. Speed and volume are the inputs of the algorithm. Fuzzy inference engine is used to judge whether the incident happened or not. Besides that, the flow of fuzzy-logic-based algorithm is constructed. Compared with increment algorithm, the new algorithm has higher detection rate and lower false alarm rate.By means of micro-simulation using traffic software integration systems, data of traffic flow was acquired. Micro-simulation process includes two parts, normal state of traffic flow and abnormal state of traffic flow. Through the first part, threshold of the algorithm is acquired. By means of the second part, rules of fuzzy logic are determined. Finally in the dissertation, performance of the fuzzy-logic-based incident detection algorithm is assessed by detection rate, false alarm rate and mean time to detect.