Traffic Incident Detection With Floating Car Data In Urban Road Networks

Traffic automatic incident detection (AID) is the process that incident is detected by analysis of traffic flow data which is from both fixed detectors and moved detectors locating at urban road networks. Traffic incidents, such as vehicle collision, vehicle malfunction, road maintenance, and so on, constantly occur in urban road networks. Except that the"point"wastage coursed by incidents, initiated traffic congest and barrage have more serious negative influence on transportation network. Secondary incident may take place if original incident is not clear away promptly. Traffic flow in urban road networks has different characteristics from free flow in freeway and expressway. Many models and arithmetic of AID for freeway and expressway are not fit for urban transportation. With the rapid development of GPS and communication technology, and data communion of governments, corporations and research institutions, floating car data (FCD), which is also called probe vehicle, become the essential composition of traffic information. Probe vehicles cover a large scale of scope of road networks and numerous real time data can reflect road state. The dissertation is composed of quality control of FCD, map matching, comparison of AID algorithms and appraise of effect of detection models. This paper elaborates the influence of incident on characters of traffic on the condition of signal control. CUSUM algorithm is utilized for free flow status while K nearest neighbor methodology is applied for congestion status. AID model of multiple isomerous data is established on the basis of data fusion technology. The experimentation result shows that these AID models are availability.The dissertation mainly includes five segments as follows.(1) On the basis of induce of achievement and direction of AID, all the methodologies via FCD and other traffic data are contrasted and their usability are analyzed.(2) Quality control of FCD is discussed. Many errors are inducted because of position of GPS and other interferes of commutation. On the basis of analysis of errors of original FCD, delete, repair and other disposals are carried on to offer clean data for detection models.(3) Map matching for massive network is the pivotal process. The average frequency of transmission of FCD from thousands of probe vehicles in HANGZHOU city is approximate one minute. Because of real time is the basic desire of AID, it is need to balance efficiency and precise.(4) AID in urban road network is the core theory of the article. Urban traffic flow is express as characteristics of discontinuous flow with signaled intersections. It is the key process of quality assurance that elimination of the effort of periodic signal on traffic flow. Differential flow is proposed to solve the problem. Traffic congest induced by incident has different impact on traffic flow on unequal road environment. CUSUM is adapting to AID of free flow, by contrast, K nearest algorithm is utilized to AID of congestion flow. AID of multiple heterogeneous traffic information based on data fusion is proposed in the dissertation.(5) Real data and artificial data are utilized for evaluate detection models. Quality index involve two aspects. One of them is accuracy which consist of false alarm rate (FAR) and detection rate (DR), the other is timeliness which is denoted by mean time to detect (MTTD).In conclusion, the dissertation is summarized. The creative research achievements and other important research directions were pointed out.