Research On Freeway Traffic Flow Predication And Incident Warning Method

The improvement of highway incident management significantly affects thehealthy development of the national economy, the safety and comfort of people’stravel and social peace and harmony. First, all stages involved in traffic incidentmanagement are analyzed from the perspective of "crisis management". Theframework structure of a highway traffic incident management system is presentedand the functions of all parts are overviewed. According to the summary of researchstatus about the fundamental function, the "accident beforehand processing" is chosenas the focus of this paper due to its insufficient research. The study is conducted basedon the logic levels of the traffic flow forecast-traffic accident warning-accidentbeforehand process. Main research contents and results are as follows:1. The research on short-term traffic flow forecasting method and the forecastsystem. First, a state vector selecting algorithm based on genetic algorithm andcohesive degree index is proposed. Experiments of the algorithm under four differentscenarios are conducted. It is demonstrated that the proposed algorithm cansignificantly improve the accuracy of nonparametric regression forecast and neuralnetwork forecast. Then a data driven short-term traffic flow forecast system which canbe used for large transportation network and multiple points real-time forecast is putforward. Actual data experiment and simulation results show that the proposedforecasting system can cope with the situations under the conditions of large networkand high data flow, and can obtain better forecast performance.2. The research on the space-time distribution of highway incident analysis andaccident warning algorithm. The space-time distribution of highway traffic accident isanalyzed on the basis of accident data of Jin Ji highway, and find that highway trafficaccidents have obvious relationship with the time of day, date, location, model scaleand traffic flow which proves the highway accident regularity of space and time. Ahighway traffic accident beforehand warning algorithm based on fuzzy reasoning isput forward. The abnormal state of traffic flow is identified by analyzing the real-timecollection of traffic flow parameters. Experiments results show that the proposedmethod can accurately predict the risk of traffic accident and can be used as referencefor decision-making of managers. 3. The research on highway incident beforehand response. A schedulingalgorithm for optimizing highway patrol vehicle is proposed. Based on the vehiclerouting problem, the accident risk is employed in the objective function. Then atwo-stage heuristic algorithm is designed to solve the proposed problem. Experimentsunder different highway network conditions are conducted. The results show that thealgorithm can improve the accident coverage of patrol vehicles and reduce theresponse time.In summary, our paper focus on "accident beforehand processing" in the highwayincident management. Several methods are put forward including the traffic flowforecast, accident warning and the handle method after warning. The approaches arevalidated by field and simulation data. Results indicate that our approaches are helpfulfor the focused problems...