Research On Urban Traffic Congestion Discrimination And Guidance Model

Nowadays all the big cities in the world have faced the serious problem of trafficcongestion. The traffic congestion causes some explicit costs and implicit costs. Theexplicit costs mainly include sizable economic losses caused by fuel consumption andtravel delay. Some research finds that the economic loss caused by traffic congestionin one country is about 2.5% of its GDP. The traffic congestion causes some implicitcost such as air pollution, energy consumption and health risks as well and it is muchharder to evaluate the loss caused by the rising number of lung cancer patients resultedfrom air pollution. With the rapid development of its economy and the continuousimprovement of people’s lives, the number of motor vehicles is rising greatly than everand almost all the big cities encounter the traffic congestion. Therefore, the trafficcongestion has become topical, urgent and important issue which all the governmentsneed to care. China is promoting the techniques of Internet of Things and Internet ofAutomobiles which provide technical guarantee for the intelligent management oftraffic congestion. In contrast to the previous research of technical background andbasis, this dissertation comprehensively applies such techniques as the electronics,information, communication and artificial intelligence calculation to the informationmanagement of urban traffic congestion, mines the traffic information data andanalyzes the classification of traffic congestion pattern, which will be advantageous tounderstand traffic congestion state timely and correctly and provide the congestiondirecting strategy with traffic congestion information support.First, this dissertation analyzes the complex factors influencing urban trafficcongestion and the causes and characteristics of different types of congestion. Second,on the basis of the massively collected traffic information, it uses the ordereddecision-making theory to construct the decision-making model of traffic congestionand obtains the attribute sets of traffic congestion and the discriminant results of thecongestion degree by feature selection and attribute reduction. Third, in view of thecongestion data of the sections and intersections, this dissertation arrives at thediscriminant sets of traffic congestion and the ordered discrimimant results by takingadvantage of Matlab operation. Moreover, the dissertation puts forward theclassification of traffic congestion patterns by repeatedly training a great number ofcongestion samples. According to the discriminant results of urban traffic congestionand the classification of traffic congestion patterns, it also sorts the congestion degreeof the traffic congestion points in the road net, constructs the information system ofurban traffic congestion and provides the congestion directing strategy decision withinformation support.For the low traffic congestion, this work constructs a timely controlled modelbased on the optimal control theory. On the basis of the store-and-forward controlmodel, this dissertation considers that the queue length increasing of each phase wouldincrease the congestion risk. Under the constraint condition of signal timing, itproposes such a constraint condition that average dissipation rate should replace thequeue length and improves the nonlinear optimal control model with the objectivefunction of the minimum cost. Through the Paramics simulation and comparing themodel performance before and after the improvement, the dissertation regards that thenonlinear optimal control model is considerably effective in directing the low trafficcongestion.For the general traffic congestion, this dissertation discusses the strategic modeland algorithms of directing congestion collaborated between control and guidance.Based on the first and second principles of the theory of dynamic allocation, itexplores the interaction between the traffic control and guidance and considers thegreen ratio and traffic volume as parameters to adopt the integrated coordinationpattern. Based on the Armijo principle, the dissertation also improves the R-PGAalgorithms, shortens the step length rule, increases the searching speed, and decreasesthe congestion direction time through numerical experiments. Through the Paramicssimulation and comparing the model performance before and after improving theR-PGA algorithms, the dissertation regards that the improved R-PGA algorithms isobviously effective in directing the traffic congestion close to saturation.For the heavy traffic congestion, this dissertation studies how to determine thecongestion guidance node by employing the congestion guiding model based on thedistributed real time path mark guidance and look for guidance path by adopting thealgorithms of path generation and path update. Through Paramics simulation, it alsoverifies from such aspects as the depth of knowledge, the allocation parameters of theLogit pattern and the guidance time interval. This study finds that the ATISinformation is of great value to traffic guidance and concludes that the accurate ODdemand matrix, the routing interval and the estimates of traveling time all directlyinfluence the distribution of traffic flow. Based on the simulation test, it identifies theimpact of the model parameters on the guidance effect in congestion region.