Chaotic Characteristic Analysis Of Highway Traffic Flow And Its Application In Flow Forecasting

Real-time and accurate prediction of highway traffic flow is an importantprerequisite for establishing reasonable regulation of traffic management program toinduce drivers to select the convenient road. Highway is a complex opening system, andthus the traffic flow exhibits the cyclical characteristic and uncertainty, which is theperformance of chaotic system which has the non-linear and self-similar characteristics.Thus, the thesis studies the character of highway traffic flow based on chaos theory,establish chaotic prediction model, improve highway traffic flow prediction accuracy,and then enhance the capacity of highway induction and regulation.Based on chaos theory, the thesis analyzes the chaotic characteristics of traffic flowon highways, and proposes a method to reconstruct aphase space, and then builds ashort-term traffic flow forecasting model based on chaos. Overall, the work of thispaper can be summarized as follows:Correlation analysis and verification of chaotic characteristics of traffic flow onhighway.Firstly,the paper adopts the maximal Lyapunov exponent algorithm andcorrelation dimension method to verify the chaotic characteristics of time series oftraffic flow, velocity and occupancy on highways.And then the paper analyzes thechaotic Correlations of the traffic flow parameters and comes to a conclusion to supportthe study of traffic flow forecasting model by using multiple parameters based on chaosbuilded in the fifth section.Prediction of chaotic time series based on maximal Lyapunov exponent may bringtwo results, and few literatures has studied on it. The paper introduces Markov chain toimprove it. The improved method makes the gradient of time series as state variables,and then builds the state transition matrix on the basis of Markov chain will be used toverify the direction of the forecasting results’ evolution.The traffic flow system of highways is a complex chaotic system. Only using thetime series of traffic flow to reconstruct phase space may not be able to outline thechaotic characteristics of the system completely. In view of this, The thesis fuses thetime series of highway speed and occupancy into a new phase space which contains thechaotic characteristics of speed and occupancy based on Bayes theory firstly. And thenfuses the traffic flow phase space and the new phase space into one through the methodof conditional entropy enlarge dimensionality with traffic flow phase space point used as a basic reconstruction component and the new phase space point used asreconstruction variables.By using the traffic flow data of Yuwu Highway, the paper verifies the forecastmodel with single parameter and multiple parameters individually. The result shows thatthe improved maximal Lyapunov exponent method is valid and feasible. The chaoticcharacteristics of highway system can be better described with the use of the modifiedreconstruction of phase space with multiple parameters. Based on this, the establishedforecasting model with multiple parameters has high precision.