Research On Traffic State Prediction And Route Guidance In Network Environment

Traffic state prediction and route guidance is a key technology for ITS.The traditional fixed traffic information collection methods have limited density and high maintenance cost.As multi-dimension and nonlinear features,the traffic flow is difficult to predict.Because of complex road network,search algorithms are inefficient and easy to trap in local optimum.Therefore,traffic state prediction and guidance technology is difficult to achieve high efficiency,accuracy and real-time.To solve above problems,on the basis analysis of domestic and foreign research condition,the paper puts forward a technology of traffic state prediction and route guidance in network environment,of which some critical algorithms such as map matching,speed prediction,raffic state classification and visualization and traffic guidance are in-depth studied.The details are as follows:(1)A map matching algorithm based on D-S evidence theory is put forward.Due to the poor quality and not directly applicable of original GPS data,six data cleansing rules are established according to vehicle GPS characteristics to filter error data,to improve the efficiency of data processing;The candidate space is established by candidate points and candidate paths;four weight factors,driving speed,direction,proximity and space correlation degree,are used to the filter candidate space;The map matching based on D-S evidence theory is used to accomplish the first matching,and the optimization method based on trajectory shape matching is applied to implement GPS points precise matching.Through experimental analysis,the accuracy of the proposed method can reach 96.5%,so the efficiency is higher.For a variety of GPS data with different frequencies and error range,its effect is good and has the strong practical value.(2)The road speed prediction algorithm based on GA-SVR regression is presented.Through the analysis of traffic flow,it is classified on the basis of working days and weather forecast model to improve the prediction accuracy;ε-SVR prediction model and gaussian kernel function are chosen to make the SVR model more precision.It is more in line with the multi-dimensional and non-linear characteristics of traffic flow;the parameter selectting of SVR model is realized by using the parameter optimization method based on GA.Through experimental analysis,in the case of sufficient data,the prediction accuracy of the proposed model can reach 89.2%,and can meet the demand of the actual traffic flow parameter prediction.(3)The traffic state classification and visualization algorithms based on improved FCM are proposed.Through GPS data of ICV,the improved FCM algorithm is trained to obtain clustering centers and corresponding membership for six levels of traffic states.The paper adopts HIS color space to transform traffic speed as color dimension to meet the demand of global traffic visualization.Meanwhile,six colors are chosen for local traffic visualization.Through experimental analysis,the classification algorithm can accurately classify the traffic states with regional information,and the visualization algorithm can meet the real-time(5 min)interacting demand of traffic state.(4)Based on the analysis of traffic trip cost,dynamic network construction algorithm and traffic guidance algorithm,the dynamic traffic guidance algorithm using PSO-ACO is proposed.The PSO algorithm is firstly applied for the path optimization,and the results are used to calculate the initial state of the improved ACO algorithm.It improves the execution efficiency and optimization stability of ACO algorithm.Through experimental analysis,the dynamic traffic guidance algorithm using PSO-ACO avoids the PSO to fall into local optimum and the low search efficiency of the ACO,and it has high efficiency,strong convergence.The PSO-ACO is easy to realize global optimal.(5)The traffic condition prediction and guidance technology in network environment is established.The system topology structure is analyzed to determine the scope of the paper.Through the function module and algorithm performance analysis of software system,the system requirement planning is determined including map matching,road speed prediction,state classification and visualization,traffic guidance.And the hardware and software development platform are determined for the traffic prediction and guidance system in network environment.As a whole,the reliability of the proposed algorithm and the system feasibility are verified.The experimental results show that the prediction accuracy of the research results can reach 89.2%,and the real-time performance of the guidance algorithm can meet the actual requirements.And the paper can provides some technical support for the network environment of traffic prediction and guidance system.