Research On Key Technologies Of Centrally Dynamic Traffic Flow Guidance System Under Large-scale Network

Rapid growth in vehicle ownership is bringing more and more urban traffic congestion problem. The socioeconomic development and people's daily lives are both seriously affected. Central traffic flow guidance system (CTFGS) is one of the core systems of intelligent transportation systems (ITS), which is recognized internationally as the best way of traffic flow guidance to solve traffic problems. Based on the national high-tech R&D program (863 program), this paper mainly researches on the key technology of centrally dynamic traffic flow guidance system under large-scale network.The mainly work is as:1) Achitecture design and key technologies theories analysis of centrally dynamic traffic flow guidance system under large-scale network This paper gives out the logical and physical architecture, reviews and analysizes the key technologies of centrally dynamic traffic flow guidance system under large-scale network.2) Link travel time estimation technology based on floating car Floating car can obtain travel time data. This paper designs a set of map-matching technology based on fuzzy logic method, which can select different criteria depending on the time interval for GPS data collection. A path travel time dividing method based on vehicles'travelling characteristics is proposed, and the average link travel time is estimated under the condition of sufficient and insufficient sample.3) Traffic information fusion technology based on fixed detector and floating car Fix detector and floating car are complementary with each other on traffic information collection. This paper presents a speed estimation method based on the regression model with ACE transform, which is used to obtain the link travel time data with cycle detector data. The fusion model based on TSGA-LSSVM is proposed to fuse the cycle detector data and floating car data. The results of model test based on real data show that the fusion model has higher precision than any single method, and TSGA is faster than GA for parameter optimation.4) Traffic information short-term forecasting technology based on sample classification fittingTraffic guidance based on the forecasted dynamic traffic information can void the lag problem of guidance. This paper presents a traffic information short-term forecasting technology based on sample classification fitting, which classifies the history sample based on KSOM method, and establishes sevel BP neural network forecasting models. The results of model test with the real data show that KSOM-BP neural network method has higher precision than BP neural network based on the whole history sample being trained5) Central guidance path optimization parallel computing technology under large scale networkPath optimization under large scale network always is the Gordian knot. This paper presents a fast path optimization method named MLHND-TQQ through the reseach on multilevel hierarchical network decomposition method and TQQ shortest path computing method. The results of test based on real network data show that this technology can fully meet the real-time demands for central guidance.Finally, work summary in this paper and the suggestion on future work are given. Above the research achievements give theoretical significance as well as guidance value for the research and application on centrally dynamic traffic flow guidance system under large-scale network.