Research On Bus Arrive Time Prediction Based On SVM/H_∞

As the development of social economy, the acceleration urbanization process, continued growth in vehicle ownership, the transport infrastructure and road management exiting many problems, the traffic congestion is more and more serious. As an important part of intelligent transportation system(ITS), the development of advanced public transportation system(APTS) is the most efficient approach to resolve the traffic problems. Real-time and accuracy bus arrival time prediction can not only provide the passengers with different travel choice but also give some guidance to the managers. In the meantime, the complexity of traffic flow and unpredictable traffic conditions make it hard to predict the bus arrival time. So it is meaningful to do research in real-time and accuracy bus arrival time prediction both theoretical and practical.Based on the discussion of bus arrival time prediction model and related technology,aimed at improving the prediction accuracy and its practicality, on the use of public transportation vehicles GPS data, take full account of the arrival time and the impact of buses arriving prediction accuracy factors, combined with support vector machine (SVM), Kalman and H∞filtering ideas, the SVM-based public transport vehicles arrival time prediction algorithm is designed, for small samples of data SVM has strong forecast ability of statistical learning. But SVM is based on historical data, it can not reflect the real-time traffic information dynamically,an algorithm based on SVM/Kalman is proposed. Considering the buses running process affected by the surrounding environment (high-rise buildings, plants, viaduct) likely to cause GPS signal reflection and attenuation and complex traffic conditions (stop cars, congestion) to the observed data may bring unknown noise to the actual data, and Kalman filtering algorithm has the assumption that the statistical properties of the noise characteristics must be Gaussian. Then an algorithm based on SVM/H∞is proposed.In order to improve forecasting accuracy, based on the analysis of GPS data and its statistical characteristics, the raw data is pre-processed. Finally, in use of the data from the bus line one in Jinan City, several experiments are designed and the experimental results show that the algorithm based on SVM/H∞performs better than the methods based on SVM and SVM/Kalman, and it also reveals its higher robustness. And it can give some guidance to the development of advanced intelligent public transport to improve the efficiency of public transportation has some significance.