Research On Forecasting Vehicle's Velocity Based On Data From GPS Floating Cars

Running speed is not only one of the important parameters of traffic state, but also an indicator needed to forecast in Intelligent Traffic System. We can get road information such as traffic flow and density from speed, which could be used to forecast future situation and so on. Previously, prediction based on data collected by fixation equipments is not almost real-time. However, floating cars scatter homogeneously in the section, send information (such as time, speed, latitude and longitude coordinates, direction) to the center timely through board GPS devices and wireless communication equipments. Therefore, data collected by floating cars has established a basis for the running speed forecasting studies.This paper has research on short-time traffic speed prediction models after reviewing floating car technology and short-time prediction home and abroad. Based on data from floating car, combining the history of network operation and real-time traffic data, we have built K-neighbor forecasting model, gray-Markov forecasting model and weighted-Markov Model. We proposed a method of finding neighbor based on cluster analysis and the Pareto effective solution in K-neighbor model, combine Markov theory with grey theory in gray forecasting model and add traffic station division into weighted Markov forecasting model. Finally we give a demonstration of link vehicle's velocity prediction to verify models' validity based on the floating car date gathered from Hangzhou city.