Urban Link Travel Time Estimates Based On The Low-frequency Sampling GPS Data

As a key parameter characterizing the degree of congestion, travel time has more and more attention of traffic managers and researchers. In recent years, the research of link travel time estimation based on GPS floating car data is many. But for the low-frequency sampling GPS data, traditional algorithm’s estimation error is high, and a significant impact of intersection delay time often isn’t considered. So the paper studies a travel time estimation algorithm considering intersection delay for low frequency sampling GPS data.The algorithm contained three main parts. The first was the GIS and GPS data preprocessing and low-frequency map-matching algorithm. GIS data preprocessing was including direction road divided, steering road divide and road intersections influence domain divided, and GPS data preprocessing was including GPS filtering and abnormal stop state judgment. Then the paper used the map matching algorithm based on vehicle direction angle and implemented the algorithm for achieve certification. The second was the low-frequency floating car path estimation considering the impact of signal control. Focusing on the case of adjacent GPS match points across a miniature road network, the paper established the low-frequency probe vehicle path estimation model. AHP evaluation system for link impedance was set up, and the impedance evaluation factors were calculated for each section using the fuzzy comprehensive evaluation method. When the link impedance factor was instead of link length, the most "short" path was the floating car most likely travel route. The third was the low-frequency link travel time estimation considering the intersection delay. The boundary of intersection affect area and Link travel time were divided firstly,. Then, based on the four GPS distribution types in link boundary for the low-frequency GPS points, the paper calculated intersection delays and improved interpolation method to calculate travel time. Finally, an example was given to verify the accuracy of the algorithm.For the low-frequency sampling GPS data, The algorithm had higher estimation accuracy compared with the traditional method, especially in the sections of large intersection delay. And The algorithm is able to provide a wider range ways to apply for floating car technology.