Applications Of Vehicle GPS Data In Urban Transportation

When GPS data collected by probe vehicles become increasingly available,there has been a proliferation of studies leveraging such data to solve transportation problems.This thesis focuses on the analysis,modeling,and application of GPS data to solve problems on the infrastructure and travellers of the transportation system.We first design a map inferencing algorithm based on low-frequency GPS data to generate road networks,which is the infrastructure of the transportation system.Using the generated road network,we study two problems,that is,the link scoring models based on GPS data for the map production and updating process,and the insurance claim models based on GPS data in car insurance.The first research problem is the map inferencing algorithm based on GPS data.Although existing map inferencing algorithms can provide satisfactory results under highfrequency GPS data,for example,GPS data collected every second,their performance deteriorates when the reporting frequency of the GPS data is low,for example,every 30 seconds.We propose a shortest-paths based approach to infer road networks under lowfrequency GPS data.Our approach works with the GPS locations reported by the vehicles directly,rather than the trajectories formed by connecting consecutive GPS locations reported by each individual vehicle.We first perform an initial inference of the underlying road network through an iterative process involving a shortest-paths based algorithm.We then apply a series of map refinement techniques to further improve the appearance of the inferred road network.We compare the performance of the proposed algorithm against two popular trajectory-based algorithms in the literature.Using low-frequency GPS data collected in Changsha,we find that the proposed algorithm outperforms the two algorithms by 19.66% and 17.53%,respectively,in terms of F-score.We then carry out sensitivity analysis using a high-frequency GPS data collected in Chicago.We find that our algorithm is more robust with respect to the increase in reporting intervals.The second research problem is the link scoring models based on GPS data.When evaluating the inferred road network,there are many situations where it is desirable to single out problematic links that do not match the ground truth network.In this research,we develop models to identify problematic links in the inferred road network by estimating their matching probabilities,that is,their likelihood to match links in the ground truth network.We first build a fractional regression model using information from the inferred link and the underlying GPS data,including the length of the inferred link and the density and dispersion of the GPS points in the neighborhood traversed by the inferred link.We then develop a latent class model to capture the spatial and temporal variations in the density of the GPS data.We evaluate the performance of both models using real data collected from Changsha.Results show that the fractional regression model and the latent class model can reduce the mean absolute errors of the predicted matching probabilities by 27.83% and 37.45%,respectively.The third research problem is the insurance claim models based on GPS data.In the usage-based insurance(UBI),GPS data is used for insurance pricing.In this study,we first quantify drivers' familiarity with their driving trips using GPS data from Beijing.We then build a latent class model to study the behavior heterogeneity of the drivers.In the latent class model,the drivers can be classified into two classes,where the first one is the low risk class with 0.54% of the samples accident-involved,and the second one is the high risk class with 20.66% of the samples accident-involved.The significant attributes for estimating the probability of claim for the low risk class are the percentages of urban and night time driving,and drivers' familiarity with their driving trips;while those for the high risk class are the number of hard brakes per hour,the percentage of high-speed driving,and drivers' familiarity with their driving trips.