Modeling, Analysis and Application of Big Traffic Data for Intelligent Transportation Systems
- Publication Type:
- Thesis
- Issue Date:
- 2019
Open Access
Copyright Clearance Process
- Recently Added
- In Progress
- Open Access
This item is open access.
Intelligent Transportation System (ITS), an integrated system of people, roads, and vehicles by utilizing information and communications technology, has emerged as an efficient way of improving the performance of transportation systems, enhancing travel security, and providing more choice to travelers. Recently, it has been seen that the big data era for ITS is coming due to the wide use of traffic detectors like traffic cameras and GPSs. These traffic detectors can collect various types of traffic data that significantly contribute to the development of ITS, which has the benefit of the public with convenient and safe travel.
With big traffic data, data-driven methods provide powerful and theoretical support for data modeling, analysis, and applications. However, existing methods still suffer from some shortcomings. First, traffic predictors usually use black-box methods to capture the spatiotemporal correlation between traffic. As a result, it reduces the flexibility of predictors due to the time-varying spatial-temporal correlation caused by frequent variation of road conditions. Second, it is impossible to cover all urban areas with traffic detectors. Thus, data absence and data sparsity have an essential impact on the reliability of travel state monitoring in a large road network. Lastly, most big data applications are based on the centralized method for processing and analyzing data, which consume more time and computational resources, optimal decision making. These make research on big traffic data in ITS become both exciting and essential.
In this thesis, a physically intuitive approach is developed for short-term traffic flow prediction that captures the time-varying spatiotemporal correlation between traffic, mainly attributed to the road network topology, travel speed, and trip distribution. Experimental results demonstrate its superior accuracy and lower computational complexity compared with its counterparts. After that, a novel methodology is presented to estimate link travel time distributions (TTDs) using end-to-end (E2E) measurements detected by the limited traffic detectors. The experimental results show that the estimated results are in excellent agreement with the empirical distributions. Lastly, a distributed scheme is proposed for taxi cruising route recommendations based on taxi demands predicted by the proposed Graph Convolutional Network (GCN) based method. Experiment and simulation are both implemented. Experimental results validate the accuracy of the proposed taxi demand predictor. Simulation results indicate that our proposed taxi recommendation scheme is better than its counterparts in the aspects of minimizing the number of vacant taxis and maximizing the global revenue of taxi drivers.
Please use this identifier to cite or link to this item: