Methods for real time prediction and visualization of traffic using smartphones

Abstract

The advances in the capabilities of smartphones and their widespread popularity have opened up new avenues for improvement in Advanced Traveller Infonnation Systems (ATIS). Increased computation power, storage capacity and better internet connectivity have made smartphones the optimal choice in providing more useful real time information to the travellers and allowed the implementation of intelligent algorithms on smartphones. The compressed traffic prediction method is one such algorithm that provides accurate real time predictions of traffic speeds by explicitly predicting speeds only at a small number of links in the network . The objective of this thesis is to illustrate and evaluate the different approaches to disseminate and visualize compressed traffic prediction data using an android application. An android application that allows effective visualization of traffic data overlaid on a map and perform other geospatial tasks is created. The application provides an illustration of the entire road network of Singapore, where road segments are colored according to average speed of the particular segment, which can be overlaid with rainfall patterns or road incidents, for visual analysis. For the back end, different methods are introduced and evaluated. First, a server based approach is tried out where the traffic information is stored and predictions are done on the server. This is followed by a hybrid method where the computation of traffic predictions alone is done on the server while the device generates the spatial features locally. Next, both computation of traffic predictions as well as visualization of traffic conditions is performed on the smartphone. In this case, the server behaves only as a data collector from where the smartphone fetches current traffic data. The performance of smartphones in each of the above methods is studied and the advantages and disadvantages of each of the proposed method is highlighted. Further, the feasibility of using predicted traffic data to address real world problems such as optimal route planning and travel time calculation is explored.