Real time vehicular traffic estimation using cellular infrastructure

Abstract

Availability of city wide accurate traffic information enables optimal flow of vehicles in a road network. Intelligent Transportation Systems (ITS) play major role in generating fine grained traffic information. However, in developing countries like India, limited ITS infrastructure is available and city wide manual traffic surveys are the basic source of traffic information. Manual traffic surveys are carried out by government agencies once every year or even less frequently and this instantaneous traffic information is extrapolated to presume traffic condition in a region for the whole year. The generated traffic information has limited application and government generally use it for planning transportation infrastructure development. Cellular infrastructure is widely deployed in India. As per TRAI Press Release No. 38/2013, there are more than 867 million cellular connections in India and cellular density is reported to be more than 70% [1]. Aim of our work is to study feasibility of using cellular infrastructure to generate useful traffic information. Our preliminary experiment with a vehicle carrying GSM modem shows that it is possible to track regions through which vehicle traverses just by using raw data about cell ID updates. The experiment also establishes need for a sophisticated map matching algorithm for determining exact route of a vehicle. We develop a map matching algorithm which can work with large location errors and show using simulations that it is possible to generate useful traffic information such as origin-destination of a trip, route and duration of a trip with in reasonable error bounds even with location error of 250-500 meters. However, for generating accurate travel time estimates for individual road segments, lower location error bounds are needed.