Comparing data from mobile and static traffic sensors for travel time assessment

Travel time and speed measures on road networks provide key information to identify critical spots of congestion and evaluate the scale of this phenomenon across an urban area. Many technologies are currently available to measure travel time and speed, but each has its limitations. As part of a wider project aiming to develop travel time reliability indicators, this paper presents a comparison and validation of data collected through four different non-intrusive techniques: floating cars, GPS traces, Bluetooth detections and video processing. First, some background information regarding the project and the modeling of travel times and speed on highway networks is provided. Then, a comparison between the various sources of data is presented. Finally, the characteristics of the various data sources are discussed based on the relevance and the availability of the sources. INTRODUCTION With the increasing problems of traffic congestion in urban areas, transportation planners need better tools to assess its evolution through the years. Travel time measures on road networks provide key information to identify critical spots of congestion and evaluate the scale of this phenomenon across the area. Many technologies are currently available to measure travel time at specific locations: license plate matching, loop detectors, Bluetooth (BT) device matching, and video processing. Traditionally, at a larger scale, travel time is estimated using mandated floating cars (FC) that run on specific routes to gather data on the road network. Nowadays, ad-hoc, or randomly collected, data can also be obtained from GPS devices aboard individual cars and commercial vehicles. In addition, interest for video data is growing since various computer vision techniques permit automated traffic monitoring and data collection. However, each of these data collection methods has limitations that must be addressed through the right aggregation method. As part of a wider project aiming to develop travel time reliability indicators, this paper presents a comparison and validation of data collected through four different non-intrusive techniques: FC sample, GPS traces, BT detections and videobased traffic data. These sources are available for various locations of the Montreal highway network and time spans: the first two mobile sources, GPS and FC, provide traffic data on large spatial areas, but limited in time, while BT devices and video data may be collected continuously, but from static sensors at specific locations. First, some background information on the research project and available sets of data is presented. Then, the analysis framework is presented along with a statistical description of the travel times and spot speed data extracted. The next section presents the possible comparisons between travel condition patterns identified in the various datasets. The conclusion and perspectives are finally provided. BACKGROUND AND INFORMATION SYSTEM Project: assessing the reliability of the highway network In 2008, the Quebec Ministry of Transportation (MTQ) mandated Polytechnique to assess the potentialities of historical FC data to provide relevant estimates of the reliability of the Greater Montreal Area highway network. Six years of FC data were analyzed and used to model the evolution of travel times and variability of travel times over time. Also part of the project was the evaluation of the available technologies to monitor the evolution of travel times and/or speed on the main highway network. Experimentations are currently conducted to compare outputs of various data collection methods. The purpose of this comparison process is to formulate recommendations with respect to the value and relevance of available datasets, tools or technologies to provide critical information for the strategic planning of the network by decision makers. Evaluating congestion on road networks Better understanding the use of transportation network is a key factor for the enhancement of transportation planners and models. Hence, more and more studies are conducted to identify the various negative drawbacks of increasing congestion in urban areas. Issues related to sustainable development are actually providing new incentives to better understand and monitor congestion on transportation networks. For instance, studies on environmental impacts of congestion (Nesamani et al. 2005) as well as on the modification of activity rhythms of households and individuals, namely the reduction in shopping related trips (Schmöcker et al. 2006), are being conducted. Also, numerous research works are linked to the definition of sustainability and the various indicators to assess its level. Litman (2008) provides a very interesting view as well as an extensive list of indicators assisting the objective assessment of sustainability. Indicators such as commuting travel time or delays due to congestion are set as economical indicators in the evaluation framework. Along with the increasing need to better understand congestion problems is the increasing availability of information technology that output multitudes of data on the movements of objects (people, vehicle, phones and other devices). This data can be processed to contribute to the measurement of critical indicators such as travel times or speeds. The emergence of services provided by companies (such as INRIX and Google) relying on distributed data collection and aggregation is an obvious witness of the spreading availability of multiple layers of transportation-related data and of the value of such information. In the United States, the recurrent reports by Schrank and Lomax (2003, 2005, 2007) illustrate the relevance of estimating comparable congestion-related indicators for the main areas of this country. In the same perspective, FHWA (2004) is pursuing a continuous effort to develop and estimate indicators through its Mobility Monitoring Program. This program relies on the data outputted from detectors available in some thirty areas. Another interesting contribution is from the Pennsylvania DOT (Szekeres and Heckman 2005) that implemented a Congestion Management System at the state level.