Exploratory Analysis of a Smartphone-Based Travel Survey in Singapore

Future Mobility Sensing (FMS) is an innovative smartphone-based travel survey system that was field tested in 2012/2013 together with the Household Interview Travel Survey (HITS) in Singapore. This paper presents findings of exploratory analysis of the data collected in this test. Clustering of day patterns from FMS data reveals large day-to-day variability of user behavior, which cannot be captured by a snapshot with a one-day survey. Even the high crosssectional variability from the larger sample size of a traditional survey would not have achieved the comprehensive set of heterogeneous patterns as provided by FMS. Some common problems in traditional surveys – such as under-reporting of trips, over-estimation of travel times, and inaccuracy of locations and times – can apparently be reduced by FMS. The FMS data, as compared to HITS, have higher resolution and better accuracy. In addition, FMS is well suited to collect multi-day data as additional costs are marginal and user burden reduces over time. FMS offers a promising technology for next generation travel data collection. Zhao, Pereira, Ball, Kim, Han, Zegras, Ben-Akiva 3 INTRODUCTION The unprecedented and increasing penetration rate of smartphones together with advances in mobile sensing technology have greatly expanded the means of collecting various forms of personal transportation data. Traditional self-reported travel surveys typically suffer from problems such as limited sample size, under-reporting of total completed trips, imprecision in reported trip start and end times (1). Smartphone-based surveys present the opportunity to collect more detailed and precise data needed for emerging agent and activity-based behavioral models. Developments in this field (2, 3) suggest that location-enabled technologies can reduce the number of erroneous “no travel" days and missed trips; improve accuracy of reported trip times, locations and paths; and, reduce respondent burden. The usage of mobile technologies for automatic surveying is not new. GPS-based logging surveys have been widely implemented worldwide and largely successful as a supplement to traditional household travel surveys (4, 5, 6, 7). However, pure GPS logging suffers from some limitations. Financially, the agencies conducting travel surveys must purchase and distribute the GPS collection devices, which can be a significant investment. Also, the participants may forget to carry the GPS loggers with them for the duration of the travel survey, and they may still face a recollection problem when completing their travel diary. In contrast, smartphones provide some clear benefits. For instance, users are accustomed to carrying their phones with them constantly, with adequate battery life, thus decreasing the likelihood of missing trips. Furthermore, smartphones contain a combination of sensors – such as Bluetooth, WiFi, accelerometers, and GPS – which expand upon pure positioning data, providing a richness which can be used to infer activity and mode information. These attributes make smartphones ideal “life-loggers.” The Future Mobility Sensing (FMS) system capitalizes on these “life-logging” capabilities in a next-generation travel survey system, leveraging increasingly pervasive smartphones, advanced sensing and communication technologies and a machine learning architecture (8). With a web-based prompted recall user interface, FMS delivers a previously unobtainable range of data, more closely reflecting what people do, not what they say they do. FMS was field-tested in Singapore in conjunction with the Singapore Land Transport Authority’s (LTA’s) Household Interview Travel Survey (HITS) 2012. The pilot recruited more than 1500 users and produced a large set of rich and detailed travel/activity data, validated by the respondents (9). This paper reports on the exploratory analysis of this unique dataset, demonstrating the capabilities of this kind of survey platform to reveal interesting and diverse user day patterns and overcome some of the known issues of traditional travel surveys. The remainder this paper has five parts. Section 2 gives an overview of the FMS system. Section 3 describes the field test with the HITS survey in Singapore and the data collected. Section 4 presents an example to illustrate the difference between the data collected in HITS and FMS. Section 5 presents the exploratory analysis results and the final section concludes. FMS SYSTEM FMS consists of three separate, but inter-connected, components: the smartphone app that collects the sensing data; the server that includes the database as well as the data processing and learning algorithms; and the web interface that users access to view and validate the processed data and answer additional questions to supplement the validated data. Figure 1 shows three components and the data flows among them. Zhao, Pereira, Ball, Kim, Han, Zegras, Ben-Akiva 4 Smartphone App The smartphone app, available for both Android and iOS platforms, collects data from a multitude of the phones’ sensors, including GPS, GSM, accelerometer, and WiFi. A major design objective of the FMS app is non-intrusiveness, i.e., the app runs in the background of the phone, silently collecting sensor data without user intervention. This aims to minimize the app’s influence on participants during their normal daily activities. In addition, the application is designed to be lightweight (in terms of memory use), easy to use, and energy efficient, using various approaches to minimize battery consumption (10), a major concern for location-based applications. The sensor data collected on the phone are transferred to the back-end server through either the cellular network or WiFi, based on the user’s preference. Backend Server Raw data collected via the app are uploaded to a database where a series of algorithms are used to process the data and make inferences about stops, travel modes and non-travel activities (11). To minimize the user’s interaction burden, the backend algorithms translate raw data into trips and activities. The first round of stop detection is made based on location and point-of-interest (POI) data. GSM, WiFi and accelerometer information are used to merge stops that would otherwise be interpreted as distinct stops. Travel modes are detected based on GPS and accelerometer features, as well as public transit location information. Short duration stops that are unimportant from a data validation standpoint (such as stops in traffic) are deleted for the purposes of presentation in the web interface. Travel destinations (e.g., home, work, shopping, drop-off) are also inferred based on previous validations by the user, POI data, and other contextual information. Web-interface The web interface provides a platform that enables users to review and “validate” their processed data in the form of a daily timeline or activity diary (Figure 2). Validation involves filling in missing information and amending incorrectly inferred data about modes of travel used for particular trips and specific activities engaged in at inferred “stop” locations (destinations). The validated data are uploaded and the algorithms learn from the user validations to subsequently make better inferences. The website is flexibly designed to enable supplementary data collection, such as information pertaining to a specific trip (e.g., how many people the user traveled with or what, if any, fee was paid for parking), during the activity diary validation stage. The LTA pilot involved a helpdesk made available to users through a web-chat or phone call; users were encouraged to have a session with a helpdesk representative for assistance during their first data validation.