Travel Pattern Identification Utilizing Smartphones

1. Introduction

The used use of Smartphones has brought in an interest by survey designer in the potential of using them for collection of travel data. In Australia for example, in 2013, 73% of the population between the ages of 15 and 65 owned Smartphones and this is projected to rise to 93% by 2018 (Frost & Sullivan, 2013). According to Fan, Chen, Liao and Douma (2013), the appeal for use of Smartphones for collection of travel data is cost-related, pragmatic, and technical. Through Smartphones, a substantial cost saving is expected cost all that is required is an app that the participant will simply download and install on their own Smartphone as compared to having a device given to them in the case of Global Positioning System (GPS). Pragmatically, it would be expected that an individual would keep his/her Smartphone charged as compared to a stand-alone GPS device and technically, Smartphones provide a variety of location tracking capabilities and sensors among them Wi-Fi, GPS, and accelerometers which can be used to infer movement. Despite the above benefits, there are some barriers which include, there are still some members of the society who don’t own Smartphone’s especially the elderly, there is biasness in downloading and using apps, and the issue of battery life. This paper therefore seeks to establish how Smartphone’s can be used for the purpose of determining travel patterns of persons.

• Background

According to Parlak, Jariyasunant and Sengupta (2012), Smartphones have become increasingly useful in travel survey as to replace GPS devices which were the first geo-positioning technology. The potential of using mobile communication devices in identification of travel behavior has been discussed my numerous authors. The increased and accelerated adoption of Smartphones in recent years has in turn increased the adoption of this technology in travel and position data collection. Since the start of this new technology, various authors have contributed in the knowledge pool of using Smartphones for the collection of travel behavior. This use of applications for collection of location and travel data can be divided into two main groups based on use cases; user initiated and background event detection strategies.

Applications that are referred to as user initiated user detection are those that an interaction between the user and the application is required for detection and logging of significant events occurring during the targeted time period (Clifton and Muhs, 2012). On the other hand, background event detection applications are those that using services and sensors in the background for collection of constant stream of data. The collected and streamed data is then handled through a verity of methods and single processing strategies depending on the objective of the data collection event. Sensors can be used for detection of additional data attributes for example mode or for provision of finer definition of the activity location.

The app of the UbiActive (Chen et al., 2012) and Quantified Traveler (Jariyasunant, Sengupta and Walker, 2012) are examples of projects that are taking advantage of the sensor element in Smartphones to automate detection of travel detail. The UbiActive uses three Smartphone sensor – GPS, accelerometer, and magnetic sensors – and records travel data as to derive the distance travelled, the mode of travel, and the time taken in the travel activity as well as physical activity the individual conducted during the travel event. The Quantified Traveler app uses accelerometer and GPS to determine location and the mode of travelling used in the event. The streamed data is then analyzed and translated into a travel footprint which is a quantification of monetary, environmental, and health measures.

2. Available Smartphone sensors

There are a variety of Smartphone sensors that can be used in determining travel and location activity. The specific sensors for any given device do vary, but the average device, especially the modern models have GPA, gyroscope, accelerometer, light sensor, and compass as well as Wi-Fi, Bluetooth, microphone, and cellular radios (Rissel, Greaves, Wen, Capon, Crane and Standen, 2013). Nevertheless, some of the high end Smartphone devices have additional sensors which include humidity sensors, thermometers, magnetometers, and barometers. Some of these sensors can be used in the detection of movement as a distinct activity from travel.

• Location determination

GPS is the most popular location sensor, but in Smartphones, location can be effectively and sufficiently captured through network-based triangulation through Wi-Fi and cellular networks. This technology is battery-friendly, an advantage over GPS which leads to fast battery depletion. Moreover, the network-based location capturing technology has been shown to provide better accurate data for dense urban regions and for frequently visited locations (Greaves, Ellison, Ellison, Standen, Rance and Rissel, 2014). Network-work based location data capturing Smartphone capabilities are available from providers such as Microsoft and android systems. These two provides according to Jariyasunant et al., (2012) have a large database containing numerous geocoded Wi-Fi networks and cellular towers which is used for the purpose of determining Smartphone location. This kind of service – network-based location – requires that the Smartphone be connected to the internet for location capturing.

The fact that internet connection has to be active for the location capturing means incase the individual forget to turn on internet connection, then location can’t be determined. To cut on the dependency of the internet, modern location logging apps work in a similar strategy by maintaining a database of landmarks for example Wi-Fi and cell id for frequently visited locations e.g. work place (Roorda, Shalaby and Saneinejad, 2011). When an individual is at work, all available location identifiers for example Wi-Fi network and Bluetooth devices are stored along with their GPS coordinates in the database and by simply detecting these landmarks, the location is recognized and the coordinates thereof even if there is no GPS single or active internet.

Through observation of these known landmarks, applications are able to determine the arrival as well as departure time from the area (Oshin, Poslad and Ma, 2012). Arrival time in the location is marked as the time when these landmarks are detected and departure time as the time when detection of the area landmarks ceases. Likewise, in some public areas for example in cafes which have their Wi-Fi networks bearing the name of the business, it is possible to use location identification Smartphone capabilities to predict the activity going on in the location, or simply what the individual is doing (Parlak et al., 2012).in outdoor environment, network-based location service is normally less accurate as compared to GPS and it doesn’t provide speed at which the individual is travelling at.

• Data collection modes

Studies have shown that the use of new technologies is mainly to replace to supplement the traditional travel survey methods for example travel diary (Clifton and Muhs, 2012; Roorda et al., 2011; Greaves et al., 2014). In addition to using Smartphones for the collection of travel behavior data at the micro level, they present a potential to be used in monitoring traffic, observation of overall transportations movement patterns and dynamic within a city or a roadway network. Some of the applications being used in this front include Mobile Millenium being implemented in San Francisco and bay area and WikiCity being used in Copenhagen, Rome, and Amsterdam (Paek, Kim and Govindan, 2010). The benefits of using GPS and Smartphone application in transportation are tangible to travel behavior modelers, system operators, decision makers, and transportation planners.

3. Location sensors

The way which geographical coordinates for longitude and latitude are captured by Smartphones is determined by the operating system on which the device is running on (Saeedi, Moussa and El-Sheimy, 2014). Location capturing methods using GPS can be through GPS or AGPS (Assisted GPS). In an android device, the developer can choose in determining the specific method to be used in obtaining location fix. This is done by specification for the location manager to access course-grained (cell tower and Wi-Fi access point transliteration), fine-grained location (GPS triangulation), or best location provider (a combination of all the three sensors depending on accuracy, fix retrieval time, and battery requirements) (Shen and Stopher, 2014). However, for iOS, the choice of the location provider to use is not available as the system automatically chooses the best location provider, and the best method for position triangulation is determined by the circumstances under which the location is being examined.

• Accelerometer

Recently, Smartphones have improved accelerometers largely as a result of the increased gaming and entertainment function of these devices. Early devices in particular android and Apple phones have an accelerometer but it only measured is a single dimension (Oshin et al., 2012). Current Smartphones have accelerometers measuring in 3 axis therefore capable of being measured in the x, y, and z dimensions. The 3 dimension accelerometers can be used for the detection of changes in the device status which are as a result of travel.

According to Clifton and Muhs, (2012) the primary use of accelerometers include determination of mode as well as general identification of movement which is an indication of participation in an event in a new local. Previously, the use of accelerometers to capture travel behavior data depended on GPS data to determine mode. For the case of GPS fixes, they have to be obtained through high frequency so as to differentiate between modes which is rather challenging due to the high demand for power and the relatively short Smartphone battery life. Research by Fan et al. (2013) on how to process accelerometer data integrated the three dimensions of accelerometers and used the total to determine the count measures which resulted to an indicator of physical activity. In classifying accelerometer data, both higher level and lower level statistics have been implemented as in Saeedi et al. (2015).

The feature extraction process as well as several of the lower level statistics and the Fourier transform coefficient for classification is also used in processing accelerometer data (Roorda et al., 2011). The other method used is the testing of several classification schemes and the best performing method which makes use of decision trees which are followed by a Markov model. Other research works have been used for comparing different location capturing methods – Wi-Fi, cell tower, and GPS – in addition with accelerometer sensors with the objective of showing the need to fuse data so as to detect mode. As stated by Oshin et al. (2012), it is essential for data captured through an accelerometer to be fused with other sensors so as to derive more accurate modes and activities. For example, in the detection of mode, omission of GPS leads to a 10.4% accuracy decrease for classification of mode.

• Wi-Fi signal strength

Wi-Fi according to Clifton and Muhs (2012) has been discussed as being a location capturing method so long as there is access point(s) with known physical location to be used for triangulation. However, there is less study on the use of Wi-Fi sensors for monitoring the behavior in stationery which consumers less power. Signal strength in Wi-Fi access points provides the potential of being used to indicate the speed of movement. If Wi-Fi signal and strength are moving in and out over short periods of time, then it can indicate high speed of movement. The advantage of Wi-Fi access points and Wi-Fi signal strength is that, they don’t have to be open access points and the device doesn’t have to be connected to the Wi-Fi network. Also, the physical location is not a primary requirement, even though it is useful to have it, for the detection of movement.

All Wi-Fi access points broadcast a Mac address and in most cases, they have a name associated used to refer to them, and each Smartphone has the capability to measure a number of components among them, whether the device is connected, Mac address, the SSID, and the signal strength for each of the nearby and available access points (Rissel et al., 2013). Wi-Fi data use in positioning is limited to simple classification of movement behavior as compared to the advanced and more complex classification of mode through accession of accelerometer data. Even through according to Parlak et al (2012) the number of studies in this topic are limited, using Wi-Fi presents an opportunity for reduction of battery drain experience when running the app.

• Data quality

Paek et al. (2010) argues that the simultaneous use of all the sensors available in a Smartphone presents a potential for collecting data that is detailed and of higher and better quality that when using a single sensor. However, as Shen and Stopher (2014) point out, there are a number of barriers hindering this realization. The first barrier is the fact that combining data captured by the various sensors is not straightforward because different sensors provide data that is dis-similar and seemingly contradictory for example, while accelerometers show movement, GPS no change in location (Saeedi et al., 2014). The second barrier is that using many sensors for a continued time requires more power which translates to reduced battery life for Smartphone devices to a point where they are rendered unusable even for the day to day communication functions of a phone.

While it is a fact that using several sensors at a go has the potential for improved data, the challenge of power presents a limitation, but to get the best and accurate data while still saving on the battery, Jariyasunant et al. (2012) suggests selection of the best combination of sensors which wouldn’t be too heavy on the battery. To harness data capturing and battery power, some sensors have been placed in Smartphone devices running on android and iOS. According to Shen and Stopher (2014), the most popular and widely available Smartphone location sensors with capability to determine location are Wi-Fi and cell network. To capture location through these sensors involves collection of the unique identifiers that are broadcasted by Wi-Fi access points and cell towers. By use of know Wi-Fi access points and cell towers that have been collected and stored in a database for android and iOS, position is determined through triangulation between the known Wi-Fi and cell towers close to the Smartphone. This estimation is with a measure of accuracy being in meters.

4. Conclusion

It is obvious that Smartphones present an enormous opportunity to improve the way in which location data is captured and collected within persons. It is however true that, there still are substantial barriers. Smartphone have sensors that can be used in determining travel and location activity. The specific sensors for any given device do vary, but the average device especially the modern models have GPA, gyroscope, accelerometer, light sensor, and compass as well as Wi-Fi, Bluetooth, microphone, and cellular radios. Among the challenges that still persist is the issue of power and battery life. This barriers stand between use of the available sensors in synchrony for better and more accurate location data. Geographical coordinates for longitude and latitude are captured by Smartphones is determined by the operating system. GPS is the most popular location sensor but in Smartphones, location can be effectively and sufficiently captured through network-based triangulation through Wi-Fi and cellular networks which are battery-friendly and an advantage over GPS which leads to fast battery depletion. The type of data collected has the potential for deterring several attributes of movement for example mode and activity and Wi-Fi network presents these both capabilities. Given Smartphones cannot operate all the location sensors simultaneously, only some have to be used in android and iOS which are the compatible operating systems.

5. References

• Clifton, K. and Muhs, C. D. (2012). Capturing and Representing Multimodal Trips in Travel Surveys. Transportation Research Record: Journal of the Transportation Research Board, 2285, 74–83.
• Fan, Y., Chen, Q., Liao, C.-F. and Douma, F. (2013). UbiActive : A Smartphone-Based Tool for Trip Detection and Travel-Related Physical Activity Assessment, in Annual Meeting of the Transportation Research Board. Washington, D.C.
• Frost & Sullivan, (2013, august 4). Australian Mobile Device Usage Trends 2013. Retrieved from http://www.frost.com/prod/servlet/press-release.pag?docid=282630973
• Greaves, S. P., Ellison, A. B., Ellison, R. B., Standen, C., Rance, D., Rissel, C., et al. (2014). A Web-Based Diary and Companion Smartphone app for Travel/Activity Surveys, in International Conference on Transport Survey Methods. Leura, Australia.
• Jariyasunant, J., Sengupta, R. and Walker, J. (2012). Overcoming battery life problems of Smartphones when creating automated travel diaries, in International Association of Travel Behavior Research. Toronto, Canada.
• Oshin, T. O., Poslad, S. and Ma, A. (2012). Improving the Energy-Efficiency of GPS Based Location Sensing Smartphone Applications, in 2012 IEEE 11th International Conference on Trust, Security and Privacy in Computing and Communications, 1698– 1705. Ieee, Liverpool, United Kingdom.
• Paek, J., Kim, J. and Govindan, R. (2010). Energy-efficient rate-adaptive GPS-based positioning for Smartphones, in 8th International Conference on Mobile Systems Applications and Services. San Francisco, United States.
• Parlak, S., J. Jariyasunant, and Sengupta, R. (2012). Using Smartphones to Perform Transportation Mode Determination at the Trip Level, in The 91st Transportation Research Board Annual Meeting, Washington, D.C.
• Rissel, C., Greaves, S., Wen, L. M., Capon, A., Crane, M. and Standen, C. (2013). Evaluating the transport, health and economic impacts of new urban cycling infrastructure in Sydney, Australia – protocol paper. BMC public health, 13, 963–970.
• Roorda, M. J., A. Shalaby, and Saneinejad, S. (2011). Comprehensive Transportation Data Collection: Case Study in the Greater Golden Horseshoe, Canada. Journal of Urban Planning and Development, Vol. vol. 137, no. no. 2, pp. 193-203.
• Saeedi, S., Moussa, A. and El-Sheimy, N. (2014). Context-aware personal navigation using embedded sensor fusion in Smartphones. Sensors, 14, 5742–67.
• Shen, L. and Stopher, P. R. (2014). Using SenseCam to pursue “ground truth” for global positioning system travel surveys. Transportation Research Part C: Emerging Technologies, 42, 76–81.