Data from mobile technologies 1
Collecting and Processing Data from Mobile Technologies
Peter R. Stopher, Institute of Transport and Logistics Studies, The University of Sydney, Australia
8TH INTERNATIONAL CONFERENCE ON SURVEY METHODS IN TRANSPORT
Annecy, France, May 25-31, 2008
Resource paper for Workshop B4:The Collection and Processing of Survey Data using Mobile Technologies
Abstract
In the past decade or so, mobile technologies that can be used to track the movement of people, freight consignments, and passenger and freight vehicles have evolved rapidly. Initially, this paper reviews the major categories of such technologies, including mobile telephones, Global Positioning System (GPS) devices, Personal Data Assistants (PDAs), and the use of laptop computers and advanced communications systems. In reviewing these, the paper proposes a number of attributes for classification that are considered useful, both from the viewpoint of collecting data by these technological means, and also from the viewpoint of processing. The next section of the paper reviews what has been reported to date about the willingness of people to engage in such technologically-based surveys, and about any reported biases in the make up of the sample obtained, especially in comparison to the biases that are commonly found in more conventional paper and interview-based surveys. From this information, the paper draws lessons about the nature of the samples that can be achieved and the extent to which such samples may be considered either more or less representative of the populations from which they are drawn.
The third main section of the paper addresses the issue of data processing. Of particular concern here is the processing requirements for logged data, where the technological device is capable of recording rather limited information about the travel undertaken, and processing software may need to impute other characteristics of the travel that are normally required for travel analysis. A common thread for both passive and interactive devices, however, is the large quantity of data that can be collected, and the needs for processing to handle such data quantities. A further issue explored is that of the reliability of data that are to be entered by respondents in interactive devices, and special concerns that may arise in processing data collected in real time, in order to be able to make use of the data for prompting or interrogating respondents. The fourth section of the paper discusses the differences between data from mobile devices and data collected in conventional self-report surveys. These differences are discussed in relation to the user of the data and explore some of the potentials that may exist for changes in modelling results from using such data. A further issue that is explored in this section is that of imputed data from mobile technology devices, as compared to imputation in conventional travel surveys.
The paper concludes by drawing conclusions about the usefulness of mobile technologies to collect and process data, outlining the requirements that such data collection may present to the analyst, the limitations of such data collection, and the advantages that it brings compared to conventional methods. The paper speculates as to what the possible future might look like in terms of the extent to which such mobile technologies may be used either to supplement or replace conventional methods of data collection. The paper also draws conclusions on how ready the technology is today and what advances one may expect in the short and medium term from this form of technology. Limitations on the technology are also discussed.
Introduction
In each of the past three conferences in this series, there has been something included in the conference on the topic of mobile technologies. In the 1997 conference, the topic of mobile technologies being applied to travel activity measurement was sufficiently new that a workshop was not devoted to it, but there was, instead, a paper on the topic in a session on “New Technologies” (Murakami, Wagner, and Neumeister, 2000). By the next conference in 2001, a workshop was devoted to the topic of “Using Technology to Improve Transport Survey Quality”, with two papers that were published on the topic, and a workshop summary (Wermuth, Sommer, and Kreitz, 2003; Wolf et al., 2003; and Murakami, Morris, and Arce, 2003). In the following conference in 2004, there were two workshops devoted to the topic, one on web-based and one on non-web-based technologies, the latter being one that was primarily orientated to mobile technologies, again with two published papers and a workshop summary (Wolf, 2006; Kracht, 2006; and Lee-Gosselin and Harvey, 2006). In this conference, there are three workshops that deal with some aspect of mobile technologies. This progression from one paper in a paper session in 1997 to three workshops in 2008 clearly illustrates the rapidity of the evolution and adoption of mobile technologies for measuring travel behaviour. In this conference, there are concerns being raised about non-response in surveys using mobile technologies, the design of surveys using some form of mobile technology, and this workshop which focuses on the collection and processing of the data from such surveys.
Principally, there are two technologies that have been the focus of most of the work on developing mobile tracking technologies to inform on people’s travel: global positioning system (GPS)-based devices and mobile telephone-based devices, generally using GSM (Global System for Mobile communications). The former use the twenty-four satellites that are currently orbiting the earth at a high level and compute position from triangulation to at least three or four in-view satellites, whilst the latter use triangulation and signal strength from base stations to determine position. These two technologies can, themselves, be combined or incorporated into various secondary technological devices. For example, GPS receivers can be used as stand-alone devices, or can be incorporated into a Personal Data Assistant (PDA), a portable computer, or a mobile telephone (Murakami and Wagner, 1999; Doherty et al., 2001; Kochan, 2006) . Likewise, GSM can be used as a stand-alone device. The Personal Handy-Phone System was developed and tested in Japan to provide positioning information to within ±60 metres and has been used as the underlying technology for the P-POINTS system (Personal Positioning Information System), which uses Pico cells to improve accuracy (Shibuya et al., 2000). GSM can also be incorporated into a PDA, or other communications devices, and may be combined with a GPS device (e.g., Ohmori et al., 2006; Hato, 2006; Itsubo and Hato, 2006).
At present, the principle difference between GPS and GSM devices is the accuracy of positioning and the availability of signals for positioning. GPS devices generally still require a clear view of the sky, although increasing sensitivities are permitting these devices to obtain increasingly reliable position information when there is no clear view, and even when some amount of structural material lies between the device and the sky. Nevertheless, they still do not record in tunnels, and experience some problems in urban canyons – locations where the device is surrounded by nearby tall buildings. However, GPS devices will usually work equally as well in rural areas as in urban areas, and are relatively unaffected by the density of development in the vicinity. So far as accuracy is concerned, GPS devices today are usually able to provide position information to within ±5 metres or less. GSM devices, on the other hand, will provide the most accurate positioning in dense urban areas, where base stations are closely spaced, and will provide little or no position in remote areas where base stations either do not exist or are extremely widely spaced. GSM devices will provide position in tunnels and other locations where GPS struggles or is unable to provide a position, including inside buildings. However, even in dense urban areas, position information from GSM devices is usually not much better than ±40 metres, and may be in excess of 100 metres even within suburban areas (Varshavsky et al., 2006). For this reason, when it was mandated in the US that mobile telephones should be able to report the position of the person using a mobile phone to dial emergency services, most telephone manufacturers added a GPS chip to the mobile telephone, so that more accurate positioning would be possible.
Apart from these basic features of GPS and GSM positioning, there are a number of attributes that could usefully be used to classify mobile technologies and which are useful in considering the issue of collection and processing of the data from these devices. In the next section, these attributes and classifications are put forward.
There is another new positioning system that is emerging at present and which could potentially replace GPS and/or GSM positioning in due course. This is Wi-Fi positioning. Recently, there have been releases of software that uses Wi-Fi to locate positions. The device that can be used for this is a PC, laptop, Tablet PC, smart phone, or RFID tag. The more densely an area is populated with Wi-Fi signals, the more accurate is the positioning. In a recent paper (Cheng, Chawathe, LaMarca, and Krumm, 2005), it was found that position could be fixed to within 13-40 metres, depending on the characteristics of the local environment. One of the advantages of Wi-Fi location over GPS is that it works well indoors and does not require line of sight to the sky, and is not subject to the reflection of signals that is a problem for GPS in dense urban areas. However, the technology has not yet received any sort of widespread implementation, so is somewhat a futuristic concept at this time. It is noted here for completeness in discussing mobile positioning technologies, but is not dealt with further, since there are no known applications of it in transport research to date.
Position can also be determined using a Radio Frequency Identification (RFID) tag. These tags can be passive or active. Passive tags have no battery and are provided with energy from a reader or writer that incorporates the RFID tag, such as a cash register or a mobile telephone. The passive tag does not appear to have a potential use for locating travellers. An active tag has its own power source and has a much longer communication range than the passive tag. It is costly, at present, so has not received serious consideration as a locational mobile technology. However, since a tag ID can be extracted from the tag signal, there is a possibility to use it to determine the location of a user (Yamada et al., 2005). However, this technology does not appear at this time to be a serious contender.
For a more comprehensive overview of mobile technologies that may be applicable to transport and travel surveys, the reader is referred to a report by Wolf et al. (2006). This report provides quite a comprehensive review of what was available in 2005-6, although it should be noted that significant advances, at least in GPS, have occurred since the publication of that report.
Classifying Mobile Technologies
As used in current practice, there are a number of attributes that differ in the devices used which will have greater or lesser effects on the collection and processing of data.
In-Vehicle or Portable
The first attribute suggested here is whether the device is designed for in-vehicle use alone, or is a portable device to be carried by a person. In the case of an in-vehicle device, it is not necessary for the device to have its own power source, because it can be connected to the vehicle battery. Further, the bulk and weight of the device is relatively unimportant, and installation can be either a simple plug in to the accessory slot in the vehicle, or can be a more permanent installation in the boot of a car or elsewhere. The in-vehicle GeoLogger®, developed by GeoStats, is an excellent example of an in-vehicle device and has been used extensively in the US and elsewhere. The paper by Bradley, Wolf and Bricka (2005) describes and shows this device. This in-vehicle device represented a major breakthrough in the development of usable GPS devices for a major household survey, following the Lexington proof-of-concept equipment (Wagner, 1997).
In-vehicle devices can also be set up to capture other information from the vehicle, such as speed, acceleration, engine condition, etc. It is also relatively easy to add to an in-vehicle device some form of dead reckoning secondary device, such as a gyroscope or inertial system that will provide position information when the primary technology (GPS or GSM) is compromised. A good example of a GPS in-vehicle device that taps into other information is the one used by Georgia Tech in the Commute Atlanta project (Jun, Ogle, and Guensler, 2007). Laptop computers and other devices may also be included in this category, especially when equipped with either or both GPS and GSM technology to provide positioning information.
Portable devices, on the other hand, are required to be small and lightweight, must incorporate their own power supply, and are generally not able to incorporate other functionality, unless it is very small in size and light in weight and can be placed inside a common housing. For portable devices, the size and weight and the duration of battery power have been key issues in their development. Early devices, such as the bicycle device developed in the Netherlands (Draijer, Kalfs, and Perdok, 2000), were bulky and heavy, with the first one weighing in at a rather substantial 2 kilograms (including batteries). Since that development, however, personal, portable devices have become substantially smaller and lighter in weight, and battery life issues, which dogged most of the early designs – requiring either a battery with sufficient bulk and weight to compromise somewhat the portability of the device, or having a much smaller and lighter battery that did not provide adequate battery life – have largely disappeared as power requirements of the GPS devices have diminished and battery technology has improved.
GSM devices have largely been adaptations of conventional mobile telephones, with size, weight, and battery life attributes that are similar to the mobile telephone, although requirements to transmit position data frequently, or to perform other functionality has been reported to reduce battery life significantly. Other portable devices include PDAs equipped with a GSM or GPS chip that allows positioning data to be obtained. These are usually constrained to the available dimensions of current PDAs on the market, together with the addition of, possibly, a GPS receiver.
Interactive or Passive
Devices can be interactive, in which case the user is expected to enter data that will be recorded along with the position records. This normally requires combination of the GPS or GSM device with a PDA, laptop computer, or other data entry device. The first GPS devices to be tested in the Lexington Proof-of-Concept Project coupled a PDA with a GPS receiver and logger (Wagner, 1997). Respondents were asked to enter such information as the identity of the driver of the vehicle (this was an in-vehicle device), the purpose of the trip and the identity of other persons accompanying the driver. Similarly, work on the CHASE method involved the development of an electronic questionnaire that was loaded into a PDA to supplement position data collected by GPS (Doherty et al., 2001). There are many other examples of interactive devices, but most have in common the idea of linking a PDA or laptop computer to a tracking device.
Alternatively, devices can be passive, meaning that respondents/users are required simply to carry the device with them or place the device in a vehicle, possibly also being required to turn the device on and maybe turn it off. No data entry is required of the respondent. Most passive devices require the user only to turn the device on, and sometimes to turn it off. No data entry is required by the user. Passive devices may also include either firmware that turns the device off and on, depending on detection of movement or GPS signals, or a response device that detects movement, such as a vibration sensor, which turns the device off and on.
While not a feature of the device itself, another aspect of the interactive versus passive classification comes from the manner in which data are collected. Using a passive device, the data may be collected and then processed without further reference to the user, or the user may be asked to respond to a prompted recall survey (Stopher, Bullock, and Horst, 2002; Stopher, Collins, and Bullock, 2004; Lee-Gosselin, Doherty, and Papinski, 2006; Tsui and Shalaby, 2006; Wolf, 2006; Stopher, FitzGerald and Xu, 2007; inter alia). In a prompted recall survey, respondents are provided with maps showing the travel recorded by the GPS and are then asked to respond to a series of questions about the travel, including such information as the mode(s) of travel used, the purpose of each trip, the identity and number of accompanying persons, and possibly costs related to each trip. In the evolution of these methods, increasing use is being made of the Internet as the means to both provide the maps and obtain the input information.