California Department of Transportation
 

Mobile Survey

Mobile Survey is a comprehensive system, designed for conducting automated travel surveys. It takes advantage of capabilities and growing use of smart phones by the general public, and will provide a low-cost, efficient and highly scalable method of conducting travel and transportation surveys by utilizing smartphone apps and automation algorithms. The main apps will cover all major smartphone platforms and will be used for a continuous travel survey. The continuous survey collects core data elements for a typical travel survey. Most detailed activity information can be detected by the system through automation algorithms and geocoding. Main applications may be used for household travel surveys and most other travel surveys regardless of size and complexity. The main apps may need to be modified for some specialized surveys. The system is being considered for development at Caltrans.

Surveys may target specific geographic areas or demographic groups, and conducted at state, national or global level. Targeting may be accomplished by sampling or through postprocessing. Volunteers will use their own smartphones to download the apps and participate in the survey. Recruitment of volunteers will be mainly through promotion, advertising and incentives, but may also include traditional recruitment methods such as mail and phone calls.

HOW DOES IT WORK?

This is how the continuous survey which uses the main applications will work. Specialized surveys work the same way, except for recruitment which may be different depending on the characteristics of the survey.

Recruited survey participants are directed to a website to register, read instructions, fill out a demographic questionnaire and download the app. They may also have the option to read the instructions and fill out the demographic questionnaire directly on their smartphones. After that, they can participate in the survey as often as they want and for as long as they wish. To start the survey, the participants identify their first location (origin) on the first screen of the app, start travelling and answer questions on the app as they appear. Questions appear when triggered by the server-side algorithms after a stop is detected. They should be answered only when it is safe for the participants. We encourage the participants not to use the app while driving. Answers don’t need to be in real time. Following attachments show the demographic questionnaire and the activity flowchart that will be programmed in the app.

Demographic Questionnaire Mobile Survey Flowchart

Activity data and all information collected by GPS are transmitted to a central server in real time. The data is integrated with demographic information, and compiled for postprocessing and data mining.

The collected data will be presented by utilizing a data analysis tool through an on-line interface that works with the collected data in the background. The data analysis tool works with the streaming data and will be user friendly and comprehensive to the point that it would eliminate the need for sharing the raw data with anyone.

AUTOMATION FEATURES

A memory function will enable the app to learn the participants’ travel patterns. When a participant enters a specific trip activity or purpose for a location, the location and activity that the participant has entered will be stored in the smartphone’s memory (It could also be stored on the server). When it is confirmed, the app will associate that location with the activity that has been entered previously. After that, every time a stop occurs at the same location, the activity will be entered automatically without asking the participant for input. This will greatly reduce burden on the participants, especially for locations like home, work, school and daycare that are frequent origins and destinations.

Another central feature of this system is the stop-detection algorithm.  It will detect start and end of trips by considering speed changes and other factors. The algorithm will run in real time and send a signal within five or ten minutes to the smartphone indicating that a stop has occurred. The algorithm will eliminate the need to ask participants to indicate stops and beginning/end of the trips. This is essential because our field tests show that people frequently forget to enter this information and when it is entered, the time of entry is usually very different from the actual time for stops and beginning/end of the trips. This makes the data entered by the participants unreliable. Other automation algorithms needed in this system, are for detecting activity/location and mode.

Activity/location algorithm may detect activity/location close to real time. In some cases, this will eliminate the need to rely on the memory function. In other cases, it may verify if the activity is the same as the one stored in the smartphone memory. In a similar way, the smartphone will store mode and occupancy between frequent origins and destination after adequate confirmation. The app will use the stored information to detect mode and occupancy in the future without asking for input. Mode-detection algorithm will interact with memory function the same way as activity/location algorithm does. Memory functions and automation algorithms will in general complement each other to produce more accurate data. Collecting other information like parking fee and transit fare may also be automated the same way.

Using the memory functions and automation algorithms reduces the need for user input and makes the app more efficient and user friendly. The need for user input decreases further as the survey continues and the system learns the participants’ travel patterns. The app will only ask for information that is not already in smartphone’s memory and cannot be detected by the algorithms. After a week or two (or ideally a month)*, the participants may stop interacting with the app altogether. And if they agree to leave the app running in the background, we’ll be able to collect large amounts of statistically significant data without any input by the participants. To encourage the participants to continue the survey to this point and then leave the app running, we may consider a direct incentive like paying for participants’ data plan or the entire phone bill for all the days that we receive correct data from their phones. This kind of data is not available through traditional data collection methods and will provide valuable information especially for longitudinal studies.

A set of apps for basic cellphones may be developed to expand the pool of potential survey participants. In that case, the participants will go on line to enter their trip information or correct/confirm what has been detected by the system. At the present time, about 85% of adults in the US have mobile phones. They are divided almost 50/50 between smartphones and basic cellphones.

*Length of the survey required to get to this point will be determined by field testing and analyzing the data.

LAUNCHING THE SYSTEM

After the completion of development and successful testing, the system may be made available on line to all public or private entities who are interested in conducting a travel survey. All they need to do is to recruit from their target population and direct the volunteers to the Mobile Survey public website (or app stores) to register and participate in the survey. The data will be collected and presented on line through the interface. Since no Personally Identifiable Information (PII) is included in the output, anyone can access the information extracted from the data through the on-line interface.

If the participating agency decides to pay for the survey participants’ phone and/or data plans as an incentive, the system may include a feature to facilitate transactions with the phone companies on behalf of the participating agency.

FUTURE IMPROVEMENTS

One of the main characteristics of this system is that it is flexible and subject to continuous improvements without disruption of the on-going survey(s). All aspects of the system, especially the demographic questionnaire and activity flowchart will continuously evolve and improve, based on our experience and feedback from planners, modelers, statisticians, IT professionals and the general public. The demographic questionnaire may be greatly expanded without too much burden on survey participants to facilitate detailed household travel surveys.

The data collected by this system is different in nature from what traditional travel-activity surveys collect. We may be able to develop postprocessing methods that would provide new and useful information for transportation studies that is not available through data collected by traditional methods.

We will continue to improve the automation of data collection. Advances in mobile technology will provide new possibilities for this system. We will constantly monitor new innovations and adjust our system to take advantage of the latest technology available.

Other future improvements may include addition of voice capability, and providing survey material in Spanish and other languages.

Project Manager:

Mohammad Assadi
Phone: (916) 654-4238
Email: mohammad.assadi@dot.ca.gov

Updated: October 2, 2012