Written by Brian McConnell
“What gets measured, gets managed.” — Peter Drucker
I travel a lot, so I spend a lot of time waiting in line — at security, at the gate, on the tarmac. As I was standing in line recently, it occurred to me that mobile phone systems are a ready-made source of data that can be used to do continuous time and motion studies. For the purposes of this article, I’ll focus on air travel, although the concept can be applied to any situation such as highway congestion, crowd flow in public spaces, etc. Even without high-precision location sensing, such as GPS, you can effectively track people as they move through an environment like an airport.
Imagine that you are traveling to another city. You arrive at the airport and go to the check-in area, at which point your mobile phone registers with a cell site that serves that zone. If the airport you’re in also happens to be equipped with microcells, the network will know where you are in the building, too. You take half an hour to reach the counter and check in, and then proceed to security, which is served by another microcell. Your phone jumps over to the cell site serving the security corridor, and again the network can see how long you remain on that cell site. From there you proceed to the gate area, and another zone, and finally to your plane.
If one or two larger mobile operators participated in this program, the facility could collect a good statistical sample, and could analyze crowd flow and queues in near real time. The mobile operator would provide the facility with a log feed that contains anonymized data from on-site cell sites (for example, by assigning a random number to each unique phone). This is a simple way to turn a mobile phone network into a tool for time and motion studies, and as mentioned, it is possible to do so in a way that provides a facility with great data, but also protects user privacy (though one can argue that if you are at an airport, you are in a secure public area and are being monitored via cameras and other methods anyway).
Tracking people as they move through a large facility equipped with microcells is easy enough, but what about tracking flights and real-world delays? Again, we can use the cellular network to do this. Think about what happens on a typical flight. Pretty much everyone on the plane is carrying a cell phone, most or all of which are turned off at pushback. This pattern is easy to see, as signals from anywhere from dozens to even hundreds of phones all vanish within a minute of each other. These same phones later re-appear in a different city after landing, making it easy to automatically identify clusters of phones, and by extension, individual flights (a flight from SFO to Cincinnati, for example). Using this data, one could estimate the actual time from pushback to landing quite accurately — to within a minute or so — and could use this information to cross-check data from other sources.
The cost of building this is minimal, because it leverages infrastructure that is already there. Provided that one or two large carriers participated, you could collect a statistically useful picture of crowd flow by sampling just 10 or 20 percent of the phones moving through a facility. Participating carriers would just need to write some software to create a service to expose data in the desired format.
I use airports as an example because there are consistently high densities of mobile phone users in them, and because delays are such a problem at many facilities. With better data, airport managers could gain a better understanding of how people move around, traffic patterns, and where the bottlenecks occur.
Since most networks now have low-resolution location sensing capability (for E911), it should also be possible to create general purpose services that allow other agencies, and the public, to get a real-time view of how people move through a city or highway system. In that case, however, the question of how this could be done while still protecting user privacy arises. If it is possible to find the right balance between openness and privacy, it would enable lots of practical and entertaining applications. Some examples:
- Traffic: provided location resolution is decent, one could calculate average traffic density and speeds on a street-by-street basis for an entire city or region.
- “Hotspots:” a heat map could display an estimate of the population density, allowing users to see how crowds form in a city (or if there is a large gathering somewhere).
- Trails: it would be pretty interesting to see how people move around in a city; this could be depicted as a map with different color paths overlaid on it.
City managers could use a system like this in many different ways. Public transportation is a good example. San Francisco operates one of the country’s more extensive public transportation networks, especially for a small city. Its managers may have a rough sense of how many people ride different lines, but that’s about it. With a system like this, they would be able to track passengers door to door. Someone traveling by bus or train would be easy to spot. They’d travel slowly from their home to a bus stop or rail station, travel fast or a while, then slowly again. It would not be hard for someone to write a program that would sift through data for a whole city to pick out the people who are catching a bus or train, and from that, measure ridership by station, real travel times, and other useful information.
Privacy is clearly an issue here, but it should be possible to design a service that provides statistics but does not identify individual users. Some policies that could be implemented to do this might be:
- Phones are anonymized with a different serial number each day, unless the user says it’s OK to include identifying information (opt in).
- Stationary phones do not appear in logs, only phones in motion are visible, so people could not see if a phone is active at your home (e.g. your gardener’s unscheduled ‘visits’ with your spouse).
- Data is only accessible via trusted intermediaries who could provide additional controls, opt-in/out options, etc.
- Phone trails are truncated, with a phone’s trail throughout a city randomly cut off at both the beginning and end so an observer cannot infer who it is based on start or end location
Some operators already provide services that allow you to see where your friends are (I remember Boost Mobile advertising this feature), so the capability is already there. While I am not interested in being able to see where my friends are, I think it’d be cool to be able to see different types of maps for traffic, where people are gathering, and who knows what else. I remember when NextBus first started operating in San Francisco; it’s a great service if you use public transportation often. I think there are many interesting urban mapping applications yet to be developed if this type of data were made available.