Transport Analytics Lab

MODE: Mobile network Origin Destination Estimation

The project MODE aims at estimation of travel demand in cities and metropolitan areas, via utilization of signaling data in cellular communications networks. The key applications of using cellular network data range from dynamically managing road traffic to long-term infrastructure planning.

The estimation of travel demand from cellular data requires the analysis of large amount of data from the cellular network. The cellular data includes information of the connections made to the antennas of the mobile phone operator. Depending on which type of information that can be extracted from the system, the locations of the devices are updated at different events and points in time. The basic data is the call details records (CDR) which include the antenna connected to when a call or SMS is initiated. Additional information from the system can be extracted, for example, when the movement of the device requires that the device is "handed over" from one antenna to another in order to maintain the connection.

Origin-Destination demand from CDR data

Travel demand is normally described by origin-destination matrices. The matrix describes the travel demand, with the unit of the number of travelers, for a given period of time. For a "dynamic" demand description, the matrix is "sliced" into time intervals, for example, to describe the travel demand of one hour.

In the animation to the right, a dynamic origin-destination with a one hour resolution is shown for the city of Dakar, Senegal. The location of the origins and the destinations are the approximate coverage areas of individual mobile phone antennas. The animation shown is based on mobile phone call detail records (CDR) from the operator Orange, made available throught the D4D challenge 2015.

od-animation

Potential

The use of cellular network signaling data has the potential to fundamentally change how we can analyze the efficiency of a current transportation system, estimate transport models, and predict future transportation use. By mapping the cell phone data to the transport infrastructure it becomes possible to estimate the current use of the transport system. From the results of such estimations, suggestions for improvements to the existing transport system can be generated. The outcome would be more efficient mobility and, in the long run, increased economic growth. Furthermore, in developing countries the cellular networks can provide a much better coverage than traditional sensor infrastructure for traffic and transport. Therefore, this type of data will be very important to generate decision support information for large infrastructure investments.

The specific objectives are to enhance the capability of short-term prediction of road traffic and accumulate knowledge on using mobility estimation as an enabler in addressing future challenges in sustainable development of the transport sector.

Link flows from CDR data

Call detail records can be used for determining origin-destination demand, but it also includes data about the antenna connected to along the trip. This information can be used for assigning the travel demand to the transport infrastructure. In the animation to the right, the dynamic travel demand given in the animation above is assigned to the street network. The transport route choice can be studied by filtering out a subset of the trips that are well suited for each task. We have filtered out trips with a minimum number of visited cells in the cellpath and assigned them to routes an algorithm we call Lazy Voronoi Routing.
link-animation

Project papers

The following papers are published in the project:
  • Angelakis V., Gundlegard, D. Rajna, B., Rydergren, C., Vrotsou, K., and Carlsson, R., m.fl., (2013), Mobility Modeling for Transport Efficiency: Analysis of Travel Characteristics Based on Mobile Phone Data., proceedings of NetMob 2013, MIT, Boston, 2013, (PDF).
  • Gundlegard, D., Rydergren, C., Barcelo, J., Dokoohaki, N., Gornerup, O., Hess, A., (2015), Travel demand analysis with differentially private releases, proceedings of NetMob 2015, MIT, Boston, 2015, (PDF).
  • Gundlegard, D., Rydergren, C., Breyer, N. and Rajna, B., (2016), Travel demand estimation and network assignment based on cellular network data, Journal of Computer Communications, 95, pp. 29-42, (Article link).
  • Gornerup, O. Dokoohaki, N. and Hess, A. (2015), "Privacy-Preserving Mining of Frequent Routes in Cellular Network Data," Trustcom/BigDataSE/ISPA, 2015 IEEE, Helsinki, 2015, pp. 581-587, (Article link).

Project material

Some additional material from the project:

Project partners and funding

The project is funded by Vinnova. Project partners are Ericsson, SICS, Trafikverket, City of Stockholm, and Sweco.