Optimizing public transport systems requires large amounts of continuously updated input data to model mobility patterns of citizens.
As of today, most public transport authorities are still relying on simple household surveys and manual passenger counting resulting in inaccurate, expensive and very static models.
While the electronic ticketing systems deployed in more and more cities provide more detailed input data, they are limited to public transport usage and do not provide any info about non-public transport mobility.
Mobility data from mobile networks does not have such limitations, and several IT companies are already providing solutions to derive origin/destination matrices for public transport companies from anonymized mobile network logs.
Most of these solutions are using offline processing on previously exported bulk data and the data source is usually MSC CDRs. However, these only contain location information when the user is making a voice call or sending/receiving an SMS.
Combined with the offline nature of processing, this makes these solutions unsuitable for any shorter time-scale optimization use cases.
Making better use of bigger data
The Public Transport Optimization project at the Ericsson Garage in Budapest uses real time terminal location streams from the Ericsson Expert Analytics product, making user locations available whenever there is any network activity from the terminal.
This means it’s possible to harvest mobility data based on foreground or background data traffic alongside voice calls and SMS. Alongside user mobility data, more and more public transport providers are using GPS trackers for vehicles. The regular position reports they make to traffic control centers provides a real time vehicle location stream, and is often made available to third parties via standard interfaces like GTFS Realtime.
By combining real time data from the mobile network and public transport systems, we can correlate vehicle and terminal paths to identify the vehicle that any given user is riding at a given point in time.
While lower location accuracy from mobile networks means these are only estimates, aggregating estimates for all mobile network users provides an origin/destination matrix with the number of users moving from a given origin area to a given destination area, as well as a drilldown of this number per public transport combination used. Data for non-public transport users is then estimated using heuristics on speed and other patterns.
All this is performed near-real time, with estimated latency of between 15 minutes to one hour.
The future of transport optimization
The combination of this big data analysis capability and near-real time operation opens up the public transport industry to several groundbreaking new optimization use cases.
These include continuous service adjustments to changing traffic demands, for example when new office areas or residential areas are built; immediate evaluation of the real impact of every service change, such as the percentage of people switching from car to a new public transport line or vice versa, and; optimizing replacement services during construction work or service outages.
With the innovations developed in the Ericsson Garage Budapest, public transport authorities will be given the tools to leverage bigger, better data than ever before, and ensure that the world’s cities make more intelligent use of their public transport infrastructure.