I am about to do a project that involves "trajectory" prediction of human gps tracks. Preferably I need data of "agents" over months. In need some sort of "routine patterns" in it (e.g. work <-> home). I was wondering if there are any public available data sets that are simulated (because of the privacy issues of such public data).

I am already aware of actually a public real data set from Microsoft: http://research.microsoft.com/en-us/downloads/b16d359d-d164-469e-9fd4-daa38f2b2e13/

However, to start with I'd prefer a simulated data set (no gaps, etc.).
Has anyone knowledge of such data sets?

Alternatively, I'd be thankful if someone knows an easy to use tool that can simulate agents and gps trips - would be also ok to just have car tracks.

P.S.: other public real data sets are also welcome :)


3 Answers 3


To simulate GPS tracks, consider using the Optimal Roadtrip code from Randy Olson.

Steps (taken directly from ipython notebook):

  1. Construct a list of road trip waypoints

  2. Gather the distance traveled on the shortest route between all waypoints (using Google Maps directions)

  3. Use a genetic algorithm to optimize the order to visit the waypoints in

  4. Visualize your road trip on a Google map (not important for your application)

enter image description here

  • Thank you, I like this approach more than the Twitter one (are there really guys out that post every few minutes when not sleeping? :D. However, to do that over months I'll need some sort of "GPS-location-diary" that are the waypoints.
    – elDan
    Mar 12, 2015 at 17:11

A real dataset that might be useful is GPS traces from OpenStreetMap. Not all traces can be downloaded as GPX files including timestamps, but those marked as identifiable when uploaded can be.

For an overview of the data see https://www.mapbox.com/blog/openstreetmap-gps-layer/

For a listing of available GPS traces see http://www.openstreetmap.org/traces

There is also the OpenStreetMap Planet.gpx file But this is a dump of all uploaded GPS points, and they are not split into separate traces.


It doesn't fit the criteria of "no gaps", but you can:

  1. Get geo-enabled tweets from the Twitter Public Stream API

  2. Screen for prolific tweeters (maybe 30k+ total tweets)

  3. Download all available tweets from prolific tweeters (capped at 3k most recent, I think, details)

enter image description here

Source tweet (OC)

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.