[OSM-dev] Fwd: GPX imports from Uber

Eric Fischer enf at pobox.com
Mon Jun 24 18:48:42 UTC 2013

I'm only a user of the GPX logs, not responsible for anything on the server
side, but this is fantastic news, and if you ask me, the more data, and the
more complete, the better it is, even going as far as every vehicle
sampling every second.

The main precedent I know for this is the Stamen/Yellow Cab/Exploratorium
Cabspotting project in San Francisco, which doesn't obscure trip origins
and destinations but does randomize the IDs of the cabs periodically so
that what a cab driver does one day isn't tied to what the same driver does
the next day, and only samples every minute so there is some inherent fuzz
in the origins and destinations.

The Skobbler GPX dump (http://heatmap.skobbler.net/gpx_download.html)
appears to include only partial trips so that the origins and destinations
of the trips aren't tied to each other but most of what happened in between
is still linked.

The main applications I see are improving road alignment, turn
restrictions, routing, speed, traffic volume, and origin-destination
aggregation, so I would appreciate it if your filtering tries not to
preclude any of these. Thanks again for planning to make this data public!


On Mon, Jun 24, 2013 at 11:12 AM, Jed Horne <jed at uber.com> wrote:

> I sent this to the imports list already and someone suggested I loop in
> dev.  Any advice is welcome!
> -Jed
> ---------- Forwarded message ----------
> From: Jed Horne <jed at uber.com>
> Date: Thu, Jun 20, 2013 at 2:07 PM
> Subject: GPX imports from Uber
> To: imports at openstreetmap.org
> Hi,
> my name is Jed Horne and I am a data scientist with Uber (http://uber.com).
>  My company makes an iPhone app that allows users to make on-demand
> requests for taxis, luxury sedans, and other vehicles.  We currently
> operate in 25+ cities in the United States, Europe, Asia, and Australia.
> We have GPS traces going back about three years from our drivers, and I am
> interested in contributing back to the OSM community.  I was planning on
> writing a script to anonymize and clean up our traces and export as GPX
> files (per instructions here
> http://wiki.openstreetmap.org/wiki/Recording_GPS_tracks).  However, I am
> very new to OSM contributing and was wondering if there is a set of best
> practices (how much is too much data, how to snip trips for privacy, etc.)
> or if there is someone I could work with directly to ensure that the data I
> give you is both private (for us and our clients/drivers) and useful (to
> the community).
> Specifically, I'm interested in using these traces to identify where we
> might be missing small connector roads or other features that could improve
> the accuracy of routing built on OSRM.  Another potential application would
> be to help identify areas of bad traffic or help improve speed profile
> information - I realize this isn't something currently supported by OSM but
> to the extent our data are useful for new or experimental features or data
> sets I'd like to know how to help out.
> If anyone has direct experience in this area I'm open to thoughts and
> suggestions.  Also, if anyone knows people who I should contact it would be
> awesome if you could make an introduction.  We have a very large volume of
> data that I hope can significantly improve the quality of OSM.
> Best,
> Jed Horne
> Uber Technologies
> _______________________________________________
> dev mailing list
> dev at openstreetmap.org
> http://lists.openstreetmap.org/listinfo/dev
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/dev/attachments/20130624/f6ba8900/attachment.html>

More information about the dev mailing list