[OSM-dev] "IQ Routes"-style averages from actual users driving the road

[Aaron Whitehouse]

On 12/01/11 12:15, Eric Marsden wrote:
>   Lukas Kabrt wrote some code to implement the data analysis side of
>   this idea during the last Google Summer of Code. It works by analyzing
>   GPX data. This could in theory work with the current facility for
>   uploading GPX tracks to OSM (as long as they are public), but it seems
>   likely to me that a dedicated service for saving GPS traces with more
>   privacy protection would work better for this usage.
>   
>   http://wiki.openstreetmap.org/wiki/Routing/Travel_Time_Analysis
> 

That's brilliant! That seems like a great way of doing it (I thought of
that approach, but thought it would be really difficult). I didn't guess
that all of the hard work had been done.

I have two thoughts from this:
(1) The webpage/API should probably be updated to let people tick a box
and submit it to "travel time analysis". It probably only makes sense in
the context of car GPX logs, so one approach would be to modify the
upload GPX page to let you select Car, Bicycle, Public Transport,
Pedestrian or Uncategorised (with all existing ones being treated as
uncategorised). Otherwise, the analysis couldn't tell the difference
between a car in heavy traffic and a bus stopping at bus stops. You
could probably put some cleverness in the analysis to get rid of walkers
etc, but it would be easier to just have a tick-box (checked by
default), as people then wouldn't submit inappropriate traces (say they
were curb-crawling for some reason!) Even if nobody is doing anything
with the logs at this point, it would start building up a repository of
them.

(2) The checkbox would have the advantage that the privacy issue would
be separated. One could choose to submit the track to the traffic
analysis (which would let the OSM server have access to the full
information (or at least a "trackable" copy)) *and* choose to only put
anonymous, unordered points without timestamps up for public consumption.

Given that the analysis code is already done, it would be great to have
a centralised implementation from the official OSM site (and
downloadable with the map). Then everyone could contribute to a common
resource.

Thanks all for your work so far!

Aaron