Real-Time Pedestrian and Bike Location How can we get it? What can we do with it? How can it not be creepy?
By Eric Fischer.

My summary of the discussion

There are many existing data sources that are published or have APIs that could stand as reasonable proxies for tracking people who are walking, biking, or just ambling around the city – some of this information is given away (via Foursquare) by those who are traveling, and other information is collected in real time (buses and taxis) and after the trip (travel surveys and Flickr photos). I don’t think the group agreed on any good use for this data (knowing where people are in the city right now), nor did the group come up with ways to ensure this collection is not “creepy.”

Eric’s original question involved the location of people bicycling, but the discussion spent more time talking about pedestrians. However, some techniques in tracking and data gathering could be applied to both modes.

See final paragraph for links on “further reading” that I find relevant to this discussion.

Schedule board at TransportationCamp West on Saturday in San Francisco at Public Works SF, 161 Erie Street.

[Ideas and statements are credited where I could keep track of who said what, and if I could see your name badge.]

Eric, starting us off:
We have a lot of information about where motor vehicles (MV) are in cities.
A lot of experience of city is not about being in a MV, though.

How many bikers going through intersection that are NOT getting hurt.
Finding places where people walk and where people’ don’t.

Where do people go on foot and on bikes?
As far as I know this isn’t available

Foursquare has benefits (awards) so people are willing to give the data, but we don’t want another Please Rob Me.

In SF, there are flash mobs, sudden protests, Critical Mass

Data sources:
-buses – boarding and deboarding – you can get a flow map from this. Someone said that Seattle has this data open.
-CTPP (Census Transportation Planning Package)
-city ped count
-Eric: Where people get on/off taxis.

“CycleTracks” – sampling bias, people with iPhones
-70% of handheld devices are feature phones, not smart phones. So there’s another sampling bias.

Opt-in factor
How do you sample?

SF Planning Dept. had a little program or project ask people to plot on a map your three most common walking routes.
What is your favorite street, and where do you not like to walk?

Eric: My collection tool is Flickr. Geotags and timestamps.

Magdalena Palugh: Are there incentives for commuting by bike? There are incentives for people who vanpool.
If there is incentive, I would gladly give up my data.
Michael Schwartz (SFCTA, sp?) What is difference <> SFCTA/MTA?

-If part of this is to get at where the trouble spots are, could you have people contribute where the good/bad parts are? “This overpass really sucks.”

Tom: Can you get peds from aerial images?
-Yes, but there’re too many limitations, like shade, and tree cover. Also, aerial images may be taken at wrong time (for a while the image of Market/Castro was during festival).

Brandon Martin-Anderson: What strategies have you tried so far?
-aerial images
-Flickr/Picasa location
-Street View face blur (a lot false positives)
Anything you plot looks kind of the same.

People like to walk where other people are. For safety reasons. -Good point on real-time basis.
Eric: Not a lobbying group for peds.
Eric: Find interesting places to go.
Richard: We need exposure data.

Paris bike sharing report showed that “Cycling is faster on Wednesdays.”
Europeans more open to sharing their private details – possibly because of stricter regulation on what agencies can do with the collected data. (There was a little disagreement on this, I personally heard the opposite).

Andrew: Can we use something like Xbox Kinect to track these people?

National Bike/Ped Documentation Project – same format
Seattle – 4 different groups that do annual bike counts. UW bike planning studio.

Who pays for this?
-Transportation planners pay for this.
-Private development projects (from contractor).
-Universities, NSF, Google
-Community groups –

Further reading


Mike Fleisher – DS Solutions
Andrew – @ondrae –

Notes to self

Is Census question about commuting about time or distance of “most traveled” mode?
Splunk – data analysis tool
What is difference <> SFCTA/SFMTA?