« Congestion Pricing, Positioning, and Meshed Wireless Networks | Main | The RPA has Wheels »

Urban GPS is Now

Generally speaking, the Global Position System does a great job of letting you know where you're at. A well known problem in urban areas results from tall buildings which occlude satellite signals as well as reflecting them, causing so-called multi-path errors. A good example of how these problems preclude the use of GPS for certain urban applications can be found in Transport For London's "Technology Trials." The image below shows the size of the buffer zones that would be necessary to calculate with 99% certainty that a car with GPS entered London's Congestion Charging Zone:

(source: Transport for London)

Enter, Skymeter. To enable their GPS-enabled parking and congestion charging business, they have developed algorithms for correcting GPS signals in urban canyons, as shown here (white is the position from a standard GPS chip, yellow is the post-processed SkyMeter position):

(source: SkyMeter Corp)

Toronto? Do they even have tall buildings there? Well, SkyMeter recently went to London to test their system in the same areas of London that caused TfL's GPS vendors so many problems. The results are similarly impressive (red = standard GPS, green = SkyMeter)*:

Zooming in a bit:

(check out the cool PostGIS, GeoServer, and Google Maps-powered interactive version)

To get a sense of what they were up against, take a look at TfL's calculations of the number of GPS satellites available in central London (red = less than 4, yellow = 4 to 10, green = more than 10, all in the 99% confidence interval):

(source: Transport for London)

Even after the Galileo (the euro-GPS) goes online, the test area is still no better off, according to TfL's models:

(source: Transport for London)

I think the data speak for themselves. Skymeter is currently planning a number of pilot installations including one for parking in the City of Winnipeg, and I wonder what London and Stockholm, or other cities contemplating congestion pricing (eg New York), will do with it. How much it really changes the discussion of what technologies are suitable for congestion pricing right now I'm not so sure.

* A disclaimer from SkyMeter, who was generous enough to give me their data to make those London maps:

This data is the "before" (red) and "after" (green) of the data processed by Skymeter with its first alpha test of data collected in London, UK, in heavy "urban canyon". This same data was analyzed by a third party (Mapflow from Ireland) in London. They calculated error as point to line perpendicular distance, and reported a reduction in the 90, 95 and 99 percent error quantiles of 26% 34% and 48% respectively (i.e. in meters). While this says the Skymeter process clearly removes error, this metric does not measure the second to second error variance, in otherwords, Skymeter removes absolute error AND makes the process much better behaved. This provides [1] a far-stronger evidentiary record, [2] two to three orders of magnitude greater compression, [3] spatial error bounding and [4] extremely rapid pricing-map registration (a trivial form of map matching at the data center)

This process is the first of four parts that [1] de-noises the GPS positioning signals, [2] characterizes the residual error for non-refutability, [3] bounds the spatial error, and [4] bound potential financial error. This process is part of a patented process that Skymeter claims can reduce tolling errors for GNSS-based tolling for distance-based road user charging as well as GNSS-based parking metering to arbitrary levels -- for example to one bill in 10,000, or to an arbitrary percentage such as 0.01% of a bill -- including in urban canyon.


This page contains a single entry from the blog posted on April 12, 2007 1:16 PM.

The previous post in this blog was Congestion Pricing, Positioning, and Meshed Wireless Networks.

The next post in this blog is The RPA has Wheels.

Many more can be found on the main index page or by looking through the archives.

Creative Commons License
This weblog is licensed under a Creative Commons License.