Product release: Toll billing via GPS map-matching

We’ve just released a new toll billing product that will make it easier for car-share, car rental, and fleet operators to understand the cost of tolls incurred by their vehicles.

In order to bill and account for tolls incurred on a trip, companies typically have to wait 4–6 weeks to get billed by the relevant tolling authorities. To overcome this problem, some companies pre-charge their customers or drivers an estimated amount for tolls incurred on the route, which results in inconsistency between the actual cost of a toll and what is charged by the company. Others simply wait the 4–6 weeks and bill the customer well after their ride, resulting in a poor customer experience and additional support time spent explaining these new charges.

The request content sent to the Tolls API (left) and the resulting toll rate (right).

Our new Tolls API endpoint (Tolls on a GPS trace) solves this problem by taking in a GPS trace (a series of timestamped latitude/longitude pairs) and returning all tolls along the route with their respective prices at the time they were incurred. Typically, fleet operators are already tracking the locations of their vehicles. All they have to do is to call the Tolls API with the GPS trajectory (latitude, longitude, and timestamp). Sounds simple, right? That’s the idea!

Our Tolls API covers over 80% of tollways in the United States. We’ve spent a lot of time aggregating and normalizing this data, which is commonly housed as PDFs or images on tolling authority websites or, in the case of dynamic tollways, as live feeds of various specifications. Structuring this data into a single format makes it easy for developers to query the data and understand the true cost of a trip across cities.

For example, finding the toll cost of a route from Asbury Park to Midtown Manhattan involves either deciphering three separate rate tables (left) or calling our Tolls API once (right).

If you’re familiar with GPS data, you know that it has its shortcomings. GPS coordinates often aren’t exactly on the road, due to street canyon effect and other measurement errors. GPS is also unavailable in tunnels.

Another common issue is the frequency of GPS measurements. Fleet management systems typically collect vehicle locations only once a minute. And a gap of 1 minute or even 30 seconds is long enough to make raw GPS reading unreliable for applications like matching with locations of interest. The sparsity of the measurements makes matching GPS traces with the location of toll gantries unreliable.

Luckily we have a solution! The raw GPS locations can be processed before being used for toll matching. Such pre-processing is known as map-matching or path-inference, which is snapping GPS coordinates to the underlying road network and finding the most likely path taken by the vehicle.

The Tolls API first snaps the GPS coordinates to the road network, then compares the snapped geometry with the toll gantry locations to identify which tolls were crossed. As GPS measurements come with timestamps, the toll gantry cross time can also be estimated accurately by interpolating the time between the closest GPS locations to a toll gantry. The pre-processing step enriches the raw GPS data so that the Tolls API can reliably return the toll rates. The returned toll rates are detailed and categorized by vehicle information (number of axles, weight, height, etc) and payment methods (cash, transponder, credit card, etc).

Red circles represent GPS locations reported by a vehicle. They are labeled by minute:second. The green circle shows the toll gantry with 20 meters radius as the margin of error. If the geometry of the route (the blue line) intersects with the green circle (on the right direction) the toll is considered a match. In this example, although the vehicle crosses the toll, its GPS trace misses the tolls. The Tolls API overcomes this problem by preprocessing the GPS measurements and working with the underlying road network as opposed to the raw GPS locations.

With this product, car-share, car rental, and fleet operators can see the tolls incurred on a trip immediately after it’s completed, allowing them to bill right away. Instead of having charges appear weeks after your ride, imagine closing a car-share session or returning a car and having the tolls that you passed already included on your receipt! Having this transparency is impactful both on billing predictability and transparency as well as customer relationships.

Please get in touch with us at partners@coord.co if you’d like to schedule a demo or become one of our early access users. Also, feel free to chat with us on our site coord.co — we’d love to hear from you.

Also, if you are interested in map-matching, path inference, and travel time estimation from GPS data, feel free to read my research papers on these topics or contact me at mahmood@coord.co.

This article was written in collaboration with Amy Kyleen Lute.