Part 1 - Reconnaissance and Failure

June 13

Sometimes if you’re biking or walking around certain cities you’ll see markings on the ground that look like diamonds or boxes. After I first noticed these a few years back, I found out that these are actually induction loops that have been embedded in the pavement to detect the passing of bicycles. Different variants will use different shapes and sizes (these diamond ones I’ve come across seem to be ‘ZELT’ loops by the company Eco-counter), and the technology can be used to classify vehicle types (cars, bikes, scooters, etc.). They’re most commonly used for traffic and drive-thrus.

I’ve had a question in my head for a while now: “What is the least bike-like thing I can get to count as a bike?” To figure this out I’ve first got to understand how they work + find a counter with a live display.

1. HOW DO THEY WORK?

The conductive loop has an alternating current sent through it, which generates a magnetic field. When you pass a conductor (like a bike wheel rim) over the changing field, eddy currents are induced in the object. Those currents create their own magnetic field that opposes the change in the loop’s field. This opposition changes the loop’s effective inductance, which is a measurable signal.

source: https://www.klkkus.com/zelt

Depending on the signature of inductance change over time that occurs, the system can classify what type of object likely passed over the loop. For a bicycle for instance, you have two metal wheels passing over the area closest to the loop at a somewhat predictable distance apart.

Eco-counter’s website says that their proprietary algorithm ‘SIRIUS’ uses 13 differentiating features. A patent of theirs from 2017 describes the general induction loop system but doesn’t talk much about classification details.

Another patent, US8779942B1, gives an example of how this might actually look. The loop geometry described in this patent is different, which changes the expected signature, but the core concept stays the same.

US8779942B1

“The signatures shown in FIGS. 4A and 4B were obtained using a parallelogram loop having the following geometry. The two shorter sides of the parallelogram loop are each forty-two inches long and each shorter side was located on the 10 respectively adjacent lane line. The parallelogram long loop sides were arranged at a forty-five degree angle with respect to the direction of travel along the lane. The spacing between the long loop sides matches the spacing between the front and rear wheels of a standard bicycle.”

Since the loop is set up such that it matches the bicycle’s wheelbase, the big spike/dip at (43) occurs when the front wheel and back wheel hit each end of the loop at the same time.

I made a couple simplistic simulations to see how the geometry/width of the loop impacts what the signature looks like.

To trick the bike counter we need to find a way to produce something close enough to the expected signature of a bike. Since the two wheels appear to be what produce the main changes, I’m focusing on how to recreate those. The main factors I can think of are:

• Material Properties / Conductivity
• Geometry
• Distance and orientation to loop / ground
• Distance between ‘wheels’
• Speed

I am quite curious as to just how precise the signature can get (if you know the geometry of a specific bike, can it be tracked by looking for a signature?), but maybe that’s a different post.

2. WHERE IS THE COUNT?

Even if I can mimic the signature perfectly, I won’t know if I’ve done it unless I can see a live update of the bicycle count. I’ve seen these before at certain locations around Montreal, Seattle, Waterloo, Kitchener, Winnipeg, and Vancouver (where I currently am). Unfortunately the two locations I was already aware of in Vancouver aren’t accessible right now. The most well-known one is next to Burrard Street Bridge, but the display has been covered up since 2024. There’s also one at Science World, but it’s been broken every time I’ve checked while passing by. The city has a map of where bike counters are installed around the city, but after scouring Google Street View for a while I couldn’t find any counter displays.

Ultimately I found a Facebook post mentioning a bike counter in North Vancouver. I went to check it out and confirmed that it was functional.

I also brought a few aluminum pans with me in the hopes that it might be that simple… it did not work and I’m sure I looked insane running back and forth across the counter holding pans near the ground.

So a couple conductive objects of similar size to wheels isn’t enough, at least not in the form factor I tried.

As for next steps, I’m currently trying to procure a cheap old bike from Facebook Marketplace so I can take it apart piece by piece until the counter stops working, then build up from there.

Results and hopefully a working solution will be in PART 2.

PART 2 - THE BIKE BASKET

June 14

Thanks to Neil on Facebook Marketplace who gave me a bike for free (he said on the listing that it was likely just useful for spare parts). It was quite useful!

Before I disassembled it, I took it over to the counter to do some more testing. I found that the orientation of the wheels doesn’t seem to matter much, but what does matter a lot is the distance between the wheel and the loop. I didn’t measure precisely but it seemed to be that if I held the bike anything more than ~2 inches above the ground it wouldn’t get counted. As I alluded to in Part 1, the coupling between the wheel and the loop falls off very quickly with distance. I also found that you can’t be going too slowly (there’s likely a cutoff for the time window the counter looks at to evaluate a signature).

Most importantly, I figured out that this particular single loop system doesn’t seem to care which direction a ‘disturbance’ is coming from (maybe you’d need multiple sensing zones or a classifier specifically for distinguishing reversed signatures).

Meaning, if I push one wheel over the loop, and then drag it backwards within a reasonable time frame, the signature is treated as more or less the same and it counts as a bike. This makes total sense and I probably should have realized it before. I also confirmed after removing the wheel from the frame that it is in fact the part of the bike that produces the important parts of the signature.

These findings make it easier to create some type of object that gets counted as a bike, since we can space things out temporally instead of physically. The only challenge is that it needs to go over the loop and back in a somewhat natural way.

The first thing that came to mind was the ‘swinging’ motion that happens when carrying a bag. I present: THE BIKE BASKET

In summary: I put a bike wheel in a basket, innocuously covered it with a shower curtain, and when you swing it over the loop it gets counted as a bike.

It’ll be hard to match the smoothness on display in that video in the next iteration, but all in all I’m pretty happy with this for a first attempt.