31st March 2023
The Spokesmen Cycling Podcast
EPISODE 324: Bike lanes empty? Prove gibe wrong with this €199 window-mounted traffic counting camera
SPONSOR: Tern Bicycles
HOST: Carlton Reid
GUEST: Kris Vanherle of Telraam
TOPICS: The Telraam traffic counting system uses a camera and AI to work out exactly what’s going by on the road. And it’s not just for professionals, you can buy one for your window and start measuring traffic 24/7, perhaps to prove that your road suffers from excessive speeding. Telraam’s developer Kris Vanherle of Belgium here describes his neat new system and whether it could be used to counter those tabloid newspaper columnists and others who claim that bike lanes are empty.
Carlton Reid 0:12
Welcome to Episode 324 of the spokesmen cycling podcast. This show was recorded on Friday the 31st of March 2023.
David Bernstein 0:28
The Spokesmen cycling roundtable podcast is brought to you by Tern bicycles. The good people at Tern are committed to building bikes that are useful enough to ride every day, and dependable enough to carry the people you love. In other words, they make the kind of bikes that they want to ride. Tern has e bikes for every type of rider. Whether you’re commuting, taking your kids to school or even carrying another adult, visit www.ternbicycles.com. That’s t e r n bicycles.com to learn more.
Carlton Reid 1:04
If counting sheep makes you sleepy, imagine how difficult it must be to count cars and cyclists. Boring Yes, but such traffic surveys provide essential data. I’m Carlton Reid, and today I’d like to introduce you to a different way of measuring road use. away that’s more accurate than the pneumatic tubes, you’ll often see stretched across carriageways, but which don’t measure very slow moving, or static motor traffic and aren’t even so hot on recognising bikes. The tell ramp system uses cameras and of course, a AI to work out exactly what’s going by on the road. And it’s not just for professionals, you can buy one for your window and start measuring traffic 24/7 Perhaps to prove that your road suffers from excessive speeding. This morning, I talked with tal rams developer Kris Vanherle of Leuven in Belgium. And I started by asking him to describe his neat system. And whether it could be used to counter those tabloid newspaper columnist and others who claim that bike lanes are empty. Kris there’s a there’s a today there’s a report in a in a British tabloid newspaper. And it’s very frequent kind of complaint. But it’s about empty bicycle lanes. So this particular right wing commentator is basically banging on about bicycle lanes being empty. Now you’ve got a product there that that might help to either prove or disprove that. So what’s your product?
Kris Vanherle 3:08
So our point is called Telraam. It’s it’s a small IoT traffic counting device. Actually, Telraam is a Dutch word. Funny word of play of words, because tell means counting. And is window and it’s the device what you put on the window. It’s you make your window Accounting window, but also telling them as a word in itself in Dutch is abacus. So it’s a nice play of words. So that’s, that’s our, what we do. So basically telling them is a small, low cost, traffic counting device which is owned and hosted by private citizens. So typically, traffic counting devices are used and deployed by professionals and so industry professional but also policymakers, local authorities, just to monitor traffic, what we’ve made first with the Raspberry Pi system, and now with a with a new custom made, hardware and software is a commoditized, let’s say cheap or affordable traffic counting device to also allow anybody to just start counting with with a very simple device. And indeed, like you said, that helps to objectify any discussions on traffic flows are either on there’s too much car traffic in my street, they’re speeding. Bike lanes are not being used, like this example. With data it’s just there so you can have you can at least have a discussion about data rather than gut feeling.
Carlton Reid 4:32
It’s that private citizens but that really interested me when I looked at your product information because yeah, this is not some is you there is like a professional version. You know, you can you municipalities can can get this from their own uses. But this is very much you could you might have a road that you’d quite like to your municipality to close off. You can put this in your window and then you can save that municipality. Look, we’ve got you and X number of cars coming through X number of bicycles coming through why don’t we do this? Is that the idea because because many bicycle advocacy groups already go out there manually counting this stuff. So you’re basically bringing something that that would manually counted and you’re doing this with, with AI in effect.
Kris Vanherle 5:20
Yeah, that that’s partly the objective. So just for some background, so I’ve been working as a traffic engineer over 15 years. And why we started with this project, because out of the frustration of lack of data, so we are either the technical data is lacking, and you basically need data to give property advice, or it’s usually expensive. So we started with this project to make it affordable. And our angle was to work with citizens because like you mentioned, local debt is a very hot topic, and local residents are very much willing to contribute time voluntarily to collect this type of data. Like you said, people can’t manually I’ve even seen people who have done origin destination studies with licence plates. It’s also I mean, people invest time are gorgeous, we should give them the tools to do it properly. So that so that the default so that the data which is being generated is actually useful for policy advice. We offer the device for private citizens, but we do want to work in a collaborative way. So actually, our model is to work with local authorities, and citizens together. So the most common model that we basically have is that we have a discussion with a local authority, they want to do something in their, in their jurisdiction on traffic, like closing a street, low traffic neighbourhood, for example, slow street, but they want to monitor before and after. So they buy devices from us. And we support the local authority to deploy these devices with citizens. So you get the dialogue immediately between the policymakers and the citizens. But apart from that, it’s also possible to just for a parameters to buy a sensor and start this discussion with with the local authority based on their data. So it’s a bit double. So we’re still professional with local authorities, but also enabling directly to towards citizens.
Carlton Reid 7:11
And how much is the product for one product for one person, one private citizen, how much they’re going to pay?
Kris Vanherle 7:16
Yeah, so we’re trying to distinguish. So if, because we really much in favour of grassroots organisations. So if you just want to buy a device that says no strings attached and simple, it’s one at 199 euro, which is still expensive, but if you compare to other ethical techniques, it’s like, three, four times more cheaper compared to limited use, for example. But if you work for a local authority, we set up a full project, we do engagements, we do, we have much more analytics in the dashboard. And then we’ve got the model, right? Where we are more or less, I think, maybe two to 400 euros per device all in. And that’s basically a model how we tend to survive. So provide professional services to local authorities, but give the devices for free via these local authorities. So they can contribute to the to the ethicality. And it works
Carlton Reid 8:06
as well for their for the telecom. So yeah, like a connection. Yeah,
Kris Vanherle 8:11
yeah. So the old device used Wi Fi for data connectivity, but the new device is using a SIM card. But that’s, that’s included. So for the private citizens, we want to make this one of cheap and we’re basically subsidising and because we’re the one off cost just for grassroots organisation. So the 199 is, it’s actually less than what we want, it’s costing us but we want to support these girls with initiatives are revenue models more from the project that we do with the local authorities, and provide more support and ethics and so on. And this is the model that we tried to build also trying to bridge this gap between citizens and local authorities because local traffic very antagonistic.
Carlton Reid 8:53
A minute ago, you mentioned pneumatic tubes, which are, which is often how, you know these these rubber strips that are laid across the road. There’s a bit of a box next to them, which is counting traffic. Now that’s been relatively controversial in the UK recently, in that somebody discovered that for a low traffic neighbourhood, it was in London, the traffic counts potentially could be out, because the pneumatic tubes don’t measure static traffic and effects of it if a traffic jam, you know, the cars are not going anymore, whatever, like I’d know five kilometres an hour, they don’t get counted. So it’s clearly a problem. So your product is solving that problem.
Kris Vanherle 9:39
Yes, in part. So I mean, again, this is technology. So it’s got limitations, but like you mentioned Dometic tubes. It’s been, let’s say the industry standard for a long time. I mean, I’ve I’ve bought this type of data many, many times and I know the limitations like you mentioned, slogan in traffic is an issue and because it depends on timing of signals, so if it’s really slow Although it cannot distinguish between a car passing by or it’s like a single pulse. By counting as an issue as well, I mean, there are two tubes which include by counting, but because they are less heavy, the signal which you get from the kinetic poses is less, so the accuracy is lower. So, these are all issues with pneumatic tubes and there are more, we’re trying to solve this indeed without without garlic, so its chemical based system. But that doesn’t mean it will be 100% accurate. Still, there’s still a trade off between cost and accuracy. But slow, slow going into effect for sure it’s not an issue because it’s taking objects and it is going faster, the objects are going very slow. But the object will be recognised. In fact, I think actually accuracy will be higher, if the traffic is slow, because there’s more frames where the object is, is in the view.
Carlton Reid 10:47
So the accuracy of this, you can track this just by setting your your product up on a window, and then at the same time, at a fraction of a moment in time, you then physically count. And then you tally the turn, and you say, you know how genuinely accurate this is, and you found that it is accurate.
Kris Vanherle 11:10
You can so what we have with, especially without a new device, there’s a screen on top of it. So and it gives you life counts. So if some some when some object passes by, so a car that skin cycle or glass vehicle, you will see the counter increasing. So and that’s also I think, for us important to have this transparency. So we are expecting an accuracy of 90 95% for cars and 80% ish, maybe a bit more for for bikes, it will depend on the locations we’ve tested in length on different sites. But it’s also useful for people in the field to have a device to can just validate and cross check, okay? The device on this period of time is indeed counting on a typical day, and you can do some manual validation. So this type of conspiracy for us is really important because it’s often lacking with with other technologies. So primitive tubes, we know that in some cases, it doesn’t work. In some cases it doesn’t. So you don’t know if you can rely on the data. In some cases, yes or no. In our case, what you see is what you get them and if you don’t work on your site, you will at least see it on the on the screen.
Carlton Reid 12:15
Now the advantage, I’m guessing of it, and we discussed the disadvantages of pneumatic use. But the advantage of it is, you know, their lateral there on the ground, and nothing can really obscure them. But presumably your product has its camera based things could obscure them so that there must be an optimum placement. style. So if you had it in somebody’s window, and say the bike lane that you’re trying to measure is frequently blocked by you know, traffic, car traffic and buses, and you can’t actually see the cyclists See you then that’s where your product isn’t quite so good. So how, what is the optimum placement of your product?
Kris Vanherle 12:58
That’s a good point. I mean, like you said, metal cubes. And basically you can assault anywhere where there’s where there’s room. In our case, you have to indeed, you’re limited by like a free field of view of the street. So there’s a few rules of thumb. So what we always recommend is, first of all, your first or second floor window, so not the ground floor, so you’ve got a bit of a downward angle. So that that that negates or avoids having blocking objects passing by. So that’s that’s one thing very clear, need to have a clear field of view. So no trees, preferably no parked cars in front. And not to this bit not too short, not too distant. So I mean, the houses which have like a small garden in front, like five metres or something perfect, because then you’ve got like a very nice view of the street. Now, we are very much aware that this is a limitation. So when we do a project with a local authority we have we’ve got a full onboarding mechanism candidates applying, and they upload a picture where they would instal the device, you’ve got clear instructions on what suitable sites? And then we evaluate if this is a suitable site. Yes, no, and be allowed to inform them? Yes, I mean, it will work on your site. So you can you you’re eligible for a device so so at least we have some control where the device is installed on locations where the count should work. Now, based on our experience, now we are working for two, three years now, I would say that about two thirds of the candidates that we get that probably eligible. So in most cases, if you have a clear view view of the street, the device should work
Carlton Reid 14:33
the product that you’ve got there 199 euros for the consumer, I’m assuming a bit more for local authorities, but when you when you place them either by the private citizen or the local authority, buying a bunch of them, and you place them in a window. I’m assuming there’s some sort of algorithm that you’ve got, which in This prevents double counting. If you’ve got a street and you’ve placed five of these counters next to the street, you’re not going to get you know, the counting the bike bicyclist five times the motorist five times, it kind of knows where it is where the device knows where it is. And it knows if there’s another device 500 metres away, it doesn’t doesn’t count those. So how do you how do you square that?
Kris Vanherle 15:26
Okay, that’s important, because the way we’ve set up our system is that you’ve got one device with a single code segment, what we call a code segment is not the full state, but the street, which is divided by the next intersection. So in principle, the road ethic on the street segment is the same on the full street segment, except for people who should be somewhere on the street segments or incidents, or anybody who’s not transit, basically. So we, we don’t really deploy multiple devices on a single segment, we just take one device, or if we have a volunteer as one device in a seat next segment, and we report the volume of this segment segment on a map public map for this segment, so we don’t really do this, for five devices on a single segment, it’s been done for validation purposes. So then you can take an average, for example, or you can make some small distinctions between a location a bit further up the segment and put it down. But they should be very, very much similar in terms of volume. So yeah, there’s no way of double counting in that sense. If we have to segment two devices on the same segment, we report the average of the of the of both devices on the segment. I think I didn’t mention this explicitly, but we tell them is on one part, this traffic counting device, the sensor, but it’s also an open data platform. So if you sign up with with Telecom, you are also sharing your data and we reporting that on on our website. Landing Page is basically a map with all this data.
Carlton Reid 16:58
So where where have you had your product placed in internationally where where has been counted so far? So
Kris Vanherle 17:08
we started three years ago, but earlier and mostly belted, because we live a Belgium based but gradually we have done some studies internationally. So we had we had a big European project in 2019 to 2021. That was in Dublin, Cardiff, Barcelona, Madrid and Ljubljana we’ve had a nice project in Berlin as well with with across with organisations like I can biking advocacy group, we’re deploying about I think, 100, it’s in the Berlin area. in Utrecht, Netherlands, we’re working with a governance. So again, it’s about 150 200 orders of magnitude. So gradually, also reaching the highest of of Europe, not just Belgium,
Carlton Reid 17:56
how have the results been used? Do you know how they’ve been used?
Kris Vanherle 17:59
It depends, it really depends on the use case. So I can give a few examples. So in some cases, it’s it’s really about monitoring the impact of an intervention, so an intervention low traffic neighbourhood, or whatnot. So anything which is changed, which would expect some fundamental changes to the traffic flows, and then you’d mentioned three, four months before the intervention, three, four months after you direct typical traffic before intervention, typical tactic after the intervention, you could just make comparisons. And that’s super useful. And because what’s typically be done is the stopgap measures have been dealt with the main roads, but for the small underlying roads, like the most smaller essential roads, that really being assessed when you have a change in the circulation plan or installation of a low traffic neighbourhood. And you want to understand if there’s not to say, spillover effects to adjacent streets and so on. So with this, because it’s cheap, you can you can really blanket an area with these devices. And you can do typical perfect pre intervention and typical African post intervention. So that’s, that’s one case. Another case is speeds, that’s also a good one, we have had a very nice case in Louisville, but there’s more which are similar. Where there was an issue of compliance to speed limit of 30 kilometre per hour. So you could really see from the telecom data that compliance was like 50%, half of the cars were driving faster. And there were two interventions. So one intervention was this this digital sign with your speed when you’re passing by, so you could really see when this intervention was done, because compliance increased from I think, 50% to 60%. So these things actually work, apparently. But then you could see a second adventure it was installation speed bump, and then you could see compliance with the speed limit, go to 95%. So, those are examples of how the data is being used to demonstrate the effects, impact of off interventions.
Carlton Reid 19:57
You know, one of the complaints from, from people Is the surveillance society we have here. And you know it capturing numberplate. It’s not what you are. But you’re not capturing numberplates here, you’re literally just there’s a lump going past at a certain velocity. And you are statistically measuring that. So there’s, there’s no, there’s no arguments here. Nobody can say this is surveillance society, you’re just counting numbers and the speed of the things that you’re counting. Yeah,
Kris Vanherle 20:23
exactly. And I mean, that’s really important. I mean, everybody says it, but it’s really genuine. And privacy is important for us. So it’s really, it’s really a thing. So we deliberately don’t do tracking, we deliberately don’t do number plate commissioning. We just do counts. So basically, what’s what’s coming out of the of the traffic counting device is the same type of data you would get from a pretty metric tube, a manual count, whatever, it’s just automatic, long, long duration, and there’s just much more data. But that’s the type of data which is being spit out of the device. And this is enough to do any quite some useful impact analysis of transformation. So yeah, obviously, when you would, when you do something with Ghana based systems, and especially if you give them citizens, you’ll get into controversies we’ve had our share of controversy has been every time we explain how the other device works. So it’s an example of edge computing. So the images which are being collected from the from the camera system, I instantly process on the device itself. And it’s just count data, which is going to our cloud. So you still images leaving the device, it just count data. So these types of things are very important to reassure users and stakeholders nearby. That text, no surveillance,
Carlton Reid 21:45
the objects that you are capturing, and you know, just just counting immediately and not sending anywhere else. And can you differentiate between a car, a van, a lorry and HGV a large truck, a bicyclist, a pedestrian, maybe even a cargo bike user? How how granular is your data?
Unknown Speaker 22:07
Yeah, very good question. So we’ve been working up to now with the Raspberry Pi based systems, we’re not the new flashy device you see now. And the old device categorises in four categories, so pedestrians, bikes, and cars and large vehicles, that classification is quite rudimentary and simple. So basically, it’s using Object properties. So for example, axis he have an object, so pedestrians are very narrow and very tall. So x axis ratio of the object is four. So it’s likely pedestrian while cars wide and not at all. So axis ratio is 0.5. We use a few other parameters to then classify that those objects but there is there is a clear risk of misclassification. So for example, the motorcycles and bikes there would likely be in the same category with the new device, it’s different. So it’s it’s immediate AI commission. So when you see an object and an object is associated, okay, this is clearly a bike or a pedestrian. So while we were now still bundling them in the same four categories, in fact, the device can see more categories. So you’ve got car, you’ve got no longer two wheelers. With the old device, the bikes and motorcycles with the distinction has already added. There’s pedestrians and then what we have now as large vehicles will be categorised in lorry so trucks and tailors, buses, and light light trucks. There’s still a risk of misclassifying cars and vans because yeah, sometimes they just look physically the same. But in terms of
Carlton Reid 23:39
like, what do you call an SUV? I mean, that’s that could be a van.
Kris Vanherle 23:44
So that actually actually the risk of misclassification with a new device is the same as a human would have. So if you’re counting manually, you would say, Well, is this event of an SUV? It’s the same time type of doubt that this device, but other than that there would be no clear misclassification?
Carlton Reid 24:00
And what about a scooter? Because that’s clearly a use scenario that many cities either want to clamp down on, or increase. So do you measure scooters, so these these push scooters,
Kris Vanherle 24:10
motorcycles, so for sure, but again, speed pedelec it’s a good example. So you’ll get to see the the blurrier area of very high end and fast electric bikes speed pedelecs which almost look like it’s good and you’ve got these electric scooters as well, which are more coming down from the scooter towards the bike as they insulate the grey area. I can’t tell for sure how the device will look or counters because sometimes they’re really physically look look quite the same, but anything which looks like a scooter will be counted as a motorcycle. So
Carlton Reid 24:48
the terminology I should I should I should I should have clarified that so when I meant scooters, I mean, the stand on things, steps that propelled you.
Kris Vanherle 24:58
So how would you what would you call them just apps like these small, small wheels, and basically walking on driving on pedestrian walkways, right, or
Carlton Reid 25:11
when you measure them, and you spot them tricky.
Kris Vanherle 25:15
Again, you don’t, it will be difficult to make a distinction between pedestrian and bike in that sense because they are small. So you might get missed out to the to the bulk of an object which is passing by is there is a person which is on the step, and the step itself is quite small. So, if the device misses the step, it will not see it, and it will go to this as a pedestrian. I think it’s a bit too early to make any definite statements on that we’ll have to see we just launched a new sensor. And we’ve tested extensively before, but these are specific cases which we have not had any definite validation on. So we’ll have to see.
Carlton Reid 25:53
So you mentioned a few times the old raspberry pi version, and the new version. So talk me through the history of this when when you first had this idea, your first product, what year was it? Was it for instance, crowdfunded Kickstarter, whatever, to your, your current product. So So talk me through the history of your product,
Kris Vanherle 26:15
the origin, the needs of why we started with this project was was because of relaxed data, affordable data. So we started with that from within the company where I work, which is a spinoff from the University of K Leuven, with a research can’t like, not too big. And we basically said, let’s try this, let’s let’s try to make something with it as a pie and see if it works, what studies and we did a small pilot here in Louisville. And, to our surprise, I mean, there was two key outcomes there. So to our surprise, it worked fairly well, because we made something very simple, with actually available off the shelf hardware and software. So we were wondering, why isn’t this being done before. So that was one. And the second thing is that we had no issue whatsoever, finding people who wanted to host the sensor and do some debugging and developing with us. Now, I appreciate live is a bit of a special university town. But still, I mean, we had I think 250 candidates for 100 devices, and it was like this to find them. So that made us think, Okay, this is a model which could work. So if we can further debug and improve this device, we will definitely find people to host something like this. So that was 2019 2020, I would still call this a proof of concept, we have a bi based system over time, and we did support it and also did the European project and Dublin guidance and so on, we basically came to the conclusion that if you want to reach a large audience, we have to detect the sensor. So they have a bio based system is nice for techie, tech lovers and enthusiasts who want to go to the pain of installing this because it does require some, some technical, know how Wi Fi configuration installation. So pretty soon, we have chosen to go down the path of making a dedicated sense of hardware for for this purpose, which is focused on ease of installation, to also be able to reach anybody who just want to count ethic and I mean, prefers not to go to the pain of installing it has been a buy. And so that’s been done for the past two and a half years. And it’s been a long road. We have also had research funds to do that and some development funds. And we only just launched, I think two months ago with with the new sensor. And now we really ready to scale up with this new device. So we’ll still support the Gatsby by base system because we also tech lovers and we’ve got a nice community of tech lovers and we really want to support them. But for the future, we will definitely look at the new sensor because it’s simpler, user friendly, and we will be able to reach a much larger audience for the public.
Carlton Reid 28:58
I’m going to I’m going to cut to an ad break now I’m going to go across to my my colleague David in America but I will be right back with his I do hang on there for a second. Take it away, David.
David Bernstein 29:08
Hello, everyone. This is David from the Fredcast and of course the spokesmen. And I’m here once again to tell you that this podcast is brought to you by Tern bicycles. The good people at Tern build bikes that make it easier for you to replace car trips with bike trips. Part of that is being committed to designing useful bikes that are also fun to ride. But an even greater priority for Tern is to make sure that your ride is safe and worryfree and that’s why Tern works with industry leading third party testing labs like E FB E and builds its bikes around Bosch ebike systems which are UL certified for both electric and fire safety. So before you even zip off on your Tern, fully loaded and perhaps with the loved one behind it you can be sure that the bike has been tested to handle the extra stresses on the frame, and the rigours of the road. For more information, visit www.ternbicycles.com. To learn more. And now, back to the Spokesmen.
Carlton Reid 30:18
Thanks, David. And we are here still with Kris Vanherle. And Kris is from Leuven. And I want to dig into Kris, I want to dig into your because you mentioned a couple of times now that you are you’re, you’re a traffic engineer, basically. And I’m assuming that’s exactly now. So what does that job entail? What have you been doing? How does this improve maybe the life of fellow you know, the people who are who are doing your job now? And is that product taking you away from that job. So that’s three different questions.
Kris Vanherle 30:53
15 years I’ve been doing this, I’ve seen the the space of graphic data being evolving immensely over this last 15 years or so. Cell phone data, for example, let’s start with that. It’s been I mean, when I started my job, I think cell phones weren’t weren’t even that prevalent yet. And casually. Cellphone data was being recognised as something useful for doing origin destination tracking. So you know, where people are coming from where they are going to super useful to, to calibrate different models, for example, which then allow you to do an ex ante analysis and try to model on before and before we do an intervention, how traffic would behave, if you would do an intervention. Before that was all speculation and theoretical. That type of data has really, really evolved. And it’s like now the standard go back to Google traffic, for example, is a product which provides origin destination data, travel time data, which is usually usually interesting, and has really changed the environment of traffic engineering, allowing different models to be much, much better.
Carlton Reid 32:00
Now, basically, that that’s using the Bluetooth on a phone to basically measure where people are going to and from it knows probably where people are just by speed, you know, they’re on a bicycle or in a car. And then it’s the origin destination means, you know, are they genuinely doing one mile trips? Is that Is that what you mean?
Kris Vanherle 32:18
Or they live in this area and go to that area where they work. And it’s generating congestion in between? Because there’s so many people coming from there and who choices for example, why are people or why are people taking a restaurant, for example, those things you can all make visual. Now. I mean, at the beginning of of when this data was useful, there were sincere privacy concerns which completely justified but by now you have topped with data products, which all are hashed and anonymous, and non Nice. So it’s much safer and clearer now. But still, sometime, it’s a bit creepy. But this type of data are extremely useful to calibrate traffic models, which you really need to give to make corporate data and appropriate and policy advice. So we had the epic moments 50 years ago, still, but they were scarcely calibrated, and they were to get the best. But now, I mean, definitely much more performance much more accurate. Now, the point where I want to come to and my time is still important and useful is that this type of origin destination, which is floating garden is one super useful, but it’s still just a sample. So you still you will will not have all traffic, which is having a cell phone in the car, for example. So you will not it will not provide you absolute values on a specific site. So that’s still valuable to collect them to calibrate models. And secondly, this is still just cars. So we have achieved to calibrate effort models for cars properly now, in the traffic engineering world. But bikes, bike models, only scratching the surface at this point. And, and we already know that route choice behaviour and so on with bikes is completely different. The steepness of the hill, the quality of infrastructure, the perception, or the heel safety, those are all things which are influencing good choice of of bikes. And there’s there’s definitely a capsule to get that right. And we tell him, you know, we’ll have to have accounting data not just regarding also for bikes. So it will allow to get to make the next step, let’s say and also get good corporate traffic models with with bikes. So you can do also proper policy development, which is supported by data, not just for cars, but also bikes.
Carlton Reid 34:36
And then the one of the other parts of that question was are you still doing the traffic engineering? I’m always like, taking you away from that field.
Kris Vanherle 34:47
Yeah, yeah, I’m completely consumed now by Telraam development. So I mean, we so Telraam was started as a project from within TML we we did this epic engineering but because it’s that was consultancy, Telraam as a product, so we spin it off as a different company, because it’s completely different. And I would say we’re in our own startup life and trying to get this product as good as possible. So it’s a completely different focus, to a development firm with development, use management, and so on. So, yeah, I’m, unfortunately, I’m not really doing much anymore. When I when I, when I’m lucky, and I can do a workshop with data analysis, then I can dive into the, into the telecom data and do some good old data analysis, which is really fun. So if I get the chance I come to it, but it’s more for the colleagues from from TML, who were working with a data and third parties,
Carlton Reid 35:42
TML. What’s TML?
Kris Vanherle 35:45
TML that’s the agency where the traffic engineering Bureau where it was within the spinoff from the key. So the colleagues.
Carlton Reid 35:52
So you guys, because your background, as a traffic engineer, you know, this product is going to be incredibly useful around the world. You know, this is not something that you’ve come from the world of tech, and you’ve brought this to an alien world that you are from this world, you know, this product, it will be incredibly useful.
Kris Vanherle 36:13
Yeah, yeah, that’s correct. And, and that’s why, but I’ve grown to understand over time, that’s why I at first, I didn’t understand why tech world hadn’t already done this, because it’s not that hard. And the use case is very clear. So while we came in from the immediate the application side, and we had to learn the back, let’s say, so we had to learn how this technology works and work with partners, but the application for us was evident. And we also can see it and the development towards what we think is definitely the most applicable use case. So simple, cheap, and very specific on a very specific type of traffic, counting data. So on the specific sites, no intersections, no origin destination, because once you go down that route, you’ll get you’ll end up with high end system, which will cost you multiple 1000s of euros per device. And we think there’s definitely a use case for her first cervix, very simple, low cost epicanthic. device.
Carlton Reid 37:10
So talk me through it because you said it. This is like two months in which you’ve you’ve had this the the newest iteration of your product, where do you see your company going? How many devices do you think you’re going to be able to sell to a to private citizens and be to local authorities? So what’s what’s your Give me your business plan, so not just your business? Product history, your business plan? If you don’t mind, Kris,
Kris Vanherle 37:36
we’ve seen what the current has been by this system that we’ve gotten a lot of attention from Belgium. But we’ve also seen that it’s, it requires all support. So we were a bit reluctant and hesitating to go international with this, because we want to be closer to be able to provide support. So. So while we’ve been having a lot of questions, in the meantime, while we were still working with it has been a buy and developing the new system, we have been holding back a bit on international inquiries. Now we’ve got the new sensor ready. And it’s not, I would say 90% validated, we still need to check off a few things. But we just want to open the international tab, let’s say and want to roll out as fast as possible and as many as possible, because the need is there. So if you if you’re asking me for numbers, I am hoping to deploy at least 2000 a year but rather looking at several, I mean, five to 10,000 additional new devices per year being deployed worldwide. We’re having a focus because we’re based in Belgium on Benelux or Belgium or the Netherlands. But we have somebody who’s working for us already in London. So UK will be a focused market to expand. And we already also getting some inquiries from from the US. So hopefully, we can also quickly built on that and serve that market. Now in terms of the distinction between the private citizens and the local authorities. I think that’s really the crux that’s really important for us. So the way we have gained traction in Belgium is always via a single citizen citizen who bought the device and made a lot of noise about his device and his data for the local authority, which then initiated local authority to set up a project. And then we can help to local authority or a commercial client consultancies working with a local authority to set up a network of of devices. And that’s really crucial because we like I said, halfway, I think we are almost subsidising the individual ones we want to make make it really cheap for the enthusiasts for the cars, fruits initiatives. We want to keep them but our sustainability is dependent on projects. So networks of, of devices for it’s for these commercial clients, consultancies and local authorities. So we really hope to go in, find that we’re looking for partnerships for example, large organisations who have have access to a big community who would buy these devices from us, we provide them a management platform and management tool to, to basically manage the fleet of devices in the community. But give them device for free towards volunteers. So that’s a bit the model that we were hoping for to achieve.
Carlton Reid 40:19
Took me through your dashboards, I know, tech people get really excited by their dashboards. So when you instal this, or when you when you when you when you upload the data when you’re not you, but when the device uploads the data to the cloud, and then it’s all graphed and mapped, and what have you, what do you see? So can you see not just numbers, this is the number of people who’ve cycled walked, driven past your window, we also get like peaks. So there’s a graph there showing you at 5pm you’ve got a massive amount of car traffic, bum, bum, um. So what what is your what are your stats show?
Kris Vanherle 40:57
Good question. So I invite you and your listeners to go to our website, because basically those things are all open. Also, there’s a private dashboard, but there’s also a very big public website. So I’ll just go over what’s what’s there. So we report the total numbers by day and by hour. So just to sum up of the cars pedestrians and, and the four categories. But like you mentioned, there’s also an interface where you can select a time interval, for example, three months, four months or four year, and then it calculates a typical traffic volume for this period, a typical traffic volume for a weekday or a weekend. And that, like you said, that will demonstrate these these typical patterns. So the short morning peak and a bit more prolonged evening peak. So those are interfaces, what’s also reporting on speeds and so we report the speeds of passing cars in bins of 10 kilometre per hour, we’ve also changed it to miles per hour for UK. So that’s also possible. So that will give you an idea on what what they share, or what are the speed profiles, let’s say of cars in your street. And finally, the VAT five. And speed is like a traffic engineering concept, which I have to explain. So if you would have 100 cars, and you would sort them by fastest to the slowest. It’s the 85th fastest car. So there’s 15 cars driving faster and 84 driving slower. So the V 85 gives you a good indication of what we call free flow speed. So how fast you guys drive when there’s no congestion. So on a disagree indication of how our streets are typically used by cars in terms of speed. So those are all in the open dashboard and in the private dashboard. So the telecom user has got a bit more extra analytics, for example, a monthly report with big peak hours last month has to have to increase last month to the to the to the previous month. So a bit more extra gimmicks, which are useful for the individual use of it, the core thing is that everything is open. So that also if you’re not at a telecom device owner, you can still interact with the data, you can do analysis yourself if if you want and we encourage that.
Carlton Reid 43:05
So traffic census as has been carried out since the dawn of motoring, you know I’ve I’ve, I’ve looked at various statistics down down the years of, you know, the usage of roads, and even of 1920s and 1930s, bike paths, etc. So censuses have been carried out since the dot. But they’ve always been this, you know, they pick a date, they pick up probably a 24 hour period. And that can be very, very misleading because it you know, do you do it in the winter? Do you do it in the spring, do it in the summer, or you tend not to do it, you know, loads of times, because it’s just so expensive, because it’s manual. Counting is phenomenally expensive and complicated to do, what you’ve got is a system that is 20 477 days a week, every single week of the year, blah, blah, blah, it’s just constant. So you’ve got much, much more robust data there that will build up. And I know, like, folks is the the Bicycle Advocacy Group for Edinburgh, has been doing traffic counts and traffic sensors for a long time. And they do it manually. This is going to I’m sure would excite them because it will be an adjunct to the database built up over many years. But instead of just doing it on a certain day, they’ve now got it every single day.
Kris Vanherle 44:31
Yeah, that’s correct. Yeah. What’s important is to know that because people think of data accounting data and data candidates in traffic counting today, but that’s not the case. So you’ve got different types of traffic counting. So for starters, you’ve got this floating Carta, which I explained from the cell phone site, but also the point location so and maybe supplies do what you would expect, would expect but all All these different types of traffic entered into on a single site are valuable. So even manual counts are still valuable. Although they are fragmented to show they’re very scarce. So either it’s expensive, or it’s or it’s very one off. But it’s still valuable because in some use cases, you really need it. So for example, if you want to optimise traffic lights, to want to know which site needs to have the green light alarms, and so on. For that type of thing, you need queue lengths, for example, to monitor queue lengths, is usually difficult, even with high end optical systems. So for those type of applications, you still need manual counts. But that being said, for two traffic counts for for for a long period. What well, we used to work with, like you said, every Tuesday of every Tuesday, every two or every two weeks, do a manual count, that’s gone, those times are gone. And either weapon or metric tubes or in the future with Telecom, that just the sheer volume of data will will will blow away even the bit more accurate manual counting data, which is much more fragmented. So yeah, it’s indeed, it’s indeed, the potential of a big change in mental mental change of evolution in terms of how we collect difficulty data on specific sites.
Carlton Reid 46:20
And @fietsprofessor, the on Twitter, he had a very, I think he did a an academic study on this, but he also did it on Twitter, and he kind of like, showed the video of this. So he had a camera above a promise intersection in Amsterdam. And then he talked about swarms, and flogging, and he showed how, you know, bicyclists just flock through this particular intersection, whereas cars are very regimented. You know, in narrow trails, bicyclists were going everywhere. So what that suggests is, as we know, there are an awful lot of bicyclists in Amsterdam. So does your system cope with high use bicycling city, where if there’s 10,000 cyclists going through an intersection, it can count that even though they’re flocking together? How do you count the ants?
Kris Vanherle 47:20
Yeah, that’s a very good question. And it’s very important to realise how you should tackle those issues. So in short Telecom, the sensor is low cost device, it’s not made for counts on intersections, because like you said, You’ve got you got flux, you’ve got everything, which is going, one going left going, right. I mean, if you do anything with it’s just impossible, I mean, it’s not impossible. But if you want to do those counts, right, you need the high end camera system, which will cost you five to 10,000 euros just for a single site. So we’ve deliberately chosen for, let’s not, let’s not try to fix that problem, because it’s just too complex. But you can fix the problem. If you focus on a bit more simpler sites, which are just seat segments where you have bikes passing by left, right, left, right. So it’s very clear that object is passing to one side and exiting the other side. And if you’ve got a bit of a downward angle, you can just basically see all object passing by. And that’s a face a much more simpler setup where AI systems can much easier more easily cope with. So if you would have such a situation, what we would do is not monitored the intersection, because it’s too difficult, but just take the branches, so maybe even a few sights on the branches. So you know, the absolute volumes on on, on the branches. As long as you don’t have like, I would say, a big cycling haze or something where you have like 567 people, not side to side, and maybe even swapping positions. I mean, that’s a nightmare. But just one or two side by side, as long as they’re clearly segmented. So like different objects, and it will go it will count those devices. Again, I just want to point back out about some technology choices that we’ve made. Our core focus is it has to be simple, simple to instal and, and cheap so it’s affordable. I mean, anything that you describe now it’s all solvable with technology that will just cost you a tremendous amount of money so you can indeed make a camera system and we should to an online image conditioning system and then will you have to pay for every object that you I mean, it’s all possible but will just cost you a lot of money.
Carlton Reid 49:47
Yes, now I think that’s what @fietsprofessor did. They basically tracked it and then they used AI to count every individual and in that and it was it was a highly complex and probably expensive thing to do. So, Kris, thank you ever so much for taking the time today to talk about this system. So give us the where can people buy these from, your social media stuff get? Give us all your information of how we can find you.
Kris Vanherle 50:14
Yeah. So first of all, first up our website. So www.telraam.net, I think if you go to the website, things will become very apparent, you can also engage with the data. There’s much more two links on what the device is like. And there’s a webshop. So you can if you want to devise yourself, I mean, that’s the way to go. If you want to push for that, let’s say you do this with my local authority. I’ve got to reach out to us via the website, and we’ll see if we’ll see if of course it is possible. So that’s our website. I think that’s the first place to go. We’re active on Twitter. Our handle is telraamtelraam two times next to each other. So that’s the probably the social media that we’re using the most. We’re not on Facebook, not on Instagram, and LinkedIn obviously, if you want also telraam. So that yeah, that’s how to find anybody interested, reach out to us via our web form. We are ready to go with our new sensor.
Carlton Reid 51:16
Thanks to Kris Vanherle there and thanks to you for listening to episode 324 of the Spokesmen podcast. Show notes and more can be found at the-spokesmen.com. I’m still waiting to with BBC journalists Anna Holligan and Kate Vandy who have yet to take delivery of a rather special outside broadcast unit: a tricked-out cargo bike. This will be the bike bureau, a mobile news gathering studio like no other. I’m hoping that chat can happen some time next month, meanwhile get out there and ride.