As you’ve probably noticed, the Telraam website looks a bit different. We’ve reshuffled and added some pages, and updated content. We hope the updated website adds greater clarity on what we’re doing and what our offer is to both citizens and professionals.
Without a doubt, the biggest change is the expansion of the Data subscription. We can finally show the additional benefits of the S2 device compared to the V1, counting more modes of traffic. So now we have not only the two-wheeler totals, but the actual distinction between bikes and motorcycles, or the split between trucks and buses within the category of large vehicles. There are many more benefits associated with the data subscription, such as 15 minute data, analytics and export functionalities, so you’re getting your money’s worth for €15/month. Have a look at the webpage or check the personal dashboard for your device to find out more!
With the launch of the Telraam S2, we moved to a device + subscription model. The Data subscription (€15/month) gives you access to all the data the S2 device generates. Organisations, whether it’s local transport authorities, consultants, campaign groups, universities, etc., managing a fleet of devices, can use the Network subscription (€25/month) with additional features to facilitate their project.
However, because we’re still citizen scientists in our hearts and we want to empower everyone to collect traffic data, we’ve kept the basic data open and free for personal, non-commercial use. In all transparency, it’s a delicate balance maintaining an open platform while at the same time ensuring that we have the necessary sustainable revenues to cover the costs associated with the collection, storage, analysis, and distribution of data, while sustaining a team to keep developing the Telraam sensor and platform. In this way, individuals and organisations, via the Data or the Network subscription, are indirectly subsidising personal use of Telraam and are supporting the Telraam ecosystem. We will proudly stick to that principle.
Evaluating half a year of S2 devices in the field
We’re now about half a year since the launch of our new little S2 sensor and we’ve already learned a lot. While we tested extensively prior to the launch, there’s no better test than an actual launch. The proof of the pudding is in the eating! After 6 months, we can conclude that the pudding is yummy… but we think we can still improve the recipe.
There are now over 500 S2 devices active in the field, so that’s a lot of data coming in. The good news is, most of the time, the device “just works” as designed. Installation is easy, and once installed, it most often just keeps counting. Good. Having said that, we’ve had our share of issues. As with any launch, some devices had hardware defects; we had to replace some devices. We’re also learning the hard way that the logistics of shipping boxes isn’t that simple. Luckily, we now have a good partner to help us with shipping the devices to you and we’re getting better (thank you for being patient!).
The most annoying (and unexpected) challenge we’ve seen with the S2 is the connectivity issues. We expected our sensor to have either LTE-M or NB-IoT coverage everywhere, but that doesn’t seem to be the case. Unfortunately, this is out of our control as operators have not yet fully deployed LTE-M/NB-IoT infrastructure everywhere. (Even when their websites claim they have! Sigh…) Luckily, the situation is improving and the problem should be temporary as telco operators gradually increase deployment of the necessary infrastructure. Again, in most locations it’s working fine, but if you’re in a remote area, or if you’re in the Western part of the UK, then our advice is to check first with us to see if there’s coverage on your specific location…
We’ve also had anecdotal feedback from users on the data accuracy. It is very challenging to do a systematic review of this, but overall the feedback from our pioneer S2 users on accuracy is within our expectations. Accuracy for cars and trucks counts is solid, confirming the 90-95% accuracy seen in our own tests. For pedestrians, the feedback we’ve got is different. Pedestrian counts are still highly challenging, especially in busy areas, leading to an undercount of up to 40-50%. There also seems to be a tendency of pedestrians/bike misclassification, though far less than what we saw with v1. There could be a bias in the validation data being sent to us, only reporting when performance is lower than expected. However, data accuracy is something we want to follow-up on closely so we know what to focus on when retraining the model. So don’t be shy and share your manual counting results with us! If you can send us parallel video recordings and an indication of the ROI setting of your device at the time of the validation that makes your feedback even more useful.
So, what’s next?
In the past 6 months we’ve focused on getting the S2 up and running, and building the data dashboard. Now it’s time for the next steps!
When we launched the S2, our initial goal was to have a detection script that was also able to count traffic at night. That didn’t work out, or more precisely, it worked in some locations but with insufficient accuracy, so we’re now actively filtering the incoming insufficiently accurate counts out. We had pledged to be ready with a new firmware to count reliably at night by autumn - when sunlight is getting scarce. That’s basically next month.
Sadly, we have to report that we’re not ready. We’re trying, but the challenge is proving to be greater than anticipated - at least for now… We still think night counts are technically feasible, but it’s more difficult to get reliable results and it’s going to take more time. We’ll continue our work and aim to gradually release an updated detection script which will be able to count all motorised traffic at night in a single class, without the speed, within a few months.
But we’re not finishing this update with doom & gloom! There’s some exciting news that we can now finally share with you: we’re developing an autonomous outdoor version of the Telraam S2! We’ve partnered up with Verhaert Digital Innovation (formally Bagaar) to develop a variant with a solar panel, a battery, and a safe enclosure, allowing power-independent installations outside, which will not require a window anymore. We’ll be developing and testing this in the framework of 2 pilot projects at the end of 2023 and early 2024, so if all goes well, this variant should be available for anyone early next year.