The AI 'Co-Pilot' bike light that makes cycling safer

Over 10,000 cyclists are injured, and more than a hundred are killed, each year on Britain’s roads. Often it’s because approaching vehicles don’t see them, and the cyclists only find out too late that they need to take evasive action. So can AI help? A device called Copilot - which is an AI-powered bike light that uses smart sensors to constantly watch the roads and the movement of vehicles - forewarns road users and cyclists of potentially dangerous situations with audio cues. The founder and CEO of velo.ai, Clark Haynes, explains how it works…

Clark - It's bringing a lot of the advanced tech that you find in things like the modern automobile and really bringing it to bear for safety. For the cyclist right now, it's about the size, a little bit larger than a deck of cards. It's small enough that it tucks just right under your seat. It's rearward looking, so it's watching out for vehicles behind you and it's battery powered, so it'll last for about five hours off of its internal battery. So you can just go for a ride and you're all set.

Chris - That's still pretty small to have the power of an artificial intelligence system scanning a road for you. What's it doing? Is it using cameras then to watch what's coming up behind?

Clark - Exactly. There are existing competitive products out there that use radar sensors and we just looked for cameras as a better way of understanding the world and you can do so much more with a better sensor. So the device itself is primarily composed of a camera that's watching the world, a set of lights that can react to things in the world, like other drivers, audio that can be used to cue the bicyclist and then onboard compute - a Raspberry Pi, made right in the UK paired with a special AI chip. And that's what really allows us to do all of this computer vision image processing in real time on a really, really low power budget so that we can run off of batteries.

Chris - What does the AI do with the camera pictures and then what's the output from the device to warn the cyclist?

Clark - It's watching the world, it's understanding the world from the perspective of the bicyclist and it's able to essentially detect approaching vehicles and understand what they're doing and what their behaviour is. I've actually been spending my career strapping cameras to robots and autonomous vehicles, so a lot of the opportunity here was to bring some of that intelligence to the world of cycling.

Chris - So if you're in danger of being rear-ended or what they dub in the industry, a Smid Sea, sorry mate, I didn't see you. Then. This should be able to anticipate those sorts of things that, that claim the lives of or cause injuries to cyclists and then warn them.

Clark - Exactly. That's our goal and it's a warning to the cyclist, but then it's also a warning to the driver. The design is that a few seconds before that you're going after something more like a startle reflex because we can have reactive light patterns that change and try to get the attention of a potentially distracted driver.

Chris - Have you road tested it till, you know, if it does give people enough warning to make a reasonable difference that will avoid an accident or improve the safety for cyclists?

Clark - Yeah, absolutely. We've been road testing for almost a year and a half with earlier beta versions and now with our production version out with the public, and right now these are all over the US but also a natural focus in our home city of Pittsburgh, Pennsylvania, which is a challenging city with lots of difficult roads full of potholes, hills, weather, you know, bad drivers too. All of those things. The feedback we get from our users is that, yes, drivers are giving me more space and I feel safer and more confident biking on these roads because I know what's going on behind me and that's really the goal of what we're doing. We actually just built a rig that's going to allow us to really accurately measure some of these outcomes and we intend to publish the results of that in the coming months.

Chris - How do you see the approach and the technology going forward, could this be deployed in different ways, different vehicles, different contexts, or are you going to develop this just for the cycle market more, and if so, what will you add?

Clark - Yeah, that's a great question. The biggest thing for us with cycling is that you have such a passionate set of users and a community that's focused on safety because you know, every cyclist is riding on the roads and they're wishing, Hey, I just want to come home safe. And so I think it's a great place to start. That being said, you know, we find all sorts of mobility are taking hold in our cities, particularly this kind of personal micro mobility, so this is bikes, but this is also scooters and one wheels and all of these new contraptions. We see this really coming to bear across all forms of personal micro mobility.