80% car parking timesaving? That’s intelligent CAM decision making
In this Cars of the Future exclusive, the co-founders of Hertfordshire-based Eloy, Anna Corp and Damian Horton, explain how their connected car services make parking, driving and self-driving safer and more efficient.
Congratulations on your Zenzic CAM Scale-Up success. How did you get into connected vehicle tech?
DH: “Our story goes back to 2004, when I was a maths undergrad at Oxford. I ended up doing a thesis on bifurcation theory in swarms, based on simulations like those seen in computer games – little army men running together, and how they interact with each other. It involved a lot of traffic modelling, but at the time there were no jobs in driverless cars, so I went into investment banking and started a couple of businesses. When I moved back to the UK from Australia, in 2018, I saw that the connected and autonomous (CAM) vehicle market was finally happening. I met Anna at a start-up in London and together with an old school friend of mine, Marcus Robbins, who’d done a lot of geospatial work, we decided to give it a crack.”
AC: “My background is in marketing, user experience and customer insights. One issue I’m very familiar with is companies not thinking about real-life human problems. At Eloy, we’re all about solving problems for all road users, not just car and lorry drivers – cyclists, pedestrians, horse riders, everybody. We take a much more holistic approach to making roads safer and more efficient. There’s a big push to move people on to active travel and public transport, but is that what people really want? Shared robotaxis are often presented as a utopia, but why would a mum use one when she has her own car with all the baby stuff already in the back? It’s hard to force behaviour change. A better way is to give people options and tools which they see value in, which make their lives better. Then they’ll adopt.”
Which brings us to your app…
DH: “We worked out early on that the best way to get into the connected car space was to provide a sat nav, before building in any new experiences to make roads flow better. On Boxing Day 2020, we got the email from Apple saying the sat nav component had been accepted for CarPlay.”
AC: “We joke that we’re the smallest sat nav company in the world, but it’s a prerequisite for all we plan to do. We had to get into existing vehicles.”
DH: “We’re obsessed with the situations you get into as a driver – sitting waiting to make a turn across a blocked carriageway, queuing at a mini-roundabout while everyone waits for each other. How can we make these small things better? The missing piece over and over again was multi-vehicle coordination (MVC). So we got super focused on niche use cases, like getting in and out of car parks and passing on country lanes.”
AC: “Smartphones are a good example of a product which has morphed into so much more. The best thing is we can offer connected vehicle solutions now, to provide good advice for human drivers, to prove high efficacy, and then apply them to higher level automated driving.”
And you’re already testing at UTAC’s Millbrook Proving Ground…
DH: “Yes. In October, we demonstrated our narrow road warning solution, which reduces the need for reversing to find a passing point, at The Transport Technology Forum at UTAC. That involved just two vehicles. The next phase is to get it working with 20 vehicles in a controlled environment, then up to 100, and scale from there. We’re looking for the right partners, ranging from ports and farms to construction traffic, freight and public transport – probably fleets initially.
“Early simulations indicate a 20% timesaving from MVC for country lane passing, and up to 80% for car park entry and exit. Internet connectivity is an issue (that’s for someone else to solve!), but we can deploy on sections of road where a good signal is almost guaranteed. Then it’s a question of making sure the intervention – the beeping or flashing or messaging – doesn’t outweigh the benefit. The big question is always: does it improve safety?”
Sorry, did you just say an 80% timesaving for car parking?
DH: “Yes, by using very similar modelling to how to fill an empty aeroplane. For years, it was a free for all, so the airlines tried to get organised by filling in order from row one. Mathematically, that turned out to be the slowest way, because everyone has to wait for the person in front of them. So they got clever and started filling from row 30 and working backwards. That’s actually the second slowest way, because you end up with the same problem of everyone waiting. Eventually they worked out that the best method is a structured filling pattern. You send in rows 30, 25 and 20, then rows 15, 10 and 5. They all have space to stow their luggage and what you have is a lot more manoeuvres per second. That gives you an 80% reduction in filling time.
“We looked at high density car parking in the same way. Think Silverstone on grand prix weekend, when there’s traffic chaos. If every car has an allotted parking bay and follows guidance from a sat nav, you can apply those same principles of more simultaneous manoeuvres. There are potentially further gains too, for example, by connecting live data to the local traffic lights to disperse the traffic more efficiently. The challenge is coordination, between the event organisers, the local authority, the car park operator and the attendees. The rationale is economic benefit, reduced journey times for everyone, which brings you to infrastructure investment decisions – the cost per mile benefits of these intelligent systems compared to building more roads.”
AC: “Once people see that the system works, they’ll quickly learn to trust it. I see huge opportunities in business parking for big employers. If they could save each employee 10 minutes a day, think of the extra productivity. Over a year, suddenly the business would have gained a lot of time and a lot of money.”
And you’re using artificial intelligence to optimise this?
DH: “Yes. Using SUMO simulation software, we’ve created full digital twins for car parks and certain road segments. Then we add a reward function. The AI basically tries to get the most points, a bit like the 1980’s computer game, Frogger. It’s a type of reinforcement learning that tells cars what to do in different circumstances. We’re training it for a road layout at Millbrook at the moment.
“The holy grail is getting 100% of cars using the software, transmitting and receiving your information and following the instructions. In the meantime, there’s questions around gaps in the data – how much knowledge you can you infer from modelling. Then there’s the dynamics of network effects. An interesting one, going back to our car park efficiency, is what happens if someone decides to break the rules, perhaps by stealing someone else’s slot. You can probably use a financial incentive to overcome that.”
AC: “One reason we really like the cars on a country lane solution is because it’s self-reinforcing. It’s a win-win without needing to use a monetary incentive. Both drivers benefit from additional information and overall traffic flow improves as a result.”
Sounds good to me. For further info, visit the Eloy website
