Ahead of his session at Sensors & IoT Virtual World Week, where he’ll be discussing Sensing Modalities & Actionable Data for Automated Driving, Pierre Olivier outlines LiDAR and perception’s role in self-driving cars.
Despite enthusiastic predictions five or so years ago, we’re still a long way from roads filled with self-driving cars. However, thanks to advances at companies like LeddarTech, we’re certainly getting much closer to this vision becoming a reality this decade.
According to Pierre Olivier, CTO at LeddarTech, there has been a shift in the last few years, which is accelerating advances in autonomous vehicles.
“If you go back two or three years ago, most OEMs and tier ones were adamant that they wanted to build the perception and the driving stack themselves. What’s changed is they’ve realised that it’s much more complex than they had originally envisioned. Now there’s an openness to see what the different suppliers can bring to the table to help solve the problem,” he explains.
“They have cars on the road and they still see that our processing can give greater benefit to them.”
This is welcome news for the automotive industry. By relying on specialists like LeddarTech, developments can happen more quickly than if the tier ones, the autonomous vehicle technology vendors, were trying to build their own systems from scratch to sell to car manufacturers. This is thanks in part to the realisation that self-driving vehicles were not ready for the mainstream as originally predicted.
“If you go back five years, the general mood in the industry was, Level Four is something that’s going to happen very quickly, and there’s going to be Level Four cars on the road in 2018 or 2019. That didn’t happen. And now, everybody thinks it’s not going to be until at least 2025 to2030 and so it becomes, how can we accelerate this and offer something that the customers want that’s not quite so complex to deliver. It’s really that realisation from the OEMs – that they needed help – that enabled us to bring solutions to the market,” Olivier says.
So how far off are we from reaching the ultimate goal of Level Five technology, [see here for a detailed explanation of the different levels], where autonomous driving is as good as or better than humans? Quite a way, according to Olivier, and a large part of the issue is how do we compare the two? He notes that the human driving record is surprisingly good: the total number of road fatalities may be a large number, but per mile driven, it’s actually very low.
“So far, autonomous vehicles only drive in optimal conditions. If the weather or road conditions are not good, they will stop. In contrast, we as humans, are expected/choose to muster on. So, there is no objective comparative measure. Arguably, if you look at what’s out there today, there’s enough to show that there is room for improvement. It also comes down to how diligent manufacturers and regulators are in allowing vehicles on the road,” he points out.
“Until somebody says: “Here’s how we’re going to qualify an automated vehicle, here’s the equivalent of its driver licence and now we approve it to be on the road”, there’s always going to be that level of uncertainty. Is it fully tested or is it beta software that I’m supposed to babysit to make sure that it doesn’t go awry?”
The Covid pandemic has also precipitated a downward shift in focus for the autonomous vehicle sector, and there is now more interest in Level 2+, which Olivier dubs ‘ADAS that works’.
“People were seeing even Level Two as something that was a short stop on the way to full autonomy. But now people are starting to realise that there’s a true business model for good quality ADAS. There’s been a renewed interest there.”
“The other area where there’s been more interest is unmanned delivery vehicles such as for food delivery. The focus being on removing people from the equation. People now see the real benefit in being able to have an unmanned vehicle coming to your doorstep, because you’re maintaining social distancing.”
LeddarTech’s core offering is based on LiDAR (light detection and ranging) technology, which uses light waves for remote sensing. LIDAR works by calculating how long it takes for the light from a laser to hit an object or surface and reflect back to the scanner. When used in an autonomous vehicle, LIDAR builds a 3D model of the environment around it in real time, so it can avoid obstacles and drive safely.
LeddarTech is not only a LiDAR company – it offers products across a range of areas including radar, camera and Advanced Driver Assistance Systems (ADAS) technology for industries such as mining, military, automotive and industrial. However, LiDAR remains a key focus, and where other companies have focused on the optical side for their LiDAR developments, LeddarTech has always focused on signal processing and flash LiDAR.
“It was never a question of having a mechanical widget and trying to figure how could we build a flash LIDAR. We never built anything other than flash LiDAR. It was always about improving the performance to meet automotive requirements. The way we acquire and process the signal is different from everybody else. And that’s really the DNA of the company,” Olivier notes.
LeddarTech has recently acquired two compagnies to further boost its offerings: Phantom Intelligence, for its innovative signal processing and existing customer relationships; and VayaVision, for its perception development. The technology from both firms is already seeing its way into core LeddarTech products. However, Olivier was keen to stress that R&D investment in its own technology is also crucial to LeddarTech’s growth strategy.
“We’re not set up to deliver finished products, we’re set up to deliver technology. So obviously our ability to sell that technology relies on being ahead of our customers and competitors. If you look at LiDAR system-on-chip (SoC), this is something people are currently designing, whereas we started our designs more than five years ago. That’s why our investments into our core technologies is so critical to us,” he says.
“That’s where you can also see the logic behind some of these acquisitions. Because if you can acquire technology that’s already been validated, and if they bring new products that are complimentary and we can integrate them, it ensures that we can maintain that advantage. But we’re not a conglomerate, we’re a technology company.”
Looking ahead to 2021, LeddarTech will be continuing the development of its perception stack, its SoCs and its signal processing.
“It’s a very interesting time for the industry. We’re shifting from what was a very idealistic stance a couple of years ago to one where both the vendors and the customers are now a bit more pragmatic in looking for solutions. Certainly it fits well with our approach, and it bodes well for seeing real solutions on the market shortly.”
To hear from Pierre as well as leading and innovative brands across the Automotive, Mobility and Transport sector, you can register for the Sensors & IoT Virtual World Week here.