Luminar’s game-changing LiDAR makes its way to TRI’s self-driving car

When I ask around about who has the most potential to make big waves in the LiDAR industry, one name tops the list: Luminar, the startup that emerged from stealth earlier this year after five years developing its unique LiDAR architecture from the ground up. Now, Luminar is revealing the first of its four current major partners: Toyota Research Institute (TRI), the research and development organization created by the global automaker to focus on robotics, autonomous vehicles and AI breakthroughs.

TRI is using Luminar’s LiDAR in its new Platform 2.1 autonomous test vehicle, having selected the solution because of its unmatched capabilities in terms of being able to see at a distance, and to also perceive objects that traditionally have very low reflectivity for laser light, which makes them a lot harder to pick up via LiDAR. These include common objects on the road including tires, and anything else with a dark, matte surface.

“When developing this product, we remained in stealth for five years, ensuring that by the time we came out we were already working with some of the best programs out there,” explained Luminar CEO Austin Russell in an interview. “We now have a product that we had the utmost confidence in – something that isn’t just an idea on a post-it note with some questionable physics, but actually a real-world system that we believe can truly enable these autonomous vehicle systems to finally have that better-than-human level perception.”

Luminar had to got back to basics to develop a LiDAR that fit this description, and owns much of its stack, rather than sourcing anything from off-the-shelf suppliers. It has a facility in Orlando, Florida where it has managed to recruit some of the world’s leading optical engineering talent (Orlando has a lot of that talent thanks to the aerospace industry presence there, Russell told me) and it has even acquired some of its suppliers along the way to vertically integrate all the pieces that come together to help create its LiDAR.

“[Luminar] gives a company such as Toyota or TRI such a huge head start and advantage with this problem, because of the opportunity to leave everybody in the dust with this fundamentally new level and quality of data to be able to work off of,” Russell said. “The way the that the software is architected and the autonomous vehicle is built out really changes when you have this new level and quality of data. Ultimately, everyone’s going to have to adopt this type of platform if you want to be able to succeed in the long run.”

[gallery ids="1547677,1547676,1547675"]

Russell says that there’s opportunity for TRI and Toyota to “leapfrog the rest of the industry” when it comes to the pace of its autonomous vehicle technology because of the sensing advantages that its product provides vs. other industry leading alternatives, including stalwarts like Velodyne, whose large multi-laser arrays adorn most current-generation test cars you’ll see on roads.

Lumina’s solution also uses only a single laser to achieve its class-leading results, which is remarkable in term of also offering cost and production scale benefits. Russell says that even so, it didn’t approach the problem from the angle of optimizing for cost efficiency at the sake of performance, as some competitors have done.

In the end, Russell is still very realistic about timeframes for seeing full autonomy on the road – he notes that disengagement rates for even the best performing self-driving test cars in use today are at about one in every 1,000 miles – which compares to accident rates of about one in every 1 million miles for human drivers. That’s a large delta to make up, but Russell and Luminar think crossing the virtual Rubicon in terms of exceeding human perceptive abilities will be a big part of narrowing that gap.