Co-founder and CEO Dave Grannan, 57, spoke with Staff Reporter Pete Bigelow about the potential of the technology on the May 3 episode of "Shift: A Podcast About Mobility." The full interview can be heard at autonews.com/shiftpodcast. Here are edited excerpts.
Q: Tell me a little bit about Light.
A: We started in 2013, based on the ideas and inventions of my co-founder, Rajiv Laroia, who had a very novel invention around computational imaging and specifically something called multi- aperture. So, using many cameras simultaneously and putting that information together with the ability to create very high-value images.
What makes your depth-imaging system unique?
Our system works on the same basis as human eyes. We like to say "inspired by nature and powered by technology." As a human you use a process called parallax to see depth in the world. You have two eyes. You're always looking at an object from slightly different perspectives at the same time, and you're able to sense depth. We do the same thing with cameras.
This part is not new. But what is difficult and what's really challenging are two things: One, calibration, particularly in something like autonomous or assisted driving. Cars are very rugged environments. So how do you keep multiple cameras calibrated 30 times a second, frame by frame, in real time and harsh conditions? That was a problem that hadn't been solved and was one of our key breakthroughs. The other one is we're not just limited to two cameras. We can use three or four in a given array or direction. And no one had figured out the algorithms to solve a problem like that before Light. So those were really two of the essential breakthroughs that allowed us to do this.
At what range can it perceive depth and detect objects?
We can measure depth or range distances anywhere from a few centimeters, out to over a kilometer from the vehicle.
It's a tremendous range that is really useful in a lot of important cases, whether it's Class 8, Level 4 automated trucking, adaptive route planning, the ability to see half a mile down the highway that a traffic jam's building and reroute the vehicle, for example.
A kilometer sounds incredible, especially when sensing range today is typically 250 or 300 meters at best.
Yeah, at least 3x better than anything that would be, say, from lidar. Just because you get a lidar return off of something 250 meters away doesn't mean you can actually identify the object or know what you're looking at.
Typically, we find that lidars are really effective not much beyond about 150 meters in best cases. But I know some of them advertise a lot longer range.
Does Clarity replace lidar?
We think lidar's an important sensing technology and certainly has use cases. The key advantages of lidar are it's very fast, it's very accurate. But it's got challenges, of course, like a lot of things.
It's very expensive. It sees very limited range. It uses more power because you're emitting a laser, and there's still some safety and interference issues to be worked out with lidar.
That said, we are partnering with lidar providers for Level 5 robotaxi-type solutions where the provider wants the redundancy, the fault tolerance that you get from additional sensors and they can afford it.
But when it gets to some of the passenger-vehicle things that we talked about earlier, especially Level 2, Level 3 kinds of solutions, I think lidar's very unnecessary and overly expensive compared to what Clarity provides.