Luminar’s lidar is on display in a Volvo concept car, and the point is to show how the laser-based sensor dovetails with Volvo's system for reading pedestrians' body language.
At the impressive distance of 250 meters, Luminar’s lidar can provide enough detail to let Volvo do what the carmaker calls “pose estimation.” This means getting an actual representation of a person, “including individual limbs, such as arms and legs—a level of detail not previously possible with this type of sensor,” Volvo said, in a release. It’s the sort of thing you’d need to know to confirm that an object is human, and then to guess which way he or she is likely to move.
Here’s a little video illustrating the trick:
The car, called the 360c concept, is still just a designer’s notion of how Level 4 autonomy would work under highway conditions. Level 4 means the car can do everything under certain constraints, typically having to do with the place, the time, or the weather.
In Volvo’s view of things, now being shown at the Los Angeles Auto Show, you’d have to get on and off the highway yourself (that’s the one constraint). Then you’d just flip a switch and zone out until you reach your exit. Or flake out, on a fully reclining bed, like so:
Sure, there are a lot of other cool simulations out there in Lidarland, but this one is backed by a major automaker. The Volvo partnership was unveiled earlier this year; a previous one, with Toyota Research Institute, had come about seven months before. More are being signed all the time.
“We have 18 different companies working with us and partnered with us,” says Luminar founder Austin Russell. “A lot of those are recent, very recent; unfortunately most of them are not public.”
Truck companies are particularly interested in long-range lidar because they stand to benefit from the stopping time that it makes possible. Luminar’s 250 meters of range provides 7-plus seconds notice at highway speeds, enough to slow even a big rig to a stop.
Luminar is notable for basing its lidar on moving parts rather than solid-state methods, and for choosing a laser wavelength that's safe for the eyes and can thus be projected at high intensity. It uses just one widely sweeping laser and receiver‚ unlike the 128 laser/receiver pairs in the top-of-the-line product from Velodyne, the company that started automotive lidar and still dominates the market. Velodyne’s lasers move visibly, rotating on the roof of the car, whereas most of the 60-odd lidar startups have preferred solid-state solutions. The idea is that nonmoving devices should be cheaper and more resistant to degradation under road conditions.
The number of cars equipped with Luminar sensors hasn’t been revealed, but the total would appear to be rather small at this point. The order of the day is to work out the performance characteristics.
“You need so many points on an object, you need a certain range for braking time, you need a certain degree of precision, to see through rain, fog, and snow, and you can’t be blinding other sensors,” Russell says. “A year ago, if you’d asked 10 OEMs [original equipment makers] for specs, you’d have gotten 10 different answers, but in the past year, and even the past six months, a lot of them have converged on what’s needed. The [differences in the] companies’ [demands] are plus or minus 10 or maybe 20 percent, but it’s nothing like an order of magnitude of difference.”
The next order will be to learn how to manufacture Luminar’s sets to automotive specifications. That could mean working to perfection after 200,000 kilometers of hard driving, and in all kinds of weather. It’s a hard problem, but it’s the kind of hard problem the auto industry is used to solving.
“We are in discussions concerning series production of real vehicles for 2021 to 2022,” he says. “So we have a couple of years to get this stuff together for that.”