Autobayng News Team
- The Rivian R2 is adding lidar, a surprising addition to a car that’s set to start around $45,000.
- That will enable far more advanced automated driving assistance features, with the eventual goal of full autonomy.
- But others, most notably Tesla, say that lidar isn’t necessary. This video shows why it might be.
The case for camera-only autonomous driving is that humans rely on just their eyes and brains, so the right cameras and AI should be enough. But if truly driverless cars are supposed to be far safer than humans, shouldn’t they use sensors that go beyond what humans can perceive?
The broad autonomous vehicle industry consensus is yes. And this week, Rivian joined their ranks by offering lidar on the upcoming Rivian R2—a remarkable add-on for a vehicle that’s tipped to start at just $45,000.
Famously, the Tesla approach to autonomy is to just use cameras and AI. But Rivian today demonstrated why that may not be enough.
For much of its existence, lidar has been so expensive and specialized that it’s largely limited to dedicated robotaxis. It’s only recently started being installed on some higher-end cars, like the Volvo EX90 and future Cadillac Escalade IQ. But lidar is cheaper now.
“About 10 years ago, lidar (units) cost in the tens of thousands of dollars,” said Vidya Rajagopalan, Rivian’s Senior Vice President, Electrical Hardware at the automaker’s Autonomy and AI Day event. “Today, you can get a good lidar for several hundred dollars.” Moreover, their resolution has only gotten better while they’ve also shrunk in size. This is also why lidar units have become increasingly pervasive on cars in China, including ultra-affordable models.
Lidar today vs. Lidar in 2016
Photo by: Rivian
But if you’re still not convinced that lidar is a no-brainer for autonomous vehicles, check out this other video Rivian released today. It shows the very stark sensory differences between a camera-based system, a combination of cameras and radar, and then both of those together with lidar added in.
In the first example, all three sensor suites perform about the same. The cameras, lidar and radar all identify the same moving cars, people and stationary objects. But when things get foggy, the differences become much clearer.
The camera-based system is limited on what it can see, just like your eyes would be. The cameras “miss” a lot of elements obscured by the fog, most notably a pedestrian crossing the street. The camera and radar combo sees a little further. But when you add in lidar, the system sees it all—again, better than a human can. Rivian Lidar Radar Demo Photo by: Rivian
In another example, lidar is able to detect some people walking along a barely lit highway at night. The camera-only system doesn’t even pick them up until they’re nearly in front of he car. It’s easy to understand why this presents a safety challenge for the AV space.
It’s not immediately clear what cameras, radar systems or lidar Rivian is basing this example on. But it’s the best illustration I’ve seen yet—pun intended—about why multiple, redundant sensor systems are needed to make an autonomous car that outperforms humans and won’t make the same mistakes we do.
Contact the author: patrick.george@insideevs.com More Rivian News We want your opinion! What would you like to see on Insideevs.com? – The InsideEVs team
