Tesla CEO Elon Musk is not a fan of LiDAR, a key technology used by leaders in the autonomous vehicle market such as Waymo. For Musk, vehicles should be able to drive themselves much like humans, through cameras and a neural network. While Musk’s stance on autonomous driving remains divisive, a recent study from Cornell researchers and comments from a full self-driving pioneer suggests that the Tesla CEO might have been right all along when it comes to using a non-LiDAR approach for autonomous driving.
Elon Musk did not mince his words when he was asked about his opinions on LiDAR during Tesla’s recently-held Autonomy Day. Musk, who actually uses LiDAR for SpaceX’s Dragon capsule, noted that the technology is useful in the right circumstances. It just so happens that autonomous driving is not one of them. Musk even predicted that companies that are deeply committed to LiDAR for full self-driving would eventually abandon the technology. “LiDAR is lame. They’re gonna dump LiDAR, mark my words. That’s my prediction,” he said.
Cornell researchers have finished a study titled Pseudo-LiDAR from Visual Depth Estimation: Bridging the Gap in 3D Object Detection for Autonomous Driving, which will be presented at the 2019 Conference on Computer Vision and Pattern Recognition in June. The study discusses a potential breakthrough for autonomous driving using strategically-placed cameras to produce stereoscopic images that can be converted to 3D data. Tesla is using a practically identical strategy in its vehicles, which was demonstrated by Sr. Director of AI Andrej Karpathy during his Autonomy Day presentation.
What is rather interesting from the findings of the Cornell researchers is that the 3D data generated from vehicles’ stereoscopic cameras were found to be nearly as precise as what laser scanners can generate. The data proved to be without distortion, and it was generated at a fraction of the cost had LiDAR been used. The researchers’ findings and model are still open for a lot of improvement, but the study does show that full self-driving without LiDAR is very much possible.
A stronger stance against LiDAR was also related recently by LiDAR pioneer Anthony Levandowski, founder and CEO of Pronto.ai, a self-driving startup that uses a non-LiDAR approach. Prior to serving as Pronto’s CEO, Levandowski was heavily involved in the development of Waymo’s full self-driving solutions. He later oversaw Uber’s autonomous driving program after his startup, Otto, was acquired by the ride-hailing juggernaut in 2016. During a recent discussion with TechCrunch, the autonomous driving pioneer, who is being accused by Waymo of breaching his confidentiality agreement with the company, described why he opted out of using LiDAR for Pronto.ai’s full self-driving solutions.
“I don’t have any restrictions on not doing LiDAR, but I do have restrictions personally of not doing things that I know are not gonna work. So, in the past, I understood, I could see what we were doing, and basically, back in 2009 or 2010, you could see that the LiDARs didn’t have the performance that you needed because you couldn’t see far enough to actually, safely, have the system react to software. It turns out that even with LiDAR, what’s missing today is not seeing more accurately or further; what’s missing is understanding what is happening to those vehicles around you and being able to predict what the motion is,” he said.
Anthony Levandowski and Elon Musk have not really seen eye to eye in the past, with the Tesla CEO criticizing the LiDAR pioneer for his stance on AI. Levandowski, for his part, reportedly insulted Musk in tweets to Travis Kalanick, Uber’s co-founder, at one point even suggesting that they start giving “physics lessons about stupid s**t Elon says.” The self-driving pioneer nevertheless proved humbled in his recent segment, stating that “a wiser person than me – and I’ll eat some humble pie here – said that LiDAR is a crutch.” When prompted by the host if he agrees with Elon Musk’s stance on autonomous driving, Levandowski admitted that “he’s right.”
Watch Anthony Levandowski’ interview on autonomous driving in the video below.