No, Nissan's Position On Lidar Is Not The Same As Elon Musk's
In the rush to make everything about Tesla and Elon Musk, important issues around automated driving are being shown through a scanner, darkly.
Tesla CEO Elon Musk's recent disparagement of the laser scanning technology called lidar, calling it a "crutch" and any autonomous vehicle developer using it "doomed," has created a massive online debate about the value of what has been considered a key technology enabling recent advancements in self-driving technology. But, like so many online debates these days, the conversation that Musk launched has generated more heat than light. The unhelpful nature of this discourse reached a new peak yesterday, when multiple media outlets twisted comments made by Nissan's Executive Vice President Kunio Nakaguro and general manager of advanced technology development for automated driving Tetsuya Iijima,to fit Musk's outlier position and in the process missed an opportunity to actually educate their readers about this complex topic.
This situation seems to have originated with a Reuters write-up of Nakaguro's and Iijima's comments at the Japanese launch of ProPILOT 2.0, a hands-free automated driving system whose use is limited to specific single-lane highways in Japan. “At the moment, lidar lacks the capabilities to exceed the capabilities of the latest technology in radar and cameras,” Iijima, is quoted as saying. “It would be fantastic if lidar technology was at the level that we could use it in our systems, but it’s not. There’s an imbalance between its cost and its capabilities.” Reuters does clarify that these comments represent Nissan's position "for now," and doesn't state that the company's position is the same as Musk's but simply by mentioning the Tesla CEO's comments they have fueled a series of misleading stories that obfuscate important nuance.
These quotes, seemingly stripped of their context, have inspired a wave of headlines equating the Nissan executives' comments with Musk's comments, at outlets ranging from The Daily Mail and Teslarati to Electronics Weekly and Left Lane News. But those two words of context in the Reuters lede was critical: speaking with The Drive, Nissan spokesman Nicholas Maxfield confirmed that Nakaguro and Iijima's comments were referring to ProPILOT 2.0 and similar limited-domain automated driving systems that can be marketed on private sale cars today, rather than a blanket disparagement of lidar technology for all future automated driving systems. Nissan has not ruled out the use of lidar for any fully autonomous shared vehicles the company may be developing, he said.
That context dramatically differentiates Nissan's position from that of Tesla and Musk, who has said that lidar is unnecessary for the electric automaker's goal of launching "Full Self-Driving" software next year that is capable of full autonomy in any and all conditions (SAE Level 5). That position is extremely controversial, bucking the conventional wisdom which holds that lidar provides critical sensor diversity and redundancy in complex domains like urban settings. Nissan's position, that lidar is unnecessary for the limited-domain highway automation in vehicles available for sale today, is far less controversial than Musk's more sweeping comments since no other automaker has taken a strong stance on the necessity of lidar to automated driving systems limited to operating in the relatively easy domain of single-lane highways.
Part of the confusion that led to media outlets equating Nissan's position with Tesla's comes from the overly-broad application of the term "self driving," which is applied equally to Nissan's limited-domain system and Tesla's Level 5 autonomous aspirations by everyone from Reuters on down. This was exacerbated by Nissan's decision to avoid any reference to the SAE automation levels when describing its ProPILOT 2.0, which could be described as either Level 2 or Level 3. Such limited systems, which require some level of "human in the loop" driver awareness or engagement for safe performance are fundamentally different than a Level 4 or Level 5 fully autonomous system that requires no human driver awareness or engagement and can be used in less limited domains than a highway-only Level 2 or Level 3 system. This difference in domain and the need for human driver awareness or engagement is central to the debate about the sensors necessary in a given automated driving system, and by collapsing this important distinction media outlets have done a disservice to their readers.
Musk's popularity, and the tendency of his fans to take his every word as gospel, has created a pattern in which comments and research are twisted to support his extreme anti-lidar position. We've seen this before, when an extremely early-stage research paper from Cornell was presented as evidence that Tesla could achieve its Level 5 autonomy goal next year without lidar, despite the considerable limitations of the research. Musk and Tesla have fueled this tendency by suggesting that Autopilot is "self-driving," helping blur the distinction between a limited-domain human-in-the-loop automated driving system like Autopilot or ProPILOT 2.0 and the Level 5 "Full Self-Driving" system it hopes to deploy next year or the Level 4 autonomous systems that are coming from other companies. Throughout the history of Tesla's involvement with automated driving technology, the company has consistently represented its systems as being more autonomous than it really is, leading to numerous examples of people using it in an unsafe manner and several fatal crashes.
What this unfortunate discourse overlooks is the core issue of sensor redundancy and diversity, which is far more important than just the value of lidar itself. The majority of Level 4 autonomous vehicle developers who believe in the necessity of lidar say its value lies in the fact that it provides another layer of sensor redundancy and diversity, on top of cameras and radar. Even Nissan radically diverges from Tesla's Autopilot sensor strategy, by including four side-mounted millimeter radar units that provide 360 degree redundant and diverse sensor capabilities for highway-only automation whereas Tesla's single forward-mounted radar keeps its system completely reliant on cameras for all but the forward quadrant. Without even discussing Tesla's extremely ambitious "Full Self-Driving" aspirations, this shows that Nissan takes a much more thorough approach to sensor diversity and redundancy for limited-domain, highway-only, human-in-the-loop systems than Tesla.
This distinction is far more real and meaningful than any similarities between Musk's extreme anti-lidar position and Nissan's more limited and nuanced position, and yet not a single story coming out of the ProPILOT 2.0 announcement seems to even notice it. Nissan's use of a driver monitoring system (DMS) for its hands-off highway-only system is another major safety-critical distinction between Autopilot and ProPILOT 2.0 that is at the center of several recent fatal Autopilot crashes, but which seems to have escaped the notice of reporters (though to be fair, Nissan hardly emphasizes its DMS in its official press release).
This all points to the need for reporters and editors, including all of us here at The Drive, to take more care to emphasize these important distinctions rather than bending every new announcement or comment to fit the contours of Elon Musk's position. Though Musk has done much to popularize a variety of automotive technological trends, it's important to understand that he has also obfuscated extremely important nuances that differentiate different kinds of automated driving systems, and issues like sensor redundancy and diversity. Given that misunderstandings about automated driving systems can have fatal results, it is incumbent on all of us who cover this space to be far more knowledgeable and detail-oriented than we have been, even at the expense of fueling a traffic-juicing celebrity controversy.