‘Lidar is lame': why Elon Musk's vision for a self-driving Tesla taxi faltered

The Guardian

6 hours ago

After years of promising investors that millions of Tesla robotaxis would soon fill the streets, Elon Musk debuted his driverless car service in a limited public rollout in Austin, Texas. It did not go smoothly.
The 22 June launch initially appeared successful enough, with a flood of videos from pro-Tesla social media influencers praising the service and sharing footage of their rides. Musk celebrated it as a triumph, and the following day, Tesla's stock rose nearly 10%.
What quickly became apparent, however, was that the same influencer videos Musk promoted also depicted the self-driving cars appearing to break traffic laws or struggle to properly function. By Tuesday, the National Highway Traffic Safety Administration (NHTSA) had opened an investigation into the service and requested information from Tesla on the incidents.
If Tesla's limited rollout of the robotaxi service was the culmination of more than a decade of work, as Musk touted on X, its struggles are also emblematic of technical decisions and fixations that the world's richest person has embraced as he pursues the goal of a fully autonomous car.
Musk has cast the concept of a driverless car as a core part of the company's future business, and, as sales have sharply fallen this year, he has vowed that its robotaxi service will rapidly and drastically expand. Yet the faltering launch this week suggests Tesla is still facing technological challenges that have attracted regulators' notice, delayed Musk's vision of a robotaxi on every corner, and highlighted the gulf between it and its driverless rival, Waymo.
The robotaxi launch featured about 10 cars traveling in a limited area of Austin with safety drivers in the passenger seat. The pilot included other restrictions, such as not operating in bad weather or during certain nighttime hours. Rides, which the company offered to a host of handpicked influencers, cost $4.20, in keeping with Musk's proclivity for cannabis memes.
'Tesla self-driving can be deployed anywhere it's approved. It does not require expensive, specialized equipment or extensive mapping of service areas,' an official Tesla account posted on X the day of the launch. 'It just works.'
Footage from at least 11 rides showed that the trial run did not pan out as flawlessly as Tesla's tweet suggested. In one case, a robotaxi failed to make a left turn and instead drove into a lane meant for oncoming traffic, then corrected itself by driving across a double yellow line. Other videos appeared to show the cars exceeding the speed limit, braking for no discernible reason and dropping passengers off in the middle of an intersection.
The videos drew the attention of the NHTSA, which said in a statement it was aware of the incidents and had contacted Tesla to obtain more information.
Musk, meanwhile, posted throughout the technical failures and regulatory inquiry, retweeting pro-Tesla influencers who praised the service. One account Musk posted showed off a video of a robotaxi stopping to avoid running down a peacock crossing the road, and another told followers: 'Don't listen to the media.'
Musk has long insisted that using only cameras on driverless cars is the singular way to achieve true self-driving capability. Tesla's consumer vehicles come with what it calls 'autopilot' and 'full self-driving' features that allow drivers to cruise on the highway hands-free. They rely on multiple external cameras to navigate, steer and brake. The company's robotaxis use similar software and also depend solely on cameras.
The reliance on cameras alone stands in sharp contrast with other autonomous vehicle companies such as Waymo and Zoox. Those companies use arrays that combine cameras and sensors, including radar and lidar. For example, the newest version of a driverless Waymo uses about 40 external cameras and sensors, whereas a Tesla with one of the latest versions of full self-driving uses about eight external cameras, according to an analysis by Bloomberg. Lidar and radar allow for self-driving cars to better detect objects in bad weather and poor lighting.
Despite the advantages to lidar and radar, Musk has been adamant that Tesla remain lidar-free. 'Lidar is lame,' Musk said during a Tesla autonomy day in 2019. 'In cars, it's friggin' stupid. It's expensive and unnecessary.'
Lidar is far more expensive, costing roughly $12,000 per vehicle, as compared with cameras, which come in at around $400 per car, according to Bloomberg. Musk maintains that camera-only technology is the most 'human' way to approach self-driving, since people use their eyes to navigate the road.
Musk's insistence on camera-only technology has landed Tesla in hot water over fatal crashes involving drivers using the full self-driving feature. The company is now the focus of government investigations and civil lawsuits, which allege that full self-driving is impeded by weather conditions such as sun glare, fog, dust and darkness. There have been at least 736 crashes and 17 deaths involving the technology, according to an analysis by the Washington Post.
'Tesla continues to have this fetishistic view that it's going to operate its system solely on cameras, despite every intelligent human being in this entire space saying that can't be done,' said Brett Schreiber, an attorney who represents several alleged victims of Tesla's autopilot failures.
'Everyone who has been following collision-avoidant technology since the 90s knows that the holy trinity is radar, lidar and cameras.'
Schreiber said he was not surprised to see the wobbling rollout of Tesla's robotaxis in Austin.
'What you're also going to see, which is the true tragedy of this thing, is people continuing to be injured and killed by this technology,' he said. 'And that's where it becomes less of a 'Oh, isn't that cute? The vehicle can't make a left' to now we're actually at someone's funeral because of the choices Tesla makes.'
Tesla did not respond to a request for comment on the lawsuits, government investigations and crashes involving full self-driving.
The differences between Waymo and Tesla's approaches to launching commercial self-driving services in dense cities don't end with the debate over lidar versus cameras. Waymo is seen widely as the frontrunner in the self-driving race in the US – a race that was once crowded with dozens of automakers, VC-backed startups and ride-share companies and has since been whittled down to just a handful of major players.
There are numerous reasons why Waymo has outlasted so many of its competitors and why it's ahead of the curve. The Google subsidiary has historically spent months, if not years, mapping cities and testing its vehicles in them before launching. In San Francisco, one of the first cities where Waymo launched its fully driverless commercial service, the company began mapping out and testing its service in 2021 before launching it to the public in 2024.
Even with a careful and gradual city-by-city approach, Waymo, which launched as a project under Google's X research lab in 2009, has encountered problems with its self-driving cars. Earlier this year, Waymo had to recall more than 1,200 of its vehicles over a software issue that was causing collisions with chains, gates and other stationary roadway barriers. The NHTSA also launched an investigation into the company last year after the agency received 22 reports of Waymo vehicles acting erratically or potentially violating traffic safety laws.
Contrast Waymo's approach with Tesla's. While Tesla is still in the testing phase of its service, its robotaxi launch in Austin is the first time the car company's fully self-driving technology is being unleashed in the wild. The company has not released information on whether, or how long, it has spent mapping out or testing the driverless technology on Austin's streets.
The launch is reminiscent of Uber's first foray into a self-driving ride-share service in 2016. The company launched a self-driving pilot in San Francisco without seeking a permit from the California department of motor vehicles, as was required. On the first day of the pilot, an Uber vehicle ran a red light. The company was forced to shut down the service a week later after the DMV revoked its registration. An Uber self-driving executive at the time had pushed the company's engineers to rush to launch the San Francisco pilot to attract more investor and public attention.
After being sued by Waymo over its self-driving operations and struggling to catch up to its competitors, Uber sold its self-driving arm in 2020.
Tesla also did not have a permit to operate its robotaxi service in Austin. Texas does not currently have a process to acquire a permit and won't have one in place until September.
While there's at the moment less visibility into what Tesla's rollout of its robotaxi service looked like behind the scenes, the automaker is no stranger to rushing to meet deadlines set publicly by Musk.
With the launch of robotaxis, Musk, who has been promising that Teslas will be fully self-driving since at least 2016, is perhaps getting closer to meeting the deadline that he set and has deferred several times over the past 10 years.