Self-Driving Cars

Elon Musk Dumped Radar for Tesla’s Self-Driving Cars. His A.I. Chief Explains Why.

June 23, 2021 12:59 pm ET

  • Order Reprints
  • Print Article

Text size

The gull-wing doors on a Model X SUV aren’t the most interesting tech available on Tesla vehicles. Getty Images

The electric-vehicle pioneer Tesla is blazing its own path in autonomous driving. It explained why in a presentation at the Computer Vision and Pattern Recognition conference, the latest development in an emerging debate about competing technologies in a new and incredibly complicated area.

Autonomous driving is really the summation, and refinement, of all the individual safety systems available in cars today. Automatic emergency braking, adaptive cruise control, lane-keeping assistance, and PMM—short for pedal misapplication mitigation, for when a driver steps on the gas instead of the brake—are available on many new cars right now. The challenge for car companies is to merge all that into a system more effective and reliable than a human.

The first problem is perception. The vehicle has to see what is around it to put those subsystems to work at the right times. This is the biggest area where Tesla (ticker: TSLA) is diverging from its automotive peers. The company recently stopped putting radar sensors in its cars, choosing to ship vehicles with only optical cameras as the eyes to its driver-assistance functions.

Every other auto maker uses a mix of cameras, radar, and lidar—laser based radar—to collect the data its cars use to help make decisions. Lidar is rare because it is expensive, while radar is inexpensive and widely used.

“Radar is a crutch,” Andrej Karpathy, Tesla’s senior director of artificial intelligence, told the conference on Monday.

Radar is used to measure things like speed and distance, but Tesla believes cameras can do that too. Disagreements between radar sensors and optical cameras were hurting the development of Tesla’s autonomous-driving functions, so Tesla decided to focus on refining the cameras.

“The vision system that we have been building over the past few years is so incredibly good that’s its kind of leaving other sensors in the dust,” Karpathy said.

After perception, or the gathering of data, comes the computing. Karpathy spent a good portion of his presentation on Tesla’s artificial intelligence and computers.

They sound impressive, but their sophistication raises a problem for investors. Anyone who isn’t up to speed on terms such as three-dimensional convolution, sensor fusion, neural networks, synthetic cortex, and tops—short for trillions of operations per second—has little basis to assess Tesla’s approach.

It also isn’t clear how Tesla’s efforts stack up relative to what other companies have achieved in working on self-driving cars.

The car parts and technology provider Aptiv (APTV) has billions of miles driven and simulated with its advanced safety systems. It also has thousands of safety engineers on staff. Aptiv can deliver a driver-assistance package similar to Tesla’s Autopilot straight off the shelf to any auto maker that wants to purchase one. But the company rarely talks about the details.

Tesla might have one advantage that can be quantified, however. It has hundreds of thousands of test vehicles in the field, the cars it has already sold, constantly learning. Karpathy says Tesla’s neural networks work in “shadow mode” on Tesla vehicles, silently making predictions for what drivers will do and adding to the robustness of the company’s driving software.

A lot is riding on Tesla’s self-driving programs. CEO Elon Musk said in the third quarter of 2020 that full self-driving technology would amount to the “biggest asset value increase in history.”

Autonomous cars, for instance, can earn their owners money by driving people around when their owners are at work or asleep. Barron’s recently estimated that $100 to $200 of Tesla’s current stock price, at about $650 on Wednesday, was based on expectations the company would deliver of higher self driving functionality to drivers by year-end.

If Tesla can make real progress and show that autonomous driving is safe, other car companies could follow its lead, selling software-enabled self-driving functions. A vast new field of innovation, and a big source of potential profits, would open up for the industry.

For now, though, investors seem to be waiting for concrete progress. Shares are down about 12% year to date, trailing behind comparable gains of the S&P 500 and Dow Jones Industrial Average.

Write to Al Root at [email protected]

Autonomous driving is really the summation, and refinement, of all the individual safety systems available in cars today. Automatic emergency braking, adaptive cruise control, lane-keeping assistance, and PMM—short for pedal misapplication mitigation, for when a driver steps on the gas instead of the brake—are available on many new cars right now. The challenge for car companies is to merge all that into a system more effective and reliable than a human.

Source: https://www.barrons.com/articles/elon-musk-tesla-self-driving-cars-artificial-intelligence-51624467471

Donovan Larsen

Donovan is a columnist and associate editor at the Dark News. He has written on everything from the politics to diversity issues in the workplace.

Related Articles

Leave a Reply

Your email address will not be published. Required fields are marked *

Back to top button