Tesla faces industry scrutiny over the best way forward for autonomous cars

Tesla faces industry scrutiny over the best way forward for autonomous cars

DETROIT -- Tesla Motors' decision to equip all of its vehicles with self-driving hardware has intensified competition among rival camps of technology and auto companies over what equipment will be on board cars of the future.

Tesla's new self-driving system will rely on cameras and radar sensors -- but not lidar, the laser imaging technology most other companies pursuing self-driving cars are using to generate precise pictures of the environment around their vehicles.

Tesla also is not using technology from Mobileye, the Israeli-based supplier of computer vision chips and software that provided components for earlier Tesla models equipped with Autopilot, a semi-automated system designed to assist with driving but not as a replacement for the driver.

Tesla CEO Elon Musk and Mobileye this summer engaged in a public dispute over the safety of Autopilot following a fatal crash in May in which the driver of a Tesla Model S was killed after hitting a truck while driving on Autopilot.

"Selling a vehicle with the latest and greatest hardware, but an unproven self-driving software package is a risky strategy," said Barclays analyst Brian Johnson.

Mobileye supplies two dozen other automakers and suppliers, and has formed alliances to pursue self-driving vehicle systems with German automaker BMW Group, U.S. supplier Delphi and chip maker Intel.

"As we move to a higher level of autonomy in vehicles, you're going to want to have more redundancy," which radar and lidar can provide, Dan Galves, senior vice president at vision safety system maker Mobileye, said in an interview. "The more sensors, the better."

A winner in the fallout from the Tesla-Mobileye dispute is Nvidia, the Silicon Valley chipmaker that will provide processors for the new Tesla system.

No automaker, including Tesla, currently offers a fully self-driving car, although most major manufacturers and suppliers are working furiously on different technology suites -- including cameras, radar and lidar -- to enable vehicles to drive themselves.

Google's self-driving cars use a combination of lidar, radar and cameras. The company has said the multi-sensor approach compensates for the limitations of each type of sensor.

Cameras do not perform well in the dark or in bright light. Ice and snow can block a camera's view. Lidar can create a 360-degree view of the car's surroundings, although it is less effective in fog, rain and snow.

The high cost of lidar, currently about $8,000, is just one of the obstacles to putting self-driving cars in consumers' hands. But the cost is rapidly declining and is expected to fall to $25-$100 in the coming years.

Ford Motor, an investor in lidar manufacturer Velodyne, said it does not plan to put a fully self-driving vehicle into production for ride-sharing fleets until 2021 and for individual consumers until 2025 or later. Toyota Motor Corp. has said it does not expect to see Level-5 vehicles -- those capable of fully automated operation without humans in all situations -- in widespread use for another 10-15 years.

Major automotive technology suppliers, including Delphi Automotive, Robert Bosch, Continental and Valeo, are all developing self-driving technology and systems that incorporate lidar.

Without lidar, Tesla "will not be able to handle all situations with this array of sensors," said Marta Hall, head of California-based Velodyne, one of the largest suppliers of automotive lidar.

UBS analyst Colin Langan raised a concern that while excluding lidar could lower the overall cost of a self-driving system -- now pegged by Tesla at $8,000 without lidar -- "it would increase negative scrutiny" if it is proven that a lidar sensor could have prevented a fatal accident.

Regulators have signaled they intend to keep tighter reins on automakers and other companies, including Google and Uber Technologies, as they roll out self-driving vehicles on public roads.

Summarizing the concerns of other automakers and suppliers involved in developing self-driving systems, Langan said Tesla's early deployment of full automation "without enough testing could have negative long-term consequences" for widespread adoption of self-driving technology.

"Lidars are becoming very inexpensive, and they will always add an additional safety layer. So in these early days of self-driving, I wonder why anyone wouldn't want to use lidars for added safety," said Sebastian Thrun, former head of Google's self-driving car project.