Is Uber Putting It in Reverse on Autonomous Vehicles?

Ronnie Schreiber
by Ronnie Schreiber

Is ride-hailing company Uber backing away from self-driving cars now that one of their test vehicles was involved in a fatality?

Following the death of a pedestrian hit by one of Uber’s experimental autonomous vehicles in Tempe, Arizona, the ride sharing company suspended the testing Uber was doing with AVs on public roads in Pennsylvania, Arizona, and California in the United States and Ontario in Canada. Now comes word that Uber has informed California’s Department of Motor Vehicles that it will not be renewing its license to test autonomous vehicles on that state’s public roads. That license expires at the end of this month.

Uber, along with the National Highway Transportation Safety Administration (NHTSA), the National Transportation Safety Board (NTSB), and the Tempe police department are each conducting investigations of the fatal incident. Already, attention has been focused on the reduced number of LIDAR sensors used on Uber’s current generation of AVs. When Uber switched its test fleet from Ford Fusion sedans to Volvo XC90 CUVs, it changed the number of the various types of sensors it uses to let the vehicles drive themselves. The Fusions had seven laser based LIDAR sensors, seven radar sensors, and 20 visual spectrum cameras. Uber’s autonomous Volvos have more radar units, ten, but fewer cameras, seven, and only a single roof-mounted LIDAR unit.

That LIDAR sensor is supplied by Velodyne, can sense things in a 360 degree circle, but it has a relatively narrow vertical range, making detection of objects close to the ground difficult. According to Velodyne, the system used by Uber also has near-field blindness and cannot detect obstacles within 3 meters of the vehicle.

The test vehicles currently used by Waymo, the Google related self-driving startup, use six LIDAR sensors, while the autonomous cars now being tested by General Motors have an array of five laser units.

In other news related to the fatal accident in Tempe, a spokesperson for Aptiv, which supplies Volvo with its own advanced driver assistance system that provides collision avoidance, lane-keeping, and other safety systems that use similar technology to those used in autonomous vehicles, said that Uber had disconnected the XC90’s native ADAS to test its own systems. Aptiv uses a combination of radar and camera based tech, which itself is supplied by Intel’s Mobileye unit. In a statement issued on Intel’s corporate website, Mobileye CEO Amnon Shashua wrote that tests using video from the Tempe accident showed that their own technology would have detected the pedestrian a second before impact.

The victim was Elaine Herzberg, 49. Her family’s attorney has announced that they have already arrived at a settlement with Uber for her death. Herzberg was trying to walk her bicycle across a four-lane road, outside the crosswalk, when she was hit by Uber’s self-driving Volvo.

[Image: Uber]

Ronnie Schreiber
Ronnie Schreiber

Ronnie Schreiber edits Cars In Depth, the original 3D car site.

More by Ronnie Schreiber

Comments
Join the conversation
4 of 15 comments
  • Bluegoose Bluegoose on Mar 30, 2018

    Uber's Volvos only have 1 Lidar sensor while the Waymo has several more. I think Uber has an inferior design. It is time for Uber to go back to the drawing board.

    • See 1 previous
    • Ect Ect on Apr 02, 2018

      It may be past time for Uber to acknowledge that they are a taxi company, not a technology company. They bring no competitive advantage to the AV business. And as a taxi company, they aren't even profitable - managed to lose $4.5 billion in 2017. Management should be focused on achieving profitability, not playing with AVs.

  • Conundrum Conundrum on Apr 01, 2018

    Poor old Li Shufu, prez of Geely. Uber planned to buy 24,000 self-driving cars from Volvo once this frippery of actual real world testing was complete. Fat chance of that big sale now. No word on whether a leading Swedish industrial designer had completed a stylized casing to beautify the lidar unit's housing sitting on the vehicle's head. Design brief: Had to be compatible with a roof rack, or 20 foot canoe strapped to the lid for their more outdoorsy Canadian customer, Mr Canoehead. No doubt the NTSB investigation will answer the following obvious question: How did the accident vehicle manage to trundle around Tempe for hours avoiding lamp posts and concrete barriers, boring its "always-on-the-ready" human attendant to tears? Then ignore the 35 mph speed limit by a human amount, just a measly 3 mph, your Honor, honest! And then smack into a human its sensors failed to notice a good 100 yards short of spec? You know, unlike any number of available pedestrian alert and braking systems currently available on drossly ordinary cars like a Corolla. Perhaps the Jedi will have to be consulted. It's interesting that Uber appears to be manning up. If it were Tesla, they'd have invented ten reasons for it not being Autopilot's fault, while absolving themselves of responsibility and using the tactic of studied omissions of the obvious to muddy the waters so much that nobody would remember WTH happened a week later.

  • Calrson Fan Jeff - Agree with what you said. I think currently an EV pick-up could work in a commercial/fleet application. As someone on this site stated, w/current tech. battery vehicles just do not scale well. EBFlex - No one wanted to hate the Cyber Truck more than me but I can't ignore all the new technology and innovative thinking that went into it. There is a lot I like about it. GM, Ford & Ram should incorporate some it's design cues into their ICE trucks.
  • Michael S6 Very confusing if the move is permanent or temporary.
  • Jrhurren Worked in Detroit 18 years, live 20 minutes away. Ren Cen is a gem, but a very terrible design inside. I’m surprised GM stuck it out as long as they did there.
  • Carson D I thought that this was going to be a comparison of BFGoodrich's different truck tires.
  • Tassos Jong-iL North Korea is saving pokemon cards and amibos to buy GM in 10 years, we hope.
Next