Good morning! Thanks for reading. Please share this newsletter and tell your friends they can subscribe here. If you have tips or feedback, just reply to this email.
Each edition, we draw on the expertise of people working in the emerging AV industry and the sectors it touches. Editors work with these Axios Experts to bring you analysis of the field. And we're experimenting with ways to better highlight their insights in the newsletter and on Axios.com.
1 big thing: Humans control the remote in some AVs
For all the talk of cars without drivers, the deployment of automated vehicles might still require humans controlling them from a distance. Most major AV companies are testing or planning to incorporate remote control — or teleoperations — in their robot-driven cars.
The big picture: AVs need help making decisions in complex situations, which is why the hype about fully self-driving cars remains unfulfilled. With humans taking the wheel via remote control, some companies hope to speed AV deployment, but questions remain about safety.
What's happening: Most AV companies plan to use teleops to some degree, including GM Cruise, Toyota, Zoox and Nissan.
- Waymo, which launched its commercial robotaxi service earlier this month in Phoenix, uses remote control operators in certain situations.
- Phantom Auto's remote technology can be applied by any automaker.
- California approved driverless testing without backup drivers, but only if the vehicles can be operated remotely.
- Starsky Robotics today filed its Voluntary Safety Self-Assessment with the federal government, outlining how it plans to use remote control drivers to deploy automated trucks for long-haul shipping.
How it works: Starsky's system — currently operating in 3 trucks — allows trained drivers to sit in an office and control a truck using computer screens, buttons, a steering wheel and pedals.
- Drivers can remotely drive a truck from a distribution center to a highway, where automation takes over. At that point, the driver is only supervising the truck to help with complex, context-based decisions.
- When the truck exits the highway, the remote driver regains control to steer the truck to its final destination.
"Computers are great at saying, 'Let’s keep between the lanes and manage the speed.' But making a judgment call on whether now is a good time to pass is really hard. It only gets worse in disorganized environments like truck yards."— Stefan Seltz-Axmacher, co-founder, Starsky Robotics
Why it matters: Teleoperation could help alleviate a labor shortage, currently pegged at 63,000 truck drivers, says Seltz-Axmacher. The American Trucking Associations warns that could worsen to 175,000 by 2026.
- Starsky Robotics envisions 1 driver guiding 30–40 trucks in an 8-hour shift.
- The company says driving from behind a desk is less taxing than the typical long monotonous stretches on the road.
Yes, but: It's hard to imagine a remote driver hundreds of miles away can make a quick judgment by dropping in on a complex situation. In most cases, the AV will simply stop and wait for further instruction.
- Another concern is that teleoperation relies on ordinary cellular networks that, if delayed, could prevent a remote operator from making a quick decision at a critical time.
- Safety groups say that it's too much like a video game, with human lives at stake.
The bottom line: The vision of self-driving cars might still include human drivers, even when they're not in the vehicle.
2. Uber gets OK to resume AV testing in Pittsburgh
The Pennsylvania Department of Transportation has approved Uber’s request to resume testing of AVs on public roads in the Pittsburgh area, AP's Tom Krisher reports.
The big picture: The decision comes about 9 months after testing was suspended following a crash in Tempe, Arizona, in which an autonomous test vehicle operated by Uber struck and killed a pedestrian.
Details: Uber filed an application to restart testing back in November, after issuing a lengthy safety report pledging to put 2 human backup drivers in each vehicle and take a raft of other precautions to make the vehicles safe.
What's next: Uber will seek to resume testing on public roads in Arizona, California and Toronto, its other test sites.
3. AI tackles the pothole problem
By Karen Lightman • Metro21: Smart Cities Institute • Axios Expert
The poor state of infrastructure poses serious safety threats to drivers today and may delay the safe deployment of AVs. One-third of the 33,000 annual traffic-related deaths in the U.S. involve poor road conditions, with potholes alone causing $3 billion in vehicle damage per year.
The big picture: AV developers have focused on improving how their cars respond to pedestrians, bicyclists and other cars on the road, but they must also address the challenges presented by infrastructure.
Where it stands: Companies and academic labs are working on technology for AVs to anticipate hazards on the road, which could also provide data to guide infrastructure repairs and investments.
- RoadBotics, a company spun out from Carnegie Mellon University, combines AI and machine learning with machine vision adapted from smartphones to identify road infrastructure issues in real time.
- Carnegie Mellon's LiveMap uses edge computing to analyze a video feed of the road and can detect and report on conditions.
- Cornell University researchers took data from sensors and accelerometers and combined them with machine learning algorithms to detect and assess potentially damaging road conditions, including potholes.
- TotalPave, a startup based in New Brunswick, Canada, has developed a Pavement Condition Index calculator that collects road conditions data in a mobile app.
What to watch: As digital infrastructure moves to 5G and as computing power increases, machine learning will be more widely implemented on the road. Until then, municipalities can use these new technologies to make the roads safer.
Lightman is executive director of Metro21: Smart Cities Institute at Carnegie Mellon.
4. Driving the conversation
Bad blood: Two excellent reads on the Carlos Ghosn saga at Renault Nissan Mitsubishi. (Hans Greimel — Automotive News; Sean McLain, Phred Dvorak, Sam Schechner and Patricia Kowsmann — The Wall Street Journal)
- Why it matters: The arrest of Ghosn — the architect of what had been viewed as a massively successful global automotive alliance — appears to stem from a power struggle with his hand-picked successor, Nissan CEO Hiroto Saikawa. The big question now is whether the soap opera will distract the companies as massive changes sweep through the industry.
He’s baaaaack: Anthony Levandowski unveiled his new autonomous driving startup, Proton.ai. (Medium)
- Details, from Axios' Kia Kokalitcheva: Levandowski, an alumnus of Google’s self-driving car project and Uber, argues that unlike most other companies, Proton is building an advanced driver-assistance system that’s focused on learning how to drive and react like humans instead of software that “tells vehicles how to drive.”
- What he’s saying: Notably, he also refers to lidar and other common autonomous driving tech as a “crutch”— quite ironic given that Uber and Waymo spent a year in a legal battle over whether he stole lidar tech from the latter.
German robotaxis: Audi pulls the curtain back on its self-driving car program. (Andrew J. Hawkins — The Verge)
- The big picture: Audi plans to spend almost $16 billion on electric mobility and self-driving technology through 2023. Its subsidiary, Autonomous Intelligent Driving, is developing SAE level 4 robotaxis for use throughout the Volkswagen Group.
5. The gender gap in speeding deaths
More than one-quarter of all highway fatalities are caused by speeding, and men are to blame more than women.
By the numbers: 10,111 people were killed in speeding accidents in 2016, up 4% from the year before, according to NHTSA. In fatal crashes among all age groups from 2010–2017, more male drivers were speeding than female drivers.
The big picture: AVs are supposed to make roads safer, but even they won't necessarily put an end to speeding if a hurried driver disengages the system to get to work on time.