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.
Today, Expert Voices contributor Robin Elliott takes us under the ocean to see how unmanned underwater vehicles are advancing.
Situational awareness: Today California's DMV will release "disengagement reports" for AV companies testing cars there. The annual reports, which include info on mileage accumulation and how frequently a human driver takes over, have become a proxy for judging a company's progress.
1 big thing: Billion-dollar bets on driverless cars
Two promising AV startups recently received eye-popping funding, but don't assume 2017's frothy hype is back.
The big picture: Blitzscaling, or investing to achieve massive growth before competitors do, doesn't work in the capital-intensive AV business. And with self-driving cars still a long way off, many contenders could start to fade in the face of massive demands for funding and talent.
What's happening: Two AV companies scored huge rounds of fresh funding in the last few days, giving them the resources to keep developing their technology and add engineering talent.
- Aurora Innovation secured over $530 million in Series B financing led by Sequoia, valuing the company at $2.5 billion. Amazon also invested, a development that has tantalizing implications.
- Nuro, a robotic-delivery vehicle startup, raised $940 million from SoftBank Vision Fund, a deal that valued the company at $2.7 billion.
Meanwhile, in more modest but still important news yesterday, May Mobility raised $22 million in Series A funding to keep developing its last-mile AV transit solution.
Quick take: These companies have something in common: a path to commercialization. Nuro and May Mobility already have tiny fleets of low-speed AVs on the road. Aurora is developing a "driver" for any vehicle and has big-name partners including Hyundai and Volkswagen.
- Leadership credibility matters, too. The co-founders of Nuro and Aurora are veterans of the 2007 DARPA Urban Challenge and some worked on Google's self-driving car project (now Waymo) before starting their own companies.
What we're seeing: With the magnitude of the challenges facing AV technology sinking in, investors have stopped chasing the dream and are narrowing their focus to companies with a commercial niche and actual revenue potential.
- In contrast to the way software apps multiply, AV companies need to map each city individually and then test their vehicles on local roads before launching — a meticulous, 6-month effort that must be repeated for each city.
- Even with massive investments of time and money, full-scale deployment remains slow and limited.
- Safety must come before speed and scale. Consider Uber's missteps in Phoenix that led to a pedestrian's death.
- Public fear of AVs remains an issue. People still have to decide whether AVs are safe, affordable, convenient — or even needed.
Where it stands: Some 60 companies are spending "high single-digit billions" annually to try to grab a commanding share of what's expected to be a $1 trillion AV market, says Mark Wakefield, managing director of advisory firm AlixPartners.
"The worry is that most of the money is going to be wasted because every one of them is hyper-convinced that they’re going to win. They’re clearly not going to win."— Mark Wakefield
The bottom line: After a lot of churning, the cream is rising to the top, says Varun Jain of Qualcomm Ventures, which exited an early-stage investment in Cruise Automation when GM bought the AV startup in 2016.
"Companies that have a difficult path to market, or don’t have the talent, will struggle. So there will be a correction. But that's healthy."— Varun Jain
2. GM and Amazon eye Rivian's EV skateboard
Amazon and GM are in talks to invest in electric-truck maker Rivian Automotive in a deal that would value the Michigan startup at between $1 billion and $2 billion, Reuters reports.
Why it matters: Rivian made a splashy debut last November with 2 new rugged battery-powered models — a pickup truck and a 7-passenger SUV. But it's the electric, semi-automated chassis underpinning those models that has likely attracted the interest of Amazon and GM.
Details: Rivian intends to market its own brand of trucks — as many as 6 models some day — but from the beginning, it has been plotting a B2B strategy to share its technology with other companies.
- Rivian's battery-powered "skateboard" platform can scale up or down and is designed to be modular, so it can accommodate many types of vehicles, from pod cars to delivery trucks and even snowmobiles.
- It includes 4 wheel-mounted motors that provide torque and maneuverability.
- The digital architecture will support fully self-driving technology (Level 4).
What we're hearing: In an interview with Axios in October, Rivian founder RJ Scaringe mused about companies like Amazon, Starbucks or Apple launching their own mobility fleets on top of a generic platform.
- At the time, he said the company was in talks with 6 potential strategic investors, both tech giants and automakers, that he wouldn't name.
- In an emailed statement, GM would only say: “We admire Rivian’s contribution to a future of zero emissions and an all-electric future.”
- As mentioned in the story above, Amazon has also invested in Aurora via a $530 million funding round announced last week.
What to watch: If the negotiations conclude successfully, a deal could be announced as early as this month, Reuters says, citing unnamed sources.
3. Increasing autonomy under the sea
Partially autonomous and remotely operated underwater vehicles are used to map oceans, observe sea life and pollution, clear underwater unexploded ordinance, and monitor oil and gas pipelines — jobs that are dangerous or impossible for divers, Robin Elliott writes for Axios Expert Voices.
What's next: Advanced communication technology, better batteries and other breakthroughs could result in vehicles with more autonomy, capable of more complex missions.
Where it stands: Unmanned underwater vehicles (UUVs) are already equipped with state-of-the-art sonar, cameras, and laser equipment and can deploy for up to 24 hours.
- Freed from road rules that constrain AVs, UUVs can more easily operate with high levels of autonomy.
- The biggest limitations to UUV missions are technological shortcomings, like limited battery life, and communications problems if remote-operated UUVs go out of range.
What's new: Academic researchers, manufacturers and militaries are trying to push UUV technology forward.
- Some researchers are designing systems that could stay underwater for up to 9 months. Longer deployments will require more efficient battery technology, and ways to supply power for underwater recharging of UUVs.
- Increasing the amount of pressure UUVs can handle is another goal, so they could descend to depths of 6,000 meters. (Human divers can handle 80 meters, at most, and current UUV systems can reach 3,000 meters.)
- ATLAS, Kongsberg, and Saab are working specifically on increasing the efficiency of mine neutralization and disposal technology. In the Baltic Sea alone, there are an estimated 50,000 mines that need disposal, which make it dangerous to build new underwater pipelines or establish new ferry routes.
- DARPA recently put out a call for an underwater AV developer for seabed-level work.
The bottom line: For UUVs to take on longer and more complex missions, underwater power transfers will have to be simplified and the systems made more reliable. But their advances — from AI and self-learning algorithms to battery technology — could offer a valuable model for developers of other kinds of AVs.
Go deeper: Read the full post.
Elliott is a program manager at Saab Seaeye, which develops underwater robotic systems. He is also a member of GLG, a platform connecting businesses with industry experts.
4. Driving the conversation
Map update: Japanese self-drive cars map developer to buy rival U.S. startup for $200 million (Reuters)
- Why it matters: High-definition mapping is critical to the operation of self-driving cars. By acquiring the company that enabled GM’s Super Cruise highway driving system called Ushr, Japan’s Dynamic Map Platform gets a foothold in North America.
10-year anniversary: Waymo CTO on the company's past, present and what comes next (Kirsten Korosec — TechCrunch)
- Background: CTO Dmitri Dolgov was there at the beginning in 2009 when Google secretly started working on its self-driving car project, and he's still there, helping to commercialize it a decade later.
"There is a huge difference between having a prototype that can do something once or twice or a handful of times versus building a product that people can start using in their daily lives."— Dmitri Dolgov to TechCrunch
Cruising: Driverless cars "will have every incentive to create havoc," warns transportation planner (Joe Romm — ThinkProgress)
- The big picture, per Stan Caldwell of Carnegie Mellon's Traffic21 Institute: Yes, cities need to consider policies to dissuade AV cruising to avoid parking fees, but taxis, ride-hailing and delivery trucks are part of the problem too.
- One solution would be for cities to redefine curbs and public parking garages from metered spaces to dynamically priced real estate, enabled by vehicle-to-infrastructure communications.
5. 1 body language thing
Researchers at the University of Michigan are studying human body language to teach self-driving cars to recognize and predict pedestrian movements with greater precision than current technologies.
Why it matters: People don't always pay attention when crossing the street, so AVs need to be on the lookout for distracted pedestrians, not just other cars on the road.
"If a pedestrian is playing with their phone, you know they're distracted. Their pose and where they're looking is telling you a lot about their level of attentiveness. It's also telling you a lot about what they're capable of doing next."— Ram Vasudevan, assistant professor of mechanical engineering, Michigan
Details: Using data collected by vehicles through cameras, lidar and GPS, the researchers captured video snippets of humans in motion and then recreated them in 3D computer simulation.
- This enabled them to create a "biomechanically inspired recurrent neural network" that catalogs human movements.
- By focusing on humans' gait, body symmetry and foot placement, they can predict what pedestrians might do next and train self-driving cars to recognize behavior.
Background: Until now, most machine learning for AVs has relied on still images.
- If you show a computer enough photos of a stop sign it will eventually come to recognize stop signs in the real world.
What's new: By using video clips that run for several seconds, Michigan's system can study the first half of the snippet to make its predictions, and then verify the accuracy with the second half.
- The researchers said they could predict a pedestrian's location within 10 centimeters after one second and less than 80 centimeters after 6 seconds. All other comparison methods were up to 7 meters off.
- "We're [now] better at figuring out where a person is going to be," says Matthew Johnson-Roberson, associate professor in Michigan's naval architecture and marine engineering department.