Delivery may carry the load for AV development

Editor’s note: A previous version of this story misstated Colin Langan’s place of employment.

Even with robots doing the driving, there's been an assumption since the dawn of automated vehicle technology that humans would be sitting somewhere in the car.

The first sketches of self-driving vehicles in the 1950s depicted chrome and tail fins, and families playing board games while gliding along electrified superhighways. Modern-era pioneers have conjured visions of people using apps to hail rides from robotaxi networks.

That's starting to change.

With innovations from interstate trucking to last-mile delivery robots, a new wave of development is focused on using autonomous driving technology to carry goods instead of people.

"The time for delivery robotics is here," said Reilly Brennan, general partner at Trucks Venture Capital in San Francisco. "If you can handle a handful of shopping bags at a decent speed, this is the time for that. It wouldn't surprise me to see some robotaxi startups shift to parcel and freight."

Companies such as Nuro and Refraction AI have formed in recent years with that task in mind. Now, with delays in deploying self-driving technology for passenger-carrying purposes and a global pandemic enhancing the prospect of delivery services for consumers seeking to avoid shared vehicles and physical stores, more companies are exploring the delivery market.

Ford has said it will postpone its launch of a commercial AV service that offers both passenger and package-delivery components so it can recalibrate those plans because of COVID.

"We can't ignore the change in consumer behavior we have seen during COVID-19," said John Rich, director of autonomous vehicle technologies for Ford Autonomous Vehicles. "It has impacted everything, from the way we work to how we shop. This change in consumer behavior, whether permanent or temporary, is something we must fully understand."

Businesses are only beginning to understand how consumers now view delivery. But experts agree the pandemic has brought new delivery opportunities. They come as automakers and tech companies examine potential operational, regulatory and safety advantages in using self-driving technology for goods delivery instead of for carrying passengers.

But delivery brings its own challenges, and it remains unclear whether delivery services are any more lucrative than the money-losing ride-hailing networks operated by Uber and Lyft.

"Just because the technology has an application doesn't mean it makes fiscal sense, and that's what this ultimately comes down to," said Ashley Nunes, a researcher who holds appointments at the Massachusetts Institute of Technology and Harvard University and has extensively studied transportation. "With packages, the value proposition in some ways is even worse because you have the added component of physically sorting and removing the packages from the vehicle. So these little bots are cute and cool, but the question is, 'Does it make sense?' "

Since his days as a participant in the Defense Advanced Research Projects Agency challenges — a series of government-sponsored AV competitions in the mid-2000s — Matthew Johnson-Roberson has sought to bring about the broad safety gains that self-driving technology has promised. He grew tired of waiting.

In 2017, Johnson-Roberson, a University of Michigan professor and leader of the Deep Robot Optical Perception Lab, founded Refraction AI, focused on food delivery via a vehicle akin to a tricycle with a cooler on top.

The prototypes travel in bike lanes at 12 to 15 mph and have a stopping distance of approximately 5 feet. With no humans aboard, "the worst thing that happens is your burrito gets smashed," he said.

Perhaps in a more nuanced fashion, the U.S. Department of Transportation would agree. In February, it granted Nuro's request for an exemption from certain federal motor vehicle safety standards written for an era when a human driver was always behind the wheel. The company now can deploy as many as 5,000 of its R2 electric delivery vehicles over the next two years without needing a traditional windshield, sideview mirrors or backup camera.

The exemption marked a milestone. It was the first one issued to a self-driving vehicle developer. In explaining their decision, officials said the company's prototypes were "at least as safe" as a vehicle that would conform to existing standards and held the prospect of leading to "safety, environmental and economic benefits" once deployed.

That a zero-occupant delivery vehicle received the first exemption, instead of a passenger-carrying AV, did not go unnoticed. Among some analysts and executives, it affirmed the idea that there's a lower barrier to entry for safety and operational reasons.

Beyond that, "ride comfort is not an issue for delivering a bag of potatoes, but sudden braking and vehicle hesitation will be unnerving for passengers," auto analyst Colin Langan wrote this month in a report on autonomous delivery titled "Will the First Scale Autonomous Vehicle Deliver Packages, Not People?".

According to Langan, delivery vehicles following pre-mapped, repeatable routes in geofenced zones were "a more compelling near-term opportunity" than AVs in which human passengers may desire a broad array of destinations.

Of the 3.2 trillion vehicle miles traveled on U.S. roads in 2018, 34 percent were related to commercial activity, he wrote. In addition, roughly 15 percent of the miles related to personal travel were for shopping, and that could shift to delivery in an autonomous era.

Langan estimated that delivery was a $1.6 trillion market today and ripe to increase. A report from the World Economic Forum published in January projects that e-commerce demand will increase the number of delivery vehicles in city centers 36 percent over the next decade.

In automated form, these vehicles could be shunted to off-peak schedules to help reduce urban congestion.

"That's the beautiful benefit to not being beholden to someone in the car," Johnson-Roberson said. "You're not bound to labor being tied to the vehicle."

If autonomous delivery vehicles can hit the road without concern for rider comfort, they also may be able to avoid scenarios that currently confound automated driving systems. Routes can be tailored around simplicity rather than a passenger's desire for the most direct or time-efficient path, which may involve tricky intersections or more congested streets.

"Circuitous routes are pretty annoying to human beings," Johnson-Roberson said. "But for delivery, you can pick routes that are not acceptable to humans."

Although taking less complex routes may be advantageous for the technology itself, Nunes warned that it comes with a trade-off in financial viability of delivery services.

"On an individual customer basis, that doesn't matter," Nunes said. "But from the firm's perspective, it does matter because it reduces productivity of the robots and the firm."

Optimizing the productivity and uptime of the vehicles is critical to success. It's one reason why Ford has plotted to use hybrid vehicles in its forthcoming commercial service instead of electric vehicles that would need hours to charge batteries. It's also why the company designed a purpose-built vehicle that can toggle between passenger-hauling and goods delivery, depending on peak travel times.

"To ensure our self-driving service is profitable, our vehicles will need to be running throughout the day, whether they are being used for ride-hailing or deliveries," said Rich, the Ford AV executive.

Further productivity concerns lie in the number of deliveries vehicles can make per trip. Nuro's prototypes contain two compartments so they can serve two customers per trip. Refraction's bots can hold a single order, but a second robot may be required for delivering larger orders. In Nunes' view, delivery robots need to carry high numbers of orders per trip, like the delivery vehicles used by UPS, Amazon and FedEx.

"Make no mistake, this makes money only if you leverage those economies of density and get as many packages on board as possible," Nunes said. "The only way this makes sense is the budget-carrier model: You cram as many people as you can get onto the plane."

For every challenge that a pivot to delivery seems to address, new ones await once those planes land, so to speak. Packages need to be sorted and extricated from AVs.

Currently, companies such as Ford, Nuro and Refraction AI expect a human to meet the vehicle and claim their order. That works for certain customers in certain locations, but not for all. Ford learned in its Miami pilot project, for example, that customers who live or work in high-rises are reluctant to meet vehicles at ground level.

Other questions abound: How do customers with disabilities claim deliveries? Might companies need customers to board vehicles and claim packages from assigned lockers? Are there liability concerns with that? If customers are not home, how do packages make it to their doorstep? Do consumers need next-generation mailboxes close to the curb? How long would food keep?

"You have three main challenges in this, and the last 100 feet is one of them," said Jeffrey Hannah, director of North America for global consulting firm SBD Automotive. "All those things still have to be worked out. And the delivery product has to have margin, and customers have to be willing to pay."

If that sounds like a cynical view of the fledgling efforts to automate delivery, he countered that none of the challenges is insurmountable. The most encouraging signal might be that COVID-19 has revealed strong consumer demand for delivery in ways not considered before the crisis.

"I'm extremely optimistic. I think this has radically exposed people to the idea of maybe never going into a retail establishment," Hannah said. "The value proposition is people's time. Even people on the fence have seen its value, and it can change retail and shopping patterns. Regardless of COVID, delivery is a hard trend."

Nuro is banking on Hannah being correct and that enough consumers find saving time in exchange for delivery fees a compelling argument. In turn, that could drive broader gains in efficiency and safety.

"Would it be safer for everyone, including ourselves, if instead of hopping in the car and racing across town to run errands, we have an autonomous vehicle do much of that for us?" asked David Estrada, chief legal and policy officer at Nuro. "Because the safest thing for you to do to avoid any road hazard is don't get on the road. So it's an entirely new way of looking at an autonomous vehicle."