A company in Ann Arbor, Mich., is testing the hypothesis that full-sized autonomous vehicles are not the answer to making short-distance food or grocery deliveries, and is instead betting on small electric vehicles.
Groceries, take-out dinners and other items could soon arrive via an autonomous robot in one Michigan city.
An Ann Arbor-based company called Refraction AI has been testing its REV-1 autonomous delivery vehicles, which are about the size of a large wheelbarrow or cargo bike. The three-wheeled electric devices are designed to operate in bike lanes, or other areas at “the margins” of the roadway in the aim to stay out of the primary traffic flow.
Their cargo area, capable of carrying up to 100 pounds, is ideal for items like grocery bags, restaurant deliveries or other goods traveling short distances — a half-mile to two and a half miles, said Matthew Johnson-Roberson, co-founder and CEO of Refraction.
“It’s much closer to things that would be couriered or brought by Uber Eats, or Grub Hub, or something like that,” he explained.
Five of the vehicles are already built, with plans to scale up to serve a wider area of the city. Refraction is currently testing the devices by delivering to its own employees, as well as the employees affiliated with two restaurants the company partnered with to supply deliveries.
The company plans to branch out into regular delivery service later this year.
“We look at the next 18 months to expand to other cities,” said Johnson-Roberson, who is also an associate professor of engineering at the University of Michigan. “We’re looking at larger cities and also college towns as well. And then we’re hoping to expand to sort of more cities beyond that, past that 18-month horizon.”
The vehicles — which are about five feet tall, four and a half feet long and roughly 30 inches wide — cost about $5,000 each. Their design was driven by a desire to travel at relatively slow speeds — 15 mph or so — and stop quickly. The REV-1 can come to a halt in about five feet.
Researchers at Refraction decided to build a device that would closely mimic the speed and weight of someone on a bicycle.
“The risk profile for that is something, we thought, the technology where it is today with autonomous vehicles, is really ready to handle,” said Johnson-Roberson.
“A lot of the hard problems with autonomous vehicles, around predicting human behavior, predicting other drivers, all these things become a lot easier, if you can just stop,” he added.
Another important design feature was that the vehicle be small enough so that if it needs to stop, it’s not obstructing regular traffic.
The rise of e-commerce and delivery services like Grub Hub have contributed to the escalating levels of traffic congestion in downtowns. Numerous cities are experimenting with technology or policy changes to better manage curb space. For example, Washington, D.C., is experimenting with the company curbFlow to smooth out the flow of goods and services on street curbs on nine city blocks.
“Our goal here is not to be the dominant road user that’s getting in everyone else’s way,” said Johnson-Roberson.
Devices like REV-1 can help cities achieve several goals, Johnson-Roberson said, pointing to sustainability gains when a full-size delivery car is replaced by a small electric bike-like vehicle. But they also ease congestion issues given that the delivery AVs are small enough to not require a whole travel lane to operate.
“We think that there are real delivery problems that need to be handled in a half mile or two miles, and we think that autonomous technologies can help with that,” said Johnson-Roberson. “And in the same way, the right solution for every problem is not a full-size car. And the right solution for every problem is also not a full-size autonomous car.”
And scaling up to a fleet of REV-1s, stationed at several parts of the city, is doable since their slow speed makes the AV technology — in this case, namely a camera — significantly cheaper in cost than a full-size autonomous vehicle.
“Really the win that we’re seeing is that by going slower, we can utilize a lot of the technology that has gotten really good and really cheap, and we don’t need to invest a ton into these other pieces of things that are sort of super expensive,” said Johnson-Roberson.