Refraction AI in Ann Arbor Launches Last-mile Delivery Robot

Ann Arbor’s Refraction AI, which created the REV-1, a low-cost, lightweight autonomous delivery robot that can operate in bike lanes and on roads, has launched. Founded by University of Michigan professors, the company has developed the solution for last-mile logistics. The company is backed by eLab Ventures and Trucks Venture Capital.
1979
Refraction AI's REV-1
Refraction AI has launched REV-1, a low-cost, lightweight autonomous delivery robot. // Photo courtesy of Refraction AI

Ann Arbor’s Refraction AI, which created the REV-1, a low-cost, lightweight autonomous delivery robot that can operate in bike lanes and on roads, has launched. Founded by University of Michigan professors, the company has developed the solution for last-mile logistics. The company is backed by eLab Ventures and Trucks Venture Capital.

“We have created the Goldilocks of autonomous vehicles in terms of size and shape,” says Matthew Johnson-Roberson, co-founder and CEO at Refraction as well as a professor at U-M. “Our platform is lightweight, nimble, and fast enough to operate in the bike lane and on the roadway, and we are tackling regional inclement weather patterns that inhibit or slow down other AV solutions.”

The robot has three wheels and is about 5 feet tall, 4.5 feet long, and 30 inches wide. It weighs about 100 pounds and can reach speeds of up to 15 mph. The inside of the vehicle holds 16 cubic feet, or about four to five grocery bags.

When a delivery arrives at its destination, a text with a keypad code lets recipients retrieve their goods. The company’s first test application is with restaurant partners.

“Consumers today expect on-demand goods of every type, and timeliness of delivery is often the key to customer satisfaction. But companies are struggling to find consistent, reliable, and economical ways to address that need,” says Bob Stefanski, managing director of eLab Ventures. “Refraction’s use of sturdy, smaller-sized delivery robots in the bike lane allows for faster technology development and covers a larger service area than competitors operating on the sidewalk. Their vehicles are also lightweight enough to deploy more safely than a self-driving car or large robot. The market is huge, especially in densely populated areas.”

The robot uses 12 cameras as its primary sensor system, along with radar and ultrasound sensors for additional safety. The system allows the platform to work in the rain or snow and is not dependent on traditional high-definition LIDAR maps.

“Our vehicle’s low curb weight at low speeds makes deployment safer than other autonomous vehicles. For example, we have a 5-foot stopping distance, compared to the typical 45-foot stopping distance that a full-sized vehicle at the same speed would need to avoid an accident,” says Johnson-Roberson. “Finally, our design and technical choices, particularly relying on cameras over HD-LIDAR, allow us to operate a more economical platform that gives us a significant competitive advantage on cost efficiency.”

Ram Vasudevan is the other co-founder and U-M professor.