Refraction AI's Matthew Johnson-Roberson on Robotic Delivery

Refraction AI: A Balanced Approach to Robotic Delivery

Refraction AI positions itself as offering a “just right” solution in the realm of robotic delivery. Based in Ann Arbor, Michigan, and recently expanding to Austin, Texas, the company secured a $4.2 million seed funding round. Its founders, two University of Michigan professors, believe that fully autonomous vehicles (AVs) are further from widespread implementation than often predicted, while sidewalk delivery presents its own set of limitations and reduced benefits.

Finding the Middle Ground: Bike Paths

The company’s REV-1 robot, initially showcased in 2019, is fundamentally built upon a bicycle frame. This three-wheeled vehicle, approximately 4 feet in height and 32 inches in width, can achieve speeds of up to 15 miles per hour. This intermediate speed allows for rapid obstacle avoidance while still exceeding human delivery speeds.

According to Refraction AI, the REV-1’s moderate speed reduces the need for long-range visibility, enabling the use of radar, sensors, and cameras instead of costly lidar systems.

From Academia to Application

Co-founder and CTO Matthew Johnson-Roberson brings nearly two decades of experience in academic robotics. He emphasizes the desire to translate advancements in field robotics into practical applications accessible to the public, a key motivator for transitioning from academia to entrepreneurship.

Evolution Since the TechCrunch Stage

TechCrunch: Reflecting on your unveiling at TechCrunch two years ago, how has Refraction AI progressed?

Matthew Johnson-Roberson: It’s been a period of significant growth. We’ve expanded from a single prototype to a fleet of 25 vehicles operating in Ann Arbor and Austin. The pandemic accelerated changes in the food delivery landscape, which we anticipated and adapted to.

We’ve focused on refining our vehicle production process and optimizing our pricing model. Initially, we offered direct-to-consumer delivery through our own app. However, we’ve since shifted to partnering directly with restaurants, functioning as a dedicated logistics provider rather than competing with services like DoorDash.

A Founder’s Journey

TechCrunch: Has your personal perspective evolved since launching Refraction AI?

Matthew Johnson-Roberson: Absolutely. This is my first venture as a company founder, despite my extensive background in autonomous vehicles. It was a steep learning curve, a true “fire hose” of new experiences. Both my co-founder, Ram Vasudevan, and I are University of Michigan professors, so building a company from the ground up was a unique and valuable process.

As we matured, we brought in experienced leadership, including a new CEO. The most challenging aspect was balancing company growth, fundraising, and navigating the pandemic’s uncertainties.

Leadership Transition: CEO to CTO

TechCrunch: You transitioned from CEO to CTO. What prompted this change?

Matthew Johnson-Roberson: We recognized the need to scale the company effectively, and that required expertise we initially lacked. Ram and I hadn’t previously managed significant company scaling. I wanted to concentrate on the technical challenges that genuinely excite me, rather than being consumed by investor meetings.

Focusing on the core technical problem – getting the robot from point A to point B – is paramount. Allocating my time to that is far more impactful than administrative tasks.

Technological Differentiation

TechCrunch: How has your technology evolved, and what sets you apart from competitors?

Matthew Johnson-Roberson: It’s surprising that few companies have attempted to replicate our approach. Many sidewalk robot companies emerged around 2015-2017, but most have since failed. We identified a gap: full-size AVs were projected to take longer to materialize, and both sidewalk robots and full-size AVs faced practical limitations for large-scale urban delivery.

Our REV-1 operates at a speed that presents challenges similar to full-size vehicles, but without the benefits of traveling in a dedicated traffic lane. We’ve had to develop unique solutions to address these challenges, forging our own path in the industry.

Unit Economics: The Goldilocks Zone

TechCrunch: How does your approach affect your unit economics?

Matthew Johnson-Roberson: The primary economic challenge with full-size vehicles is their high cost. The gig economy often shifts the depreciation costs of these expensive assets onto the delivery workers themselves. This model is difficult to apply to full-size autonomous vehicles.

Sidewalk robots, conversely, require a high delivery volume to compensate for their slow speed, necessitating extremely low prices that may not be sustainable. Our “Goldilocks” approach balances speed and cost. We achieve sufficient deliveries per hour while keeping vehicle costs manageable due to the lower speed and the ability to utilize less expensive sensors.

This allows us to offer competitive economics compared to gig economy workers, the primary competition in urban food delivery.

Navigating Regulations

TechCrunch: What is your strategy for gaining regulatory approval and encouraging adoption?

Matthew Johnson-Roberson: Surprisingly, regulatory hurdles have been less significant than anticipated. The jurisdictions we’ve operated in have been remarkably supportive, actively encouraging our expansion. It’s becoming increasingly easy to obtain permission to operate autonomous vehicles in many parts of the United States; the real challenge lies in the technology itself.

We prioritize being good neighbors, utilizing road margins responsibly and avoiding obstructions. Public acceptance has been positive, likely due to our avoidance of sidewalks.

Looking Ahead: Future Growth

TechCrunch: You recently closed a $4.2 million seed round. What are your plans for the next year?

Matthew Johnson-Roberson: We’re deploying funds to expand our Austin operations. We aim to launch in another location within the next year and significantly increase our vehicle fleet. We’re also developing our second-generation vehicle, focusing on reduced cost, improved reliability, and easier manufacturing.

Our goal is to establish a substantial delivery program in multiple U.S. cities. I believe that capital isn’t the primary limiting factor; the key is solving the technical challenges effectively. We’re optimistic that a focused approach will yield success, unlike the “throw money at the problem” strategy that hasn’t proven effective in the AV industry.