Coco Robotics, a startup known for its fleet of last-mile delivery bots, is looking to get more information out of the five years’ worth of data its robots have collected. Its answer: a physical AI lab with University of California Los Angeles (UCLA) professor Bolei Zhou at the helm.

Coco Robotics, which made the announcement Tuesday, said Zhou has also joined the Los Angeles-based startup as chief AI scientist.

When the company launched in 2020, it used teleoperators to help the bots navigate obstacles on their delivery routes. Coco Robotics co-founder and CEO Zach Rash told TechCrunch the company’s goal has always been to operate its last-mile delivery robots autonomously to cut the overall costs of delivery. Now, Rash said the company has collected enough data to dive deeper into automation.

“We have millions of miles of data collected in the most complicated urban settings possible, and that data is incredibly important for training any sort of useful and reliable real-world AI systems,” Rash said. “We’re now at the point where we have sufficient data scale where I think we can start really accelerating a lot of the research happening around physical AI.”

The decision to tap Zhou to lead the effort was a “no brainer,” Rash said. Zhou’s research around computer vision and robotics has largely focused on micromobility, as opposed to full-scale vehicles, Rash said.

Coco Robotics was already collaborating with Zhou, too. Both Rash and his co-founder Brad Squicciarini are UCLA alums and have even donated one of their bots to the school’s research lab.

“[Zhou] is one of the leading researchers in the whole world on robot navigation, reinforcement learning, and a lot of the technologies and areas of research that are highly relevant for us,” Rash said. “He’s been already very capable of recruiting some of the top researchers in the world who he’s worked with in the past to come join Coco and help accelerate things on our end.”

This new research lab is separate from the collaboration the robotics startup has with OpenAI, which allows Coco Robotics to use OpenAI’s models while the AI research lab gets access to the company’s robot-collected data.

Coco Robotics plans to use the information and research it gathers from the lab for its own purposes for now. Rash said the company doesn’t have plans to sell the data to its peers.

Rather, it will be used for the company to improve its automation and efficiency, which will mainly pertain to the local models its robots run on. Rash said they also plan to share their research findings with the cities they operate in when applicable, to help fix obstacles and infrastructure that slows their bots down.

“Success for this lab really looks at us offering a higher-quality service at an extremely low price,” Rash said. “How do we get our costs lower? How do we make this much more affordable for businesses and customers? I think that’s going to create a tremendous amount of growth in this ecosystem.”

Topics