Managing Entropy in Autonomous Logistics — It’s Software-First for Pickle Robot
It’s a busy time for logistics. E-commerce is booming, and warehouses can’t keep up with the demand. The challenges are daunting — tasks are labor-intensive, the pace is frantic, and staffing shortages are perennial. While the industry is overdue for smarter solutions, robot startup investments require careful evaluation. Should robots be fully autonomous or collaborate with humans? How many tasks — and what kind of tasks — should a robot handle? Which automated solution will provide the most productivity, while still being safe and easy to operate? What are the company’s unique economics and customer’s return on investment? Will the solution scale?
We were impressed with the answers to all these questions while evaluating our recent investment in Pickle Robot and its logistics robot, Dill. Toyota AI Ventures is joining Pickle’s $5.57M seed round alongside lead investor Hyperplane and others, including Box Group, Third Kind Venture Capital, and Version One Ventures. Based in Boston, Massachusetts, Pickle was founded in 2018 as a spin-out of LeafLabs by AJ Meyer (CEO), Ariana Eisenstein (CTO), and Dan Paluska (VP of Robotics).
Using a custom suction gripper and a robotic arm with extensive reach, Dill can unload and load packages from the back of a truck or trailer at an impressive rate of 1,600 packages per hour. The Pickle team is focusing on unloading because, while it’s exceptionally challenging for humans, it’s well-suited for a specially designed robot.
While some companies lean more heavily on hardware-based solutions, Pickle’s software-first approach uses machine learning to teach the robot to quickly adapt and improve its efficiency. That’s why Dill can successfully handle packages of many different shapes and sizes, whether they’re arranged in neat stacks (rare in our messy, entropy-rich universe), or randomly piled up in the back of a trailer as demonstrated in the video below.
In the rare case that Dill is unable to complete a task on its own, Dill can collaborate with people to get the job done. In a recent interview with IEEE Spectrum, AJ demonstrates his encyclopedic knowledge of the industry, explaining why automation can’t solve every obscure handling case on the loading dock, nor should it. For example, a robot gripper may not be physically capable of grabbing certain unique shapes, or a damaged package may fall apart. In those instances, it makes sense for machines and people to work together.
AJ believes that a better customer experience results from a focus on 98% of the cases, not wasting engineering resources chasing the 2% in the “long tail.” We agree. It’s not uncommon for robotics startups to attempt to solve too many problems perfectly, only to end up solving none of them very well. Too often, I hear founders claim their solution can solve everything for everyone. My rebuttal question is to ask them whether their solution can solve something for someone — ideally a market of paying customers. I’m happy to say that Pickle’s Dill can.
In fact, Pickle’s thoughtful approach makes us confident that Dill is the first step toward a scalable robotics warehouse solution. Dill does not require a custom-built warehouse — it can easily be retrofitted into existing warehouses of all sizes. The Pickle team ultimately envisions the future of warehouses where both robots and people collaborate to achieve optimal throughput. This resonates with our belief that technology is best when it amplifies the human experience.
Dill robots are sorting packages at customer warehouses right now. If you’re interested in the trailer unloading configuration, you can sign up for it starting in June, and those robots will begin shipping early next year. Visit the Pickle website to learn more.