172: How To Stay Focused on Your Mission and Deliver Robotic Solutions at Scale with Rapid Robotics CEO & Co-Founder Jordan Kretchmer

Engineering & simulation

The team developed 'few shot' learning capabilities that allow the system to infer object properties from just five pictures, enabling the robot to handle new objects without training specific models for each one.

“we're working on stuff, for example, that is few shot, take 5 pictures of an object, and we will infer everything about that object so that we don't even have to train models on any specific objects, and that's in the works, and things that just were not even possible three, four years ago.”

Engineering & simulation

AI enables the system to train on the concept of a 'box' rather than individual instances, allowing the robot to recognize and pick up any box regardless of color, label, or material by understanding its geometric properties and intent.

“What AI has enabled us to do is train our system on what is a box, right? And so it knows a box, has X amount of sides, it knows that it might have tape on it, might have some labels. It knows it could be brown, it could be white, it could be, you know, plastic, right? It is a box.”

Engineering & simulation

The AI-driven system uses sensing, force, torque, and camera inputs to automatically detect errors and resolve conflicts on the conveyor belt without customer intervention, making the technology accessible and simple.

“if something goes wrong, instead of the customer having to say something went wrong and send us a support ticket, the arm we should be able to tell through sensing, force, and torque, as well as the vision, the camera inputs, we know what went wrong. So the arm, because it's intent-based, knows what it needs to do in order to resolve that conflict on the conveyor belt or whatever happened.”