Solving Real-World Factory Challenges with Custom AI

The other robots will just follow suit and we can do like really complex deployments very quickly. I can give you an example of where our robots are carrying their PCB magazines. And these magazines are like finished PCB boards. And they are worth lots of money, tens of thousands of dollars in each of the load. And they have to be very carefully transported from the soldering station to the finished goods. And one of the reasons they use the robot is it's clean room-like, so they don't want too many humans to be touching these magazines and risking the exposure of EMI. But what happened was because it's such a clean place, they keep vacuuming it. and the person who vacuums the floor, they could leave the vacuum cleaner on the floor and the cable is actually lying on the floor. The robot needs to go around the cable or at least stop and give an alert. It cannot go over the cord because just that jerk can render the whole payload useless. So we had to quickly come up with a custom AI model to see what black wires, black cables are like. and then you need to detect that and that had to get into our perception stack in short time because they were trying to commission it and they were trying to go live. So this is something that our team were able to do it very quickly.