A quick understanding, not a deep dive. Automated Tire's platform is proprietary, so the specifics stay inside the company. What follows is a high-level summary of the kind of perception work I lead there. Happy to talk through my general approach in conversation.
I lead perception engineering for Automated Tire's SmartBay, a robotic platform that services tires without removing the wheel from the vehicle. That means owning the chain end to end: which sensors we use, the algorithms that read them, and getting both running reliably on the robot. You can see the platform in The Robot Report's introduction video.
What the work involves
- Evaluating and selecting depth and machine vision hardware, based on measurement against the application's needs rather than spec sheets
- Vision models that track a compressed tire's geometry during the tire change — hard, because rubber under load doesn't hold a shape that deterministic measurement can pin down
- Optimising perception to run in real time on the robot
- Connecting what perception sees back into how the robot works
Tools
- Depth cameras, stereo vision, camera calibration
- PyTorch, OpenCV, object detection
- ROS 2, Python, C++