The problem

WRO Future Engineers is an open-ended engineering problem: build a robot that operates autonomously in a modelled environment, and document the engineering decisions behind it as carefully as the robot itself.

Approach

Layered autonomy. Perception with YOLO on the Pi, and both the Python and the Go code written to be testable off-robot. The Go work also compiles the programs, which makes them faster. The documentation treats the robot as a system rather than a script.

What I wanted to learn

Computer vision on constrained hardware, and how far a small model gets you before latency forces a different architecture.