The lab is where the research side of my work lives: perception and probabilistic reasoning on real robots, tested in simulation before they ever run on hardware. It sits outside delivery timelines, so an approach can be measured honestly rather than shipped early.

Current focus is active perception for humanoid robotics: particle filter methods that let a ROBOTIS OP3 reason about where a tracked ball probably is after it leaves view, and recover it faster than a purely reactive controller would. Everything here is built to be broken, instrumented, and rebuilt until the behaviour is reliable rather than merely demonstrable.