Ludvig Ericson

Though robotics and artificial intelligence has come a long way over the years, a lot of work remains until the uprising can finally begin. The first order of business is actual spatial understanding and reasoning.

Spatial understanding is a key capability in autonomous robotic systems. Everyday constructions such as walls, floors, and ceilings still by and large elude robotic understanding. My work seeks to remedy this by using so-called spatial priors, in the context of common robotics problems, such as exploration, search-and-rescue, and in the longer term, social robotics. These spatial priors are typically encoded as deep generative models, such as GANs or VAEs.