Mobile manipulation leverages both the navigation capabilities of a mobile base and the manipulation capabilities of a robot arm. While some household tasks can be handled with navigation and manipulation decoupled, many realistic tasks like opening cabinets or moving through doorways require coordinated arm and base motion, referred to as whole body control. To avoid having to handcraft control policies, learning from demonstrations provides a scalable approach that can generalize across different task variations and environmental conditions. In this project, I want to use visual observations from expert demonstrations to train policies that achieve whole body control for mobile manipulation.