Learning to walk in confined spaces using 3D representation