Multi-Domain Walking with Reduced-Order Models of Locomotion