Combining Learning-Based Locomotion Policy With Model-Based Manipulation for Legged Mobile Manipulators | Synapse