Zero-shot Imitation Learning from Demonstrations for Legged Robot Visual Navigation