Learning Semantics-Aware Locomotion Skills from Human Demonstration

Yuxiang Yang, Xiangyun Meng, Wenhao Yu, Tingnan Zhang, Jie Tan and Byron Boots

CoRL 2022, New Zealand


Our framework adapts locomotion skills (speed and gait) based on perceived terrain semantics.

Our framework completes an outdoor trail and uses different skill on different terrains.

Additional Results on Different Terrains

Walking on Rock (0.48m/s)

Walking on Mulch (0.52m/s)

Walking on Grass (0.64m/s)

Walking on Gravel (0.76m/s)

Walking on Asphalt (1.3m/s)