#reinforcement learning
Oct 2026
Gait-parameterized RL for locomotion across morphologies and scales without reward retuning
Sungho Lee, Jungyeong Kim, Sangshin Park et al.
· Intelligent Service Robotics · 0 citations