Y. Or

A Dynamics-Informed RL framework for a monopedal hopping quadcopter is proposed, embedding a target Specific Energy into the reward to constrain the optimization to a physically viable energy manifold, ensuring stable hopping behaviour.

Optimal gaits for swimming robots can be significantly enhanced by understanding the topological changes in joint angle trajectories, leading to improved displacement and energy efficiency.

Simpler models of robot body-environment interaction surprisingly outperform complex ones when training data is scarce.