Multiple Gait
Multiple gait research focuses on enabling robots, particularly legged robots, to execute diverse locomotion patterns adaptable to various terrains and tasks. Current efforts concentrate on developing robust control algorithms, often employing deep reinforcement learning, model predictive control, or hybrid approaches, sometimes incorporating vision-based perception and bio-inspired design principles. These advancements aim to improve robot agility, efficiency, and robustness in challenging environments, with implications for applications ranging from search and rescue to industrial automation. The field is also exploring the underlying biomechanics of multiple gaits in animals to inform more efficient and natural robot locomotion.
Papers
CROSS-GAiT: Cross-Attention-Based Multimodal Representation Fusion for Parametric Gait Adaptation in Complex Terrains
Gershom Seneviratne, Kasun Weerakoon, Mohamed Elnoor, Vignesh Rajgopal, Harshavarthan Varatharajan, Mohamed Khalid M Jaffar, Jason Pusey, Dinesh Manocha
Achieving Stable High-Speed Locomotion for Humanoid Robots with Deep Reinforcement Learning
Xinming Zhang, Xianghui Wang, Lerong Zhang, Guodong Guo, Xiaoyu Shen, Wei Zhang