Researchers have developed new reinforcement learning (RL) frameworks for quadrupedal locomotion, addressing the sim-to-real gap. The first approach, utilizing NVIDIA's Isaac Sim and Isaac Lab, achieves zero-shot sim-to-real transfer for whole-body control on a Unitree Go1, demonstrating robust recovery from disturbances and achieving linear velocities of 2.0 m/s. The second framework focuses on torque-driven RL for heavier quadrupeds, enabling them to traverse rough terrain and track desired velocities without explicit velocity observation, with simulation results on a Unitree B1 reaching 3.5 m/s. AI
IMPACT Advances in sim-to-real transfer for quadruped robots could accelerate development and deployment in real-world applications.
RANK_REASON Two arXiv papers detailing new research in reinforcement learning for quadruped locomotion using simulation tools.
AI-generated summary · Google Gemini · from 2 sources. How we write summaries →