Advances in Robust Control of Humanoid and Legged Robots

The field of robotics is moving towards developing more robust and efficient control systems for humanoid and legged robots. Recent research has focused on improving the robustness of motion policies in real-world environments, particularly in the presence of uncertainties and perturbations. Innovations in reinforcement learning and adversarial training have enabled the development of more robust and adaptive control systems. Additionally, there is a growing interest in optimizing energy efficiency in quadrupedal locomotion, with research highlighting the importance of gait parameters in reducing energy consumption. Noteworthy papers include:

  • Learning Robust Motion Skills via Critical Adversarial Attacks for Humanoid Robots, which proposes a novel robust adversarial training paradigm for enhancing motion robustness.
  • Physics-Informed Neural Networks with Unscented Kalman Filter for Sensorless Joint Torque Estimation in Humanoid Robots, which presents a framework for whole-body torque control without joint torque sensors.
  • Refining Motion for Peak Performance: Identifying Optimal Gait Parameters for Energy-Efficient Quadrupedal Bounding, which demonstrates the importance of optimizing gait parameters for energy-efficient quadrupedal locomotion.

Sources

Learning Robust Motion Skills via Critical Adversarial Attacks for Humanoid Robots

Multi-critic Learning for Whole-body End-effector Twist Tracking

Physics-Informed Neural Networks with Unscented Kalman Filter for Sensorless Joint Torque Estimation in Humanoid Robots

Robust RL Control for Bipedal Locomotion with Closed Kinematic Chains

Refining Motion for Peak Performance: Identifying Optimal Gait Parameters for Energy-Efficient Quadrupedal Bounding

Built with on top of