Learning Agile Soccer Skills for a Bipedal Robot using Deep Reinforcement Learning
Deep Reinforcement Learning was used to train a bipedal robot to exhibit robust and dynamic movement skills, including walking, turning, kicking, and fall recovery, and to combine these skills in a smooth and efficient manner to play a simplified one-versus-one soccer game.
Humanoid-Gym: A Reinforcement Learning Framework for Humanoid Robots with Zero-Shot Sim-to-Real Transfer
Humanoid-Gym is an open-source reinforcement learning framework designed to train locomotion skills for humanoid robots, enabling zero-shot transfer from simulation to the real-world environment.
Continual Learning Framework for Versatile Soft Robotic In-Hand Manipulation
A Continual Policy Distillation (CPD) framework is introduced to acquire a versatile controller for in-hand manipulation of objects with varying shapes and sizes within a four-fingered soft gripper.
Active Exploration in Bayesian Model-based Reinforcement Learning Improves Sample Efficiency for Robot Manipulation Tasks
Leveraging Bayesian neural network models and active exploration strategies can significantly improve the sample efficiency of model-based reinforcement learning for robot manipulation tasks, outperforming model-free and reactive exploration approaches.
Accelerating Torque-Based Legged Locomotion Policies with Decaying Action Priors
A two-stage framework that leverages position-based imitation data and decaying action priors to accelerate the training of torque-based legged locomotion policies, enabling consistent convergence to high-quality gaits.