GymSlug: Deep Reinforcement Learning Toward Bio-inspired Control Based on Aplysia californica Feeding

Review Engineering, Biomedical

Deep reinforcement learning for modeling human locomotion control in neuromechanical simulation

Seungmoon Song et al.

Summary: Modeling human motor control has been a grand scientific challenge, with researchers exploring neuromechanical simulations and reinforcement learning to predict human movement in novel environments. While plausible motor control models have explained basic locomotion behaviors, challenges remain in modeling higher layer controls. Recent advances in deep reinforcement learning show potential for addressing these challenges, but application in neuromechanical simulations for human control remains limited. The Learn to Move competition and software platform aim to accelerate the use of reinforcement learning in this field and have attracted global participation and innovative solutions. Future opportunities lie at the intersection of human movement simulation and reinforcement learning, with plans to extend collaborative efforts in modeling human motor control.

JOURNAL OF NEUROENGINEERING AND REHABILITATION (2021)

Add to Collection

Proceedings Paper Automation & Control Systems

Reinforcement Learning for Robust Parameterized Locomotion Control of Bipedal Robots

Zhongyu Li et al.

Summary: Researchers developed robust walking controllers using a model-free reinforcement learning framework for bipedal robots to train in simulation and transfer to a real robot, using domain randomization to facilitate sim-to-real transfer.

2021 IEEE INTERNATIONAL CONFERENCE ON ROBOTICS AND AUTOMATION (ICRA 2021) (2021)

Add to Collection

Article Neurosciences