Reinforcement Learning of Human Movement

Treadmills are widely used in rehabilitation and gait analysis. However, previous studies have reported differences in terms of kinematics and kinetics between treadmill and overground walking due to physical and psychological factors. The aim of this study was to analyze gait differences due to only the physical factors of treadmill walking. The study has been published in Frontiers in Bioengineering and Biotechnology. - by Minki Jung

A human model and its gait controller were created. A patient model was created by limiting the hip joint torque of the human model. A gait assistive device with four degrees of freedom was attached to the patient model. The controller of the device was trained using a reinforcement learning method to help the patient model walk normally. - by Jonghyun Park

The skeletal model has 31 DOFs including the six DOFs in the root body. Ninety-two Hill-type muscles were attached to the lower limbs following a previous study (Rajagopal et al., 2016). The gait controller could provide excitation signals for the muscles to make it walk after a reinforcement learning to follow a reference motion (Peng et al., 2018). - by Youngjun Koo

  Walking is a result of complex process that involves neuronal control of muscles, musculoskeletal dynamics and interaction with environments. We investigate the neuronal controller of human walking using the deep reinforcement learning for the applications in orthopaedics, rehabilitation, and sports. Human gait study is closely related with bipedal robots and exoskeleton robots assisting walking. 

  보행과 같은 인체 운동은 신경제어, 근골격 동역학, 환경과의 상호 작용으로 이루어집니다. 운동을 위한 인체 근육의 신경제어 (nuero-musculo control)는 지난 한세기 동안 중추신경 제어, 반사신경 제어, Central Pattern Generator 등, 많은 이론과 가설들이 제시되었습니다(Dietz, Physiological Reviews, 1992). 본 연구실에서는 근골격 동역학 시뮬레이션 모델, 가상 운동 환경,  심층강화학습을 통하여 인체 운동의 신경 네트워크를 연구합니다. 이를 통하여 개인의 보행 특성을 반영하는 근골격 보행 제어기를 찾아내고, 다양한 가상 실험을 진행하는 방법을 개발합니다. 이는 근골격 및 관절 수술 기구의 개발, 수술 방법의 개선, 외골격 보조 장치의 개발 등에 활용될 수 있습니다.