Achieving musculoskeletal motor control using reinforcement learning

Voiceover

Presenter(s)

Elyas Larfi

Abstract or Description

In this research project, we aim to train a musculoskeletal arm model to reach a target using deep reinforcement learning methods. The complexity of this task is defined by the multi-dimensionality of human arm joints. We explore solutions for this task using a simulated environment called MuJoCo and a musculoskeletal model provided by Myosuite. The primary purpose of this research is to examine how task demands and noise impact the choice of muscle coordination pattern, and compare primate neural recordings performing the task to the produced artificial neural network activity.

Mentor

Dr. Mazen Al Borno