Leveraging biomechanics to gain insights into the sensorimotor system
Alexander Mathis
Warren Hall room 205A and via Zoom
Efficient musculoskeletal simulators, and machine learning algorithms provide new computational approaches to tackle the grand challenge of understanding the sensorimotor system. First, I’ll talk about theory-driven approaches to test what the goal of the proprioceptive pathway is. Second, I’ll show that taking inspiration from sport science, we can train reinforcement learning algorithms to carry out skilled object manipulation tasks of the hand with 39 muscles. Interestingly, these models have a number of emergent properties that compare favorably to humans and challenge the notion of muscle synergies as a simplifying control principle.
—
To request the Zoom link email jteeters@berkeley.edu. Also indicate if you would like to be added to the mailing list for announcements about Redwood Seminars.