Explore the computational principles of sensorimotor control in this comprehensive lecture by Daniel Wolpert, part of the ICTP-ICTS Winter School on Quantitative Systems Biology. Delve into topics such as the complexity of human movement, optimal control theory, state estimation, motor prediction, and Bayesian decision theory. Examine various models of motor control, including the Kalman filter and forward models, and understand how they relate to real-world applications like eye movements and arm trajectories. Learn about different types of motor learning, impedance control, and loss functions in movement. Gain insights into the normative approach to human movement control and how it can be applied to reverse-engineer sensorimotor systems. This lecture provides a solid foundation for understanding the computational principles underlying how organisms sense the world and generate behaviors.
Computational Principles of Sensorimotor Control - Lecture 1
International Centre for Theoretical Sciences via YouTube
Overview
Syllabus
Complexity of human movement control
Modest success in robotics: Manipulation
Normative approach to human movement control
Reverse-engineering sensorimotor control
Motor planning
Arm movements: Paths
Eye movements: saccades
Models
The Assumption of Optimality
The ideal cost for goal-directed movement
Motor noise is signal-dependent
Signal-dependent noise and optimal control
Pointing movements: minimize variability
Motor control in the late
The demise of the desired trajectory
Motor control in the early
Optimal Feedback Control Todorov, Kappen
Optimal control and planning
State estimation Interpreting the uncertain state of the world
Generative model of state evolution
Kalman filter is the Bayesian estimator
Motor prediction with forward model
How is eye position estimated
Motor prediction
Types of Kalman estimation problems
Minimizing delays
Types of Motor Learning
Representations in motor learning
Mechanistic models
Normative models
Impedance
Measuring stiffness
Controlling stiffness
Bayesian Decision Theory
Sensorimotor learning and Bayes rule
Loss Functions in movement
Virtual pea shooter
Predictions
Loss function is robust to outliers
Imposed loss function
Summary
Taught by
International Centre for Theoretical Sciences
