Stochastic Optimal Control in Biology and Engineering

Control under uncertainty is a fundamental problem relevant to biology as well as engineering. Optimality models have explained numerous details of biological movements. Indeed optimal control and optimal (i.e. Bayesian) estimation are becoming the framework of choice for studying sensorimotor function. However most demonstrations of optimality are limited to relatively simple behaviors. In more complex and interesting behaviors we still lack the algorithms to compute what is optimal. Continued progress requires more efficient algorithms for stochastic optimal control.
In this University of Washington program, Emanuel Todorov, of MIT and UCSD, presents a new problem formulation that greatly simplifies the construction of optimal control laws, and yields original algorithms.

To see more videos from the University of Washington visit uwtv.org.

Category:

Education

Tags:

• computer
• science
• engineering

[aldoreshgaramok]

The chick at the beginning is cute

[gespilk]

I did not like this talk. It was somewhat chaotic. Was this a lecture or an overview of research results?

There is a saying, "A good plan, violently executed now, is better than a perfect plan next week". I believe it fits nicely when we talk about control in biological systems. It does not have to be optimal, just good enough.

The model I work with when it comes to biological systems is "Cells are agents!".

Todorov kind of talks about the transition from infant to child as control problem.

[damnitjimimonlyadr]

Great video.