Whether you are watching a launch site catching a starship booster on reentry or sitting in a self-driving car that is changing lanes, the technology behind these actions is automatic control. Control is a discipline at the intersection of engineering, math and physics that provides the tools to make machines behave the way you want.
At the core of a successful automatic control design is a good physical-mathematical description of the system you want to control. This description is usually hard to get, and even if you have it, it is unlikely that you will be able to immediately design a controller for it.
This talk will highlight recent efforts by the control systems community to automate control design using data. It will describe the fundamental principles behind an end-to-end process that, after collecting experimental data from the system to be controlled, encodes the data into an optimization program whose solution yields a provably correct control algorithm.
The talk will discuss what is left to be done and how these results fit in a world where data and machine learning are taking over.