(732b) Keynote Talk: Towards the Automated Optimal Control of Colloidal Self-Assembly

Controlling colloidal self-assembly with external field-mediated non-equilibrium interferences is a viable approach to manufacturing desired structures, for advanced applications in energy harvest, drug delivery, and photonics devices. However, the scaleup and automation of this process is hindered by factors ranging from the identification of the system states to learning the system dynamics, and to the design of the optimal control policy. In this talk, I will focus on an electric field-mediated colloidal self-assembly system, to first talk about our previous efforts on how data can be used to learn the system dynamics in designing an optimal control policy to rapidly produce ordered colloidal crystals, and then discuss ongoing work on how we leverage advanced machine learning approaches to explore the automation of learning the optimal feedback control policy, especially on the state representation and policy design with a continuous action space.