I had the pleasure to attend the inaugural Conference on Learning for Dynamics and Control at MIT. The goal of this new conference is to bring together researchers at the interface between machine learning, control, optimization, and related areas, with a focus on addressing scientific and application challenges in real-time physical processes modeled by dynamical or control systems. The conference format is still in flux, but this year’s version was a single-track format with diverse high-profile speakers presenting on a range of topics including: model-predictive control, deep learning, simulating dynamical systems, robotics, traffic systems, and more. There was also two excellent poster sessions. Overall, the conference was a huge success. Over 400 people attended, and all on very short notice! Only about 3 months elapsed between when the idea of this conference was conceived and when the conference actually occurred.
My two highlights of the conference were: (1) Getting to see all the Berkeley folks I knew through my postdoc but haven’t had much of a chance to interact with since then. Many any of them are now moving on to exciting careers in academia and industry. (2) Having the opportunity to meet Alexandre Megretski for the first time and chat about robust control and IQCs. Megretski wrote some of the seminal work in these areas and it was an honor to meet him in person.
Kudos to the organizers (Ali Jadbabaie, George Pappas, Pablo Parrilo, Ben Recht, and Melanie Zellinger) for putting together such an amazing conference on such short notice! I’m looking forward to attending again next year!