In spite of the recent advances in machine learning, the design of control systems capable of fully exploiting the potential of these methods to learn from, and interact with, the environment and safely achieve complex speciﬁcations remains elusive. Modern Machine Learning methods can leverage large amounts of data to learn powerful predictive models, but such models are not designed to operate in a closed-loop environment. Recent results on reinforcement learning oﬀer a tantalizing view of the potential of a rapprochement between control and learning, but so far proofs of performance and safety are restricted to limited cases.

Led by Dr. Mario Sznaier from Northeastern University, the AI and Machine Learning Control section welcomes submissions focused on bridging the gap between control systems and machine learning techniques.

Topics considered in the scope of this section include:

convergence rates and robustness against adversarial attacks in control theoretic techniques
handling different scales of data availability and informativity
machine learning inspired approaches to learning and controlling dynamical systems, including reinforcement learning
robustness of closed loop systems incorporating machine learning elements in the loop
sample complexity of learning dynamical systems
understanding the fundamental limits of machine learning when used to learn and control dynamical systems
verification of global closed loop properties of systems when using controllers learned from limited, local data

Theoretical contributions are expected to be mathematically rigorous and advance the state of the art at the interface of Control and Machine Learning. Practical contributions are expected to describe in detail an example of application of Machine Learning to a real-world, challenging control problem, or the use of Control tools to improve Machine Learning algorithms. Papers using simulations to illustrate the use of existing Machine Learning techniques to control problems are discouraged, unless they provide new theoretical insights and performance certiﬁcates supporting the use of the proposed algorithms.

The section welcomes submissions which support and advance SDG 9: Industry, Innovation and Infrastructure.

This multidisciplinary section is at the forefront of disseminating and communicating cutting-edge scientific knowledge and impactful discoveries in the field of control engineering and machine learning to researchers, industry, policymakers, and the public worldwide.

Formal Verification of Neural Networks-Based Control Architecture for Safety-Critical Autonomous Systems