Title: Mathematical modeling and LMI-based control design for applications in the context of electrical microgrids

Abstract: Renewable generation systems have experienced a fast development in the last decades, especially considering the continuous increase in energy demand, the need to reduce the emission of greenhouse gases, and the growing interest in distributed energy systems. In this context, photovoltaic (PV) systems have been consolidated as an attractive alternative to conventional fossil fuel generation due to their high efficiency, modularity, flexible design, ability to generate energy on-site, low maintenance, and increasingly attractive cost. At the same time, the microgrid concept has become popular. A microgrid is an electrical network of power generators and loads that can be grid-connected (connected to an available electrical grid) or isolated (as a local grid).

This talk explores two applications of control design in the context of microgrids: 1) The robust static output-feedback control of a continuous-time uncertain linear model for grid-connected converters with LCL filter based on two-stages technique, aiming to optimize a quadratic performance criterion (LQR), allowing decentralization and magnitude constraints for the gain entries (suitable for practical implementations); and 2) The gain-scheduled dynamic output-feedback control of a continuous-time linear parameter varying (LPV) model for a microinverter-based distributed power generation system, aiming to improve transient response by incorporating D-stability and H∞ criteria into the synthesis conditions.

The design conditions used in both examples are developed regarding Linear Matrix Inequalities. Furthermore, the experiments consider real-world operational characteristics, providing a comprehensive evaluation of LMI-based approaches.


Title: Safety Assurances for Learning-Enabled Autonomous Systems

Abstract: The ability of machine learning techniques to leverage data and process rich perceptual inputs (e.g., vision) makes them highly appealing for use in autonomous systems. However, the inclusion of data and machine learning in the control loop poses an important challenge: how can we guarantee the safety of such systems?

To address these safety challenges, we present a controller synthesis technique based on the computation of reachable sets, using optimal control and game theory. We present new methods that leverage advances in physics-informed neural networks to compute reachable sets and safety controllers efficiently. These techniques are highly scalable to high-dimensional systems, enabling the learning of safe controllers for a wide array of autonomous systems. Furthermore, these methods allow us to quickly update safety assurances online, as new environment information is obtained during deployment. In the second part of the talk, we will present a toolbox of methods that use data-driven reachable sets to stress-test learning and vision-based controllers. By identifying safety-critical failures, these tools guide performance improvement while maintaining safety.

Together these advances establish a continual safety assurance framework for learning-enabled autonomous systems, where safety considerations are integrated across various stages of the learning process, from initial design to deployment and ongoing system enhancement. Throughout the talk, we will illustrate these methods on various safety-critical autonomous systems, including autonomous aircrafts, autonomous driving, and quadcopters.