Bridging the Gap in Prediction, Estimation, and Control of Integrated Systems

Abstract

Integrated systems are ubiquitous as more heterogeneous physical entities are combined to form functional platforms. New and “invisible” feedback loops and couplings are introduced with increased connectivity, leading to emerging dynamics and making the integrated systems more control-intensive. The multi-physics, multi-time scale, and distributed-actuation natures of integrated systems present new challenges for modeling and control. Understanding their operating environments, achieving sustained high performance, and incorporating rich but incomplete data also motivate the development of novel design tools and frameworks.

In this talk, I will use the integrated thermal and power management of connected and automated vehicles (CAVs) as an example to illustrate the challenges in the prediction, estimation, and control of integrated systems in the era of rapid advances in AI and data-driven control. While first-principle-based modeling is still essential in understanding and exploiting the underlying physics of the integrated systems, model-based control and optimization have to be used in a much richer context to deal with the emerging dynamics and inevitable uncertainties. For CAVs, we will show how model-based design, complemented by data-driven approaches, can lead to control and optimization solutions with a significant impact on energy efficiency and operational reliability, in addition to safety and accessibility.