Keynote Speakers

Autonomous functioning via real-time monitoring and information management is an attractive ingredient in the design of any high-performance system. High-performance systems require design tools that not only deliver the requisite performance, but are able to function amidst various uncertainties, and are stable, safe, robust, resilient, and optimal. Uncertainties could be present due to various reasons including malfunctions, environmental variations, ageing, and modeling errors. Safety considerations often lead to constraints on states and outputs, introducing limitations that prevent against catastrophic events or damages that could lead to irrecoverable failure modes, and ensuring secure and resilient behavior. Stability, robustness, and optimality, the mainstays of feedback control systems, are necessary properties for any autonomous system. Adaptation together with learning are control methodologies that have been employed in order to achieve these properties as well as safety in the face of uncertainties. This talk will focus on the interplay between adaptation, learning, and safety while designing high-performance autonomous systems using feedback control.

The field of adaptive control has excelled at real-time control of systems with specific model structures through adaptive rules that learn the underlying parameters while providing strict analytical guarantees. These guarantees pertain to stability and robustness in the face of imperfect learning, and guarantees of optimality when parameter learning is feasible. The interplay between adaptation and learning has been explored at length over the years when no constraints are imposed on the input magnitude or on the states. This interplay becomes more complex when considerations of safety enter the picture, as the control inputs that are required to filter out unsafe control inputs even while adapting to the uncertain parameters may compromise performance considerations. A requisite adaptive control input that guarantees the desired performance may not always meet safety constraints. While an appropriate tradeoff between performance and safety may be realizable for a nominal system, the introduction of uncertainties makes it difficult to understand when and how the tradeoffs are realized. Thus, the central question that this talk is focused around is this: how can we leverage adaptation and parameter learning to ensure safety with minimal conservatism?

This talk will examine the interplay between adaptive control and safety. On the safety side, we will explore ways of integrating design principles from the fields of adaptive control and control barrier functions. We will then show how this integration leads to a controller which guarantees safety under parametric uncertainties, with conservatism that vanishes to near-zero with no requirements on excitation. The impact of learning on this interplay will also be examined, with a focus on achieving optimal performance when parameter learning is enabled. How reinforcement learning methods can be integrated into the overall control design will be explored as well, so that near-optimal policies can be realized for highly complex control tasks. Examples from flight control will be utilized to illustrate all of the above concepts and challenges.

At the frontier of autonomy lies a fundamental question: how can we design teams of agents that reason, act, learn, and memorize cooperatively in dynamic, uncertain environments? In this talk, I will present a unified perspective that integrates learning and control as the foundational principles for building such autonomous teams. In the first half of the talk, I will discuss how my group has advanced the ability of engineered systems to reason and act optimally under uncertainty, and to learn adaptively from interactions with their environment. I will illustrate these ideas through transportation-related applications, including self-learning powertrain control, power management of hybrid electric vehicles, and optimal coordination of connected and automated vehicles. These examples highlight how data-driven control and reinforcement learning can enable autonomous systems to operate safely and efficiently in real time while achieving near-optimal performance. The second half of the talk will focus on the next frontier—building autonomous teams. I will present a theoretical framework grounded in team theory, a mathematical formalism for decentralized stochastic control problems in which multiple agents with asymmetric information cooperate toward a shared objective. I will discuss recent structural results for sequential dynamic team problems with nonclassical information structures, showing how to construct information states that remain invariant to control strategies, thereby enabling dynamic programming decompositions in decentralized settings. These results can aim toward a unifying scientific foundation—what I refer to as the Science of Autonomous Team Intelligence—where teams of agents, whether robotic, vehicular, or human–machine, can reason, act, learn, and memorize collectively, achieving coherent and safe behavior in complex, uncertain environments.