Adaptation, Optimality, and Synthesis

This research area focuses on the development of optimal control methods that can handle uncertainty, complex mission specifications, and rely on sophisticated approximation, learning, and sampling techniques to enhance scalability that is necessary for real-time deployments. Specifically, tasks address a wide range of challenges including the synthesis of approximately optimal controllers for hybrid dynamical systems, provide a number of approaches for enhancing the scalability of such synthesis methods under complex missions specifications captured by temporal logic (TL) formulas, and extend them to systems with unknown uncertainties and run-time computational limitations motivated by implications in other research topics. These design tools will be tailored in concert with nonsmooth systems research efforts so that the analysis methods are general enough to support design flexibility (e.g., allowing for non-strict Lyapunov functions to facilitate adaptive control methods). While the design tools avoid overly constraining the analysis (e.g., by avoiding an explicit discretization of the continuous part of the dynamics), they also account for the effects of adversarial conditions (e.g., design methods are proposed to account for not only the available data but also the contextual and physical knowledge about the systems of interest).


Adaptation, Optimality, and Synthesis Project Publications