Controls and Control Theory Commons^™

Data-Driven Stochastic Optimal Control Using Hilbert Space Embeddings Of Distributions, Adam J. Thorpe

Electrical and Computer Engineering ETDs

Autonomous systems are increasingly being deployed in complex environments subject to real-world uncertainty. For such systems, it may be exceptionally difficult or even impossible to compute a simple mathematical model of the system--for instance due to the presence of human elements, complex mechanics or system dynamics, or learning-enabled components. Data-driven control has recently gained significant attention in this area, where observations taken from the system evolution are used to compute an implicit representation of the system that is amenable to analysis and control. However, data-driven algorithms for control present new challenges, and require new insights to enable their use. The …