Fast Data Driven Control

--finished project(2023-2024)

here's a preliminary poster and here is the full paper submitted to L4DC! Also here is a link to our preliminary paper published at NeurIPS workshop on Gaussian Processes, Spatiotemporal Modeling, and Decision-making Systems!

In brief, we've devised an exceptionally fast, data-driven approach utilizing random Fourier features to control nonlinear input systems, prioritizing safety and stability. Remarkably, our technique matches the accuracy of complex Gaussian Process-based kernel methods. What sets us apart is the linear time and memory complexity, a stark contrast to the cubic and quadratic complexities in terms of data points.