Fall | Graduate | 12 Units | Prereq: 2.151
Provides a broad theoretical basis for system identification, estimation, and learning. Least squares estimation and its convergence properties, Kalman filter and extended Kalman filter, noise dynamics and system representation, function approximation theory, neural nets, radial basis functions, wavelets, Volterra expansions, informative data sets, persistent excitation, asymptotic variance, central limit theorems, model structure selection, system order estimate, maximum likelihood, unbiased estimates, Cramer-Rao lower bound, Kullback-Leibler information distance, Akaike's information criterion, experiment design, and model validation.
Fall 2020 Update: Fully Remote Classes - Class will meet at synchronous online lectures at the scheduled time, but for those who are unable to participate in the lectures, videos of all lectures will be posted. Furthermore, complete lecture notes and slides will be provided for every lecture. Individual or small group meetings to discuss assignments and projects will be scheduled. Personal attention will be given to each student. Emphasize cross-disciplinary studies between dynamic systems & control and machine learning. Broad topics will be covered, ranging from regression, statistical data analysis, Bayes and Kalman filters to parametric and non-parametric system identification, sub-space method, nonlinear function approximation, neural networks, deep learning, and Gaussian processes. This year also includes the Koopman Operator for exact linearization of nonlinear dynamics and its application to nonlinear Model Predictive Control (MPC). Powerful theories and useful techniques will be taught in the context of exciting applications. Students will work on Context-Oriented Mini Projects. Topics include 1) Active noise cancellation for wearable devices; 2) Self-driving cars based on Simultaneous Localization and Mapping (SLAM), 3) System identification of the cardiovascular system; and 4) Nonlinear MPC of robots using exact linearization. About 5 assignments and 2 quizzes will be given together with the Context-Oriented Mini Projects.
Ford Professor of Engineering; Director, d'Arbeloff Laboratory for Information Systems and Technology; Head, Control, Instrumentation, and Robotics