Explore a cutting-edge approach to system identification in this hour-long lecture from the Alan Turing Institute. Delve into the challenges of identifying ordinary differential equations (ODEs) in nonlinear, stochastic systems with noisy measurements. Learn about a novel method that combines physics-informed neural networks with Kalman filter techniques to accurately determine parameters in continuous-time systems. Discover how this approach leverages existing system knowledge to create more precise models, even for complex systems like double pendulums. Gain insights into the importance of robust system identification for controller design and see how this innovative technique overcomes the limitations of standard optimization algorithms.
Overview
Syllabus
Tobias Heinrich Nagel - Kalman Bucy informed Neural Networks for System Identification
Taught by
Alan Turing Institute