This thesis introduces a physically consistent Gaussian process (GP) framework for the data-driven modeling of uncertain Euler-Lagrange (EL) systems, fulfilling critical physical properties. Multivariate Bayesian model error bounds are also derived, enabling precise uncertainty quantification and tight exponential stability guarantees. Applications to uncertainty-adaptive, structure-preserving tracking control and disturbance observation are given in numerical simulations and physical experiments.
«
This thesis introduces a physically consistent Gaussian process (GP) framework for the data-driven modeling of uncertain Euler-Lagrange (EL) systems, fulfilling critical physical properties. Multivariate Bayesian model error bounds are also derived, enabling precise uncertainty quantification and tight exponential stability guarantees. Applications to uncertainty-adaptive, structure-preserving tracking control and disturbance observation are given in numerical simulations and physical experiment...
»