Joint State and Dynamics Estimation With High-Gain Observers and Gaussian Process Models

With the rising complexity of dynamical systems generating ever more data, learning dynamics models appears as a promising alternative to physics-based modeling. However, data available from physical platforms may be noisy and not cover all state variables. Hence, it is necessary jointly perform estimation. In this letter, we propose interconnecting high-gain observer framework, specifically Gaussian process state-space model. The provides estimates, which serve for training updated model, in turn, used improve observer. Joint convergence model proved high enough gain, up measurement perturbations. Simultaneous estimation are demonstrated on simulations mass-spring-mass system.