The paper presents a new version of the point mass filter, which is based on Rao-Blackwell theorem and greatly increases computational efficiency of the estimation algorithm. Due to its reliability and robustness it is appropriate for implementation in safety critical applications. THe algorithm is a result of the joint work with dr. Vaclav Šmidl from University fo West Bohemia, Czech Republic and was presented at the international conference on information fusion FUSION 2013. Based on our result, we managed to establish contact with the leading research groups on the topic of state estimation (University of Linkoping, Sweden).
B.03 Paper at an international scientific conference
COBISS.SI-ID: 27016743In Accelerator Driven Systems, high availability of the accelerator is one of its key requirements. Fortunately, not every beam trip is necessarily a failure. For example, in the proposed MYRRHA transmuter, absence of the beam for less than 3 seconds is still deemed acceptable. Predictive diagnostics strives to predict where a failure is likely to occur, so that a mitigating action can be taken in a more controlled manner, thus preventing failure of other components while exactly pinpointing the component that is about to fail. One approach to predictive diagnostics is to analyze process variables that quantify inputs and outputs of components as archived by the accelerator's distributed control system. By observing trends in their values an impending fault can be predicted. In addition, sensors measuring e.g., vibration, temperature or noise can be attached to critical components.
B.03 Paper at an international scientific conference
COBISS.SI-ID: 27015719Gaussian process (GP) models are nowadays considered among the standard tools in modern control system engineering. They are routinely used for model-based control, time-series prediction, modelling and estimation in engineering applications. While the underlying theory is completely in line with the principles of Bayesian inference, in practice this property is lost due to approximation steps in the GP inference. In this paper we propose a novel inference algorithm for GP models, which relies on adaptive importance sampling strategy to numerically evaluate the intractable marginalization over the hyperparameters. This is required in the case of broad-peaked or multi-modal posterior distribution of the hyperparameters where the point approximations turn out to be insufficient. The benefits of the algorithm are that is retains the Bayesian nature of the inference, has sufficient convergence properties, relatively low computational load and does not require heavy prior knowledge due to its adaptive nature. All the key advantages are demonstrated in practice using numerical examples. Ihe significance of the result lies in the fact that it was published at the leading event of the automatic control scientific comunity and promotes the department of Systems and Control as well as the Jožef Stefan Institute in the domain.
B.03 Paper at an international scientific conference
COBISS.SI-ID: 28011559