Sparse and hybrid modelling of relative humidity: the Krsˇ ko basin case study

This study describes an application of hybrid modelling for an atmospheric variable in the Krsko basin.The hybrid model is a combination of a physics-based and data-driven model and has some properties of both modelling approaches.In the authors’case,it is used for the modelling of an atmospheric variable,namely relative humidity in a particular location for the purpose of using the predictions of the model as an input to the air-pollution-dispersion model for radiation exposure.The presented hybrid model is a combination of a physics-based atmospherical model and a Gaussian-process(GP)regression model.The GP model is a probabilistic kernel method that also enables evaluation of prediction confidence.The problem of poor scalability of GP modelling was solved using sparse GP modelling;in particular,the fully independent training conditional method was used.Two different approaches to dataset selection for empirical model training were used and multiple-step-ahead predictions for different horizons were assessed.It is shown in this study that the accuracy of the predicted relative humidity in the Krsko basin improved when using hybrid models over using the physics-based model alone and that predictions for a considerable length of horizon can be used.

Dynamic System Identification of Underwater Vehicles Using Multi-output Gaussian Processes

Non-parametric system identification with Gaussian processes for underwater vehicles is explored in this research with the purpose of modelling autonomous underwater vehicle(AUV) dynamics with a low amount of data. Multi-output Gaussian processes and their aptitude for modelling the dynamic system of an underactuated AUV without losing the relationships between tied outputs are used. The simulation of a first-principle model of a Remus 100 AUV is employed to capture data for the training and validation of the multi-output Gaussian processes. The metric and required procedure to carry out multi-output Gaussian processes for AUV with 6 degrees of freedom(DoF) is also shown in this paper. Multi-output Gaussian processes compared with the popular technique of recurrent neural network show that multi-output Gaussian processes manage to surpass RNN for non-parametric dynamic system identification in underwater vehicles with highly coupled DoF with the added benefit of providing the measurement of confidence.