This code is based in the work of Prof. Mauricio A. Alvarez and Prof. Neil D. Lawrence SheffieldML (some modification were made such that Gaussian Processes with dimension greater than 2 can be run) and the work of Prof. Dr. Juš Kocijan. This is an implementation of dynamic system identification with multi-output Gaussian Processes. Both dependencies need to be installed before running the code. An Underactuated Ship model was used to generate data for system identification. The system is a two input , four output system. A NARX architecture was used for the Multi-output Gaussian Processes as for the comparative Neural Networks.
If you want to cite this work please cite:
Ariza R., W., Leong, Z.Q., Nguyen, H. and Jayasinghe, S.G., 2018. Non-parametric dynamic system identification of ships using multi-output Gaussian Processes. Ocean Engineering, 166, pp.26-36.
Prediction from Multi_output GPs by algorithm of Naive Simulation with full data compared to mathematical model, a) controlled surge acceleration, b) induced sway speed, c) controlled yaw speed, and d) induced roll speed
To run the System idetification, please run the file: MultiOutputGPSSI.m
Wilmer Ariza Ramirez
Australian Maritime College, University of Tasmania, Newnham TAS 7248, Australia