This repo contains the implementation of dynamic iterative PCA proposed for identification of a system where the input and output measurements are corrupted with Gaussian white noise. The code for slightly different version of DIPCA algorithm is available at this link. Feel free to check out some of the other closely related and latest works on DIPCA algorithm from my webpage.
Please refer the demo_dipca.m, readme.txt files for usage of the dipca function. You can also type the following command to get the documentation in Matlab:
- doc dipca_rot_ref
- help dipca_rot_ref
The algorithm is capable of estimating the following from the noisy data:
- Input-output orders and delay of the transfer function
- Input-output noise variances
- The coefficients of difference equation
If you are using this code, please cite the following papers:
Identification of Linear Dynamic Systems using Dynamic Iterative Principal Component Analysis, Deepak Maurya, Arun K. Tangirala, Shankar Narasimhan, 2016, IFAC-PapersOnLine, 49(7), pp.1014-19.[Paper]Identification of Errors-in-Variables models using dynamic iterative principal component analysis, Deepak Maurya, Arun K. Tangirala, Shankar Narasimhan, 2018, Industrial & Engineering Chemistry Research, 57(35), pp.11939-54. [Paper]
The above papers discuss the identification algorithm. You may refer our following paper if you are interested in performing residual analysis for any EIV system:
Optimal Filtering and Residual Analysis in Errors-in-variables Model Identification, Vipul Mann, Deepak Maurya, Arun K. Tangirala, Shankar Narasimhan, 2020, Industrial & Engineering Chemistry Research, 59(5), pp.1953-65. [Paper]
There are several other works which extend the use of DIPCA algorithm for classical systems (non-EIV) with OE and ARX model, MISO systems:
ARX Model Identification using Generalized Spectral Decomposition, Deepak Maurya, Arun K. Tangirala, Shankar Narasimhan, To appear in 24th International Symposium on Mathematical Theory of Networks and Systems (MTNS 2020) [arXiv]Identification of Output-Error (OE) Models using Generalized Spectral Decomposition, Deepak Maurya, Arun K. Tangirala, Shankar Narasimhan, In Fifth Indian Control Conference (ICC 2019) (pp. 28-33). IEEE. (Won the Best Student Paper Award) [Paper]Identification of MISO Systems in Minimal Realization Form, Chaithanya K. Donda, Deepak Maurya, Arun K. Tangirala, Shankar Narasimhan, To appear in Sixth Automatic Control and Dynamical Optimization Society (ACODS 2020), IFAC-PapersOnLine, 53(1), pp.141-146., [Paper]
An updated list of related works and their code can be accessed from my webpage.