How I get my Q and R values of a Kalman filter for a good position estimation without only trial and error, are there ideas how choosing it, or methods?
Dear Robert,
I deduce that you are applying kalaman filter to a non linear system, and that you used a motor model, applied physical disturbance to approximate the noise with white Gaussian distribution to determine the initial values of Q and R in a trial and error methodology ..etc.
I have attached some links that will help you in IPMSM modeling and a good way to determine the Q and R. the 6th link is for autocovariance least squares package (estimation tool) that will help you determine noise covariance...
Best regards,
Khaled Rady
www.nt.ntnu.no/users/skoge/prost/proceedings/npc07/DTU/dtu10.pdf
https://www.researchgate.net/post/How_to_find_Q_and_R_in_Kalman_Filter2
ww.scirp.org/journal/PaperDownload.aspx?paperID=41240
elibrary.wiley.com/doi/10.1002/tee.22034/pdf
ieeexplore.ieee.org/xpl/login.jsp?tp=&arnumber=4658964&url=http:%2F%2Fieeexplore.ieee.org%2Fxpls%2Fabs_all.jsp%3Farnumber%3D4658964
jbrwww.che.wisc.edu/software/als/
Hi Robert,
I have dealt with this issue myself and have published in this area. You can refer to my publication linked below. This paper presents a method for automatically and efficiently finding the best values for your Q, R matrices in addition to other adaptive parameters if you are using adaptive Kalman filters.
Bests,
Afshin
Article Enhanced adaptive unscented Kalman filter for reaction wheels
An IPMSM machine is a deterministic system and not a stochastic that's why we try to built a stochastic model of the IPMSM machine based on a determenistic model.All theses quantities Q , R is not an easy task to detirmine the best values.The only way is by trial and error methodologyis .There exists a systematic method to determine the initial quantities of Q and R backward in time from a final error criteria,but it is very difficult to solve theses problems (theses problems are called difference equations of two boundry conditions).
Thank you very much. It is an honour for me that you have answered here.
I read your papers about Observability Analysis of the EEMF-Model, it is fascinating, but as we know, the theory is good, at the test bench it looks different ;)
Please see the reports arXiv:1503.04313v1.pdf, arXiv:1505.07201v1.pdf and arXiv:1505.07208v1.pdf. and the papers in SADHANA December 2016 issues.
Please see the latest intech open book Kalman Filters - Theory for Advanced Applications
Edited by Ginalber Luiz de Oliveira Serra,
Article Multi-site modeling and prediction of annual and monthly pre...
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Control Theory
- Control Systems