i am working recently on a project in which i want to implement a Kalman filter as being an observer, and i couple this observer with a state feedback controller that produces control actions necessary to control a mobile robot.
The problem is that the governing equations for the mobile robot's motion contain errors, and this will affect the control actions produced, and at the same time the miss modelling error covariance matrix is hard to get as well as the measurement error covariance matrix. I though about using some arbitrary values for these matrices and go on with the filtration process, but i thought that it would be much better if there is anything like a kalman filter formulation that has the ability to estimate closer values for system parameters so better control actions can be obtained.
Thank you in advance.