Keep Covariance Matrix symmetric and positive definite while updating

Question

I am working on a Recursive least square estimate problem and realised that covariance matrix is not symmetric and positive definite after few iterations. How to ensure covariance matrix to be positive definite during update ? I tried reading few publications but unfortunately its above my skill set. I am not looking for a very optimised and computationally effective solution but some thing that is quick enough to test RLS algorithm while keeping covariance matrix positive definite.

Thanks

Answer 1

One can use Potter's Square Root Algorithm. Various implementations can be found in chapter 7, Grewal, Andrews:Kalman Filtering: Theory and Practice with MATLAB