do I need to calculate covariance matrices?
Hi, I'm using robot_localization package and I searched to improve localization I want to calculate covariance matrices of my sensors, but I faced this info :
In general, it's good to have a pre-calculated covariances for your sensors, as it can reduce the computational load during runtime and improve the accuracy of the state estimates. However, if you don't have the covariances, the robot_localization package can estimate them during runtime.
So, I am not sure is it true or not? In that case, calculating covariances then using robot_localization, is it worth to try? Also I have A LOT cpu load, that's why which one is more efficient I didn't get it, my code will be an extra task for CPU because I am planning to calculate covariances continiously , but in the above comment said pre-calculated covariances reduce computational load?!