kalman filter

All zeros is a pretty good choice, if it's a linear filter. Whatever make sense if it's an extended Kalman filter -- sometimes you have to use your noggin.

If you use an "information filter" then you track the inverse of the variance matrix, which will let you start with zero inforation -- that filter will automatically take care of initialization, but it can be more computationally intensive.

Hi! In kalman filter we will performing an updation. what is the need for the updation and when we will make an updation

Thanks, ilakkiya

--------------------------------------- Posted through

After you've done your web search, you may want to look at the Wikipedia article on Kalman filtering, which gives a good overview.

If you're trying to implement Kalman filters and you don't understand what you're doing, you're just engaging in a cargo cult, and aren't practicing engineering

I recommend Dan Simon's book, "Optimal State Estimation", and any books or courses you find you need to understand what he's talking about. It is vital, if you want your filter to do you any good at all, to understand the system model, to understand the noise model, to understand what _your_ costs are, and to understand how to modify _your_ implementation of a Kalman filter to work sensibly in your environment.