-2
$\begingroup$

I am using L3GD20 and I am trying to implement a kalman filter for it on the stm32f3 discovery board. I have though a few questions about that:

  1. After the filter gave me the values, do I have to make an average between them and those of the original model or should I use them as they are?

  2. According to this document, we don't use the original state space vectors in the filter, so how could we have "correct" space state estimated values?

$\endgroup$
4
  • 1
    $\begingroup$ Your kalman filter gives you "multiple values"? Please elaborate. $\endgroup$
    – Paul
    Nov 25 '14 at 23:54
  • $\begingroup$ What i mean is that the filter give estimated values that could be wrong and the sensor gives noisy values.Which values do i have to trust? $\endgroup$ Nov 26 '14 at 15:49
  • 2
    $\begingroup$ If you doing things properly, the output of the kalman filter should be better than relying on sensor data or your state transition model alone. $\endgroup$
    – Paul
    Nov 26 '14 at 18:49
  • 1
    $\begingroup$ @Med.ali, if you apply Kalman filter correctly, then the estimated values are one of the best values you can go with. The whole point of using Kalman filter is to eliminate noise, so of course you need to trust Kalman filter more than your sensor. $\endgroup$
    – CroCo
    Nov 27 '14 at 19:16
3
$\begingroup$

To apply Kalman filter successfully, you need two requirements namely the system must be linear (i.e. both the motion and observation models) and the noise to be Gaussian with zero mean and some variance besides the models must be specified accurately. Kalman filter is a time domain recursive filter. Meeting these requirements, Kalman filter is one of the best filters to go, so you don't need to average the filtered data. This answers your first question. For the second question, I didn't read the attached pdf, Kalman filter estimates your state vector. I don't understand what do you mean by saying "we don't use the original state space vectors"?

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.