I am new to particle filtering. I can see when particle filters are used with Ensemble Kalman's, the velocity of the states are taken care of by the Kalman.

When tracking using particle filters only with position-only measurements, how do you get the velocity for propagating the particle by the transition function ? In Kalman filters, the filters smooth through the difference in position measurements. I don't see that in particle filters. Wondering if you provide velocity as a pseudomeasurement by differencing successive measurements ?

Thanks for any help

  • $\begingroup$ Well, I kind of figured it out by implementing it. PFs kind of random walk into the right velocity/acceleration states even though they don't have explicit measurements. The weight update equations take care of it, as the gain in Kalmans do. $\endgroup$
    – jschoe
    Jul 16, 2019 at 23:54
  • $\begingroup$ I am exactly in the same dilema. Is it because you are resampling the particles of high likelihood at every step? and when you resample, you give them random perturbations until converging to the true values? $\endgroup$ Mar 3, 2021 at 0:24


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.