How to implement Bayes filter / Kalman-filter like something in pymc?

[Dennis Evangelista]

Hello pymc list:

  I am a newish user of pymc... I've looked at the tutorials and I see where they are coming from but I'm still confused on how to implement something that I think may be of use to lots of people.  

  In the simplest case, I am tracking some animals in a video.  As each position measurement comes in, I would like to update the state of where I think the animal is and what he is doing (velocity, etc wise)... but I may have many animals and over a long time.  I can think of two ways to do this but I'm not sure if they are good or recommended uses of pymc. 

1.  (which should come out looking like a Kalman filter).  Each animal has a self.mu, self.C, and self.measC; at each time step I update these as follows:

def update(self,when,obs):

  state = pymc.MvNormalCov(mu=self.mu, C=self.C)