I'm not an expert in Kalman, but isn't the use of latitude/longitude going to make the algorithm performance vary depending on where you are on earth? And also the two different scales in x/y axis makes weird to define a R value related to that coordinate system?
Wouldn't it make much more sense to do it with projected coordinates (e.g. MKMapPoints) and supply the R value in meters and convert it to MKMapPoints based on the latitude?