Use Kalman Filter in location estimating

Question

I have been trying to understand Kalman filter and how to use it. I am planning to write it in java.

I have real time location (longtitude, latitude) and speed data. I need to find the next location of a moving object. Locations are accurate there is no noise in the location data. The reason that I want to use Kalman filter is to estimate the next probable location of the object. I couldn't understand how to give the values to the matrices(Transition, Measurement,etc).

I need your help to create and understand structure of matrices. I am also open for the recommendations to the new algorithms.

Answer 1

You could take a look at some open-source implementations. The ASF provides the following:

• http://commons.apache.org/proper/commons-math/apidocs/org/apache/commons/math3/filter/KalmanFilter.html
• http://commons.apache.org/proper/commons-math/userguide/filter.html#a17.2_Kalman_Filter

The following code illustrates how to perform the predict/correct cycle:

for (;;) {
   // predict the state estimate one time-step ahead
   // optionally provide some control input
   filter.predict();

   // obtain measurement vector z
   RealVector z = getMeasurement();

   // correct the state estimate with the latest measurement
   filter.correct(z);

   double[] stateEstimate = filter.getStateEstimation();
   // do something with it
}