From: m.alawieh on 22 Feb 2010 07:40
>I like to integrate GPS and INS using kalman filter to predict the
>position of a vehicle.
>first of all i like to use GPS sensor readings with kalman filter .
>I have read lot of research papers for that purpose but I donot know how
>to use real time data of GPS sensor in the kalman filter measurement
First question is what kind of GPS/INS integration are u using. Loosely
coupled, tightly coupled or ultra tightly coupled integration. The later
two integrations gives better results but your INS should be able to accept
GPS measurments (carrier and beat pahse, or I and Q in the ultra coupled
>My question is that where i have to put the readings from the gps sensor
>in measurement equation i.e where to put numerical values from the sensor
>in the measurement equation.how these values make sense.
This also follow the ingeration part, talking about loosely coupled. Your
measurement is basically:
Z = GPS postion - INS measurment
this occurs at a 1PPS rate ( the rate the gps position is measured).
To say this in a very simple way, your INS kalman predicts the position of
the rover at a rate f_s from 50Hz up to 2500Hz, and the gps postion is
received at f_GPS = 1 Hz. When the gps signal is there you feedback the
error corrections so ur INS kalman compensates the bias and other errors
that results from long term predicitions.
As Tim said, Synchronization in this case is something u have to worry
about when u need high precision and u have high dynamics in ur system.
>also i like to ask as how can i determine the state transition matrix F
>and observation matrix H for a particular application.
>also how can i determine the measrement noise matrix from the gps sensor.
>thanks and regards