# GNS/INS (Inertial Navigation System) integration

Tags:
1. Aug 3, 2015

### ramesses

Hi
I'm working on a GPS/INS model.

Now I developped the mecanization equations for the INS.

I have a kalman filter which estimate the error state (position error with respect to north and east).

it takes as an input the diffrence between GPS and INS and gives its estimation .

I have developped a gauss-markov process which very similar to sensor noise (after applying wavlet).

Now, I don't know what can I do during the GPS outage ?

Because the mesaures are not available during this periode; I can't use Kalman filter.

Documentation that I have, doesn't say clearly what to do during GPS outage.

2. Aug 3, 2015

### Staff: Mentor

I googled Inertial Navigation During GPS Outage, and got some pretty good hits. Maybe have a look through some of them to see if they give you some ideas...

3. Aug 3, 2015

### olivermsun

During the GPS outage your GPS fix has infinite, or perhaps some really large, variance. So what happens to your Kalman filter when that happens?

4. Aug 4, 2015

### ramesses

@berkeman
Thank you :)
Kalman filter doesn't work until a new GPS's measure comes. unless the Kalman filter can't be used

Last edited: Aug 4, 2015
5. Aug 4, 2015

### olivermsun

Doesn't your Kalman filter have at least one input which is the integrating model from your INS?

6. Aug 5, 2015

### ramesses

Hello;
the system starts its loop after receiving the initiale position, bearing and speed from the GPS receiver.
this loop, presented in the first figure, is used during GPS outage.

Now, when a new measure comes from a the GPS receiver, the distance between the GPS and INS positions will be used as input for the kalman filter.

the measurement equation, as in [1],is

I followed the same model as ([1]), except I used Gauss-Markov process instead of autoregression.

The vector in the right is called the error state vector. It is used to estime the INS error.
αuN and αvN are the parameters of the 4 ordre autoregression process for the 2 accelerometer axes.

If I have well understand, the use of the process parameters inside this matrix serve for the prediction step. it gives an estimation for the evolution of the error for the epoch t+(1/INS_frequency)
but after giving its feedback, the error state vector will reset to zeros

What do you think about it ?

Sorry for my english

[1]Park, Minha, and Yang Gao. "Error and performance analysis of MEMS-based inertial sensors with a low-cost GPS receiver." Sensors 8.4 (2008): 2240-2261.