Unable to compute kalman filter innovation (measurement residuals) in the new sensor fusion and tracking toolbox

I have recently installed the sensor fusion and tracking toolbox. Some of the IMU and GPS fusion examples are useful for my application. However there does not seem to be a straightforward method of computing EKF measurement residuals, or in general, accessing the process model and measurement models. are the underlying models not publicly accesible?

Answer 1

[Brian Fanous]

The code which computes the innovations and Kalman gain should be visible to you, but are not available through a public API right now. You can inspect the source code to see these variables.

Most of the code you are looking for is in the fusion.internal.IMUBasicEKF and fusion.internal.MARGGPSFuserBase class. These are internal classes so they may change in a future release. For R2018b the computations of the Kalman gain and innovations occurs in the correctEqn() method on line 72 of the fusion.internal.IMUBasicEKF class.

Can you give us an idea of what your application is and why you’d like to see this? That might give us an idea of how we can better support this in future releases.

[Suraj Bijjahalli]

Thank you very much. I found what i needed in the MARGGPSFuser class. I needed to compute the innovations to set up a sensor fault monitoring routine. Would be useful if the API was made public.

Answer 2

[Honglei Chen]

Are you using trakingEKF?

https://www.mathworks.com/help/fusion/ref/trackingekf-class.html

There is a residual method you can use

HTH

[Suraj Bijjahalli]

Thank you for your answer Honglei. Unfortunately, I am using the insfilter (https://au.mathworks.com/help/fusion/ref/marggpsfuser.html)

Is there a way of extracting measurement residuals and the Kalman gain computed at each update cycle for this filter?