Interesting question. InsyAllah I will explain in detail soon

I am intereseted to Mr Zol Bahri Razali, waiting for your answer.

Sincerely,

Thank you Mr  Zol Bahri Razali, i am waiting.

Hello. You question is a little unclear. Do you mean that you have several Kalman filters with different state vectors, e.g. a constant-velocity and a constant-acceleration one? If this is the case, then the IMM can be written down for it. You have to be careful when switching between states of different dimensions since it can cause transients. If this is not what you are talking about, please clarify your question.

The links below may be useful to you (the first discusses the IMM and the second the problem of changing state dimensions). There is a close parallel with manoeuvring target tracking which is covered in the first link, where you often have to change from 4-D state to 6-D state (in a planar scenario).

Technical Report A Survey of Manoeuvring Target Tracking Methods

Conference Paper A Gaussian Mixture Filter for Near-Far Object Tracking

Excellent links, Graham Pulford!

Thank you Mr Graham Pulford.

Yes, i have two Kalman filters with two state vectors of a constant-velocity and a constant-acceleration. 

Could any one of you help me to implement an IMM scheme to control objectives like observer based control and thanks in advance ?

At the beginning ,I used the second order Kalman filter (state vector of a constant-velocity nx=4) but it did not appropriate for tracking a  manoeuvring target . So i used the third order Kalman filter (state vector of a constant-acceleration nx=6) to mitigate this problem, whereas it did not sufficiently smooth like the first one in the non-manoeuvring fractions of the target trajectory.

Now, i want to utilize the two models using the IMM for tracking in the manoeuvre and non-manoeuvre fractions of the trajectory target

see

Article Systematic approach to IMM mixing for unequal dimension states

A Gaussian Mixture Filter for Near-Far Object Tracking