As a simple example, I would like to implement a Kalman-Filter in a level 2 m-file s-function. Because of this, I need a covariance matrix P, which will be updated at every time step (so, the matrix of the last timestep is also needed). It would be self-evident to use a DWork variable for this. But the DWork variable cannot be defined as a matrix. Do you have any ideas how a matrix can be implemented in a s-function as a DWork-like structure?
DWork-data are vectors. You will need "reshape" function.
Be careful when using Dwork vector for saving 2D-array after reshaping it into vector. Because if you choose the Dwork vector length to be equal to the total number of elements in the 2D-array. Next, you reshape the 2D-array into vector. After that, you save the "vectorized" 2D-array into the Dwork vector. Then for unknown reason, the Dwork vector will save only half of the 2D-array elements. This will be the case no matter what you try!
But fortunately, after long effort of trial and error, I found solution for this problem. Simply by choosing the length of the Dwork vector to be double the total number of elements in the 2D-array. By doing so, half of the Dwork vector will hold all of the elements of the "vectorized" 2D-array. While the other half will hold useless values (they will be the same values you initialized the Dwork vector with in the start function).
Final note, Matlab built-in reshape function will not help you much if you choose my previously explained method. Instead, use for-loops to move back and forth between the Dwork vector and the 2D-array.
- Badges
- Science topic