How to implement interpolation using Prelookup using a MATLAB function?

I have data with size 1201 26 37 384

I want to interpolate the first three dimenssions to get the fourth F(dim1, dim2, dim3) = column with size 348.

I'm using interpolation Using Prelookup and Prelookup to interpolate the first three dimentions and get the output is the fourth diminsion.

How can I implement this using matlab function instead of the simulink blocks?

I have tried using `griddedInterpolant` inside MATLAB funciton, but seems Simulink does not like it.

Is there any ideas?

Answer 1

[Chuguang Pan]

@Amr. interpn can be used for N-D interpolation (table lookup).

Data = rand(1201,26,37,384);
[X1,X2,X3] = ndgrid(linspace(0,2,1201),linspace(0,2,26),linspace(0,2,37));
[X1q,X2q,X3q] = ndgrid(linspace(1,2,50));

Data_query=arrayfun(@(idx) interpn(X1,X2,X3,Data(:,:,:,idx),X1q,X2q,X3q,'linear'),1:384,"UniformOutput",false);
Data_output = cat(4,Data_query{:});
size(Data_output)

ans = 1×4

50 50 50 384

<mw-icon class=""></mw-icon>

<mw-icon class=""></mw-icon>

[Amr]

Thanks for your reply.

I think this approach can not be implemented in Simulink function as passing X1 ... X3 to the matlab function and doing the interpolation will make the actual simulation time slower. I have tried griddedInterpolant which was fast in MATLAB scripts, but it is not supported in Simulink functions.

[Walter Roberson]

@Amr

I am confused about whether your target is implementation in Simulink (seems like it from your comments); or if your target is outside Simulink (as requested explicitly in your initial question) ?

[Amr]

Sorry for confusion.

I wanted to implement it in a Matlab function inside Simulink.

when I said that How can I implement this using matlab function instead of the simulink blocks? I meant to use the matlab function inside simulink instead of using Simulink interpolation Using Prelookup and Prelookup blocks.

I hope I have clarified it.

Answer 2

[Walter Roberson]

Convert your 4D data into a 3D array together with an (1201*26*37) x 384 array. The 3D array values should be column indices into the N x 384 array, and "nearest" interpolation should be done (in order to ensure that the outputs are indeed column indices.)

So, interpolated F(dim1, dim2, dim3) will return column indices into the N x 384 array, and you then use the returned column indices to extract the appropriate 384 values.