Extrapolation 2-D lookup table

Hi, I'm having the following problem. I have a 2d table that contains data from some measurements. How can I do the same kind of Extrapolation that is possible in SIMULINK 2-D table lookup using interpolation-extrapolation lookup method, but in Matlab. As I figured out 'griddata' and 'interp2' can not do the job for me. This is the dimension of my data:

x [1x37]

y [1x28]

z [28x37]

Thanks in advance.

[Jette]

I came across the same question as jose tapia. I also could not find anything in the FEX. Doesn't anyone have an idea how this is done in Simulink?

[Kaustubha Govind]

Jette: Jose's question was about being able to do something in MATLAB that is already possible in Simulink using the Lookup Table blocks. It appears that you want to do something in Simulink itself. Please start a new question if one of the Lookup Table blocks is not sufficient for your purpose.

[Jette]

I do want to do it in MATLAB and get rid of Simulink models which basically only use lookup tables. I meant the algorithm or the underlying function Simulink uses.

[Titus Edelhofer]

Hi Jette, nice to meet you again ;-). Titus

Answer 1

[Matt Fig]

I don't know what kind of extrapolation is done in Simulink, but have you looked on the FEX? This might help, for example:

INTERPNE

[jose tapia]

Hi, Matt. Thanks for your answer.

INTERPNE doesn't work for me, because the data I have doesn't fit with the function.

Answer 2

[Titus Edelhofer]

Hi Jose,

in a German forum I gave a description on how the interpolation of Simulink works, see here, although I don't have a working solution. If you care for how the extrapolation works, I would translate here.

Titus

[jose tapia]

Thank's for your help. This is very useful.

[Ilham Hardy]

Hi Titus,

I know it is four years ago, but would you please describe how the 2D-LUT in Simulink works? This time in English?

I check the link you gave and translate it, but still I don't get it yet. The solution post by Jette is not complete (I mean ZI_1 is still matrix).

Thanks a lot, Ilham

Answer 3

[Mehmet Özdemir]

Hi everyone,

I was struggling with the same problem for some time and finally found a very easy solution. Save the date.

INPUT PARAMETERS

x = Column indicator of the matrix

y = Raw indicator of the matrix

v = Look up table matrix

xq, yq = Point where interpolation takes place

THE CODE

vq1 = interp1(x,v',xq,'linear','extrap') % This calculation gives an array which are the interpolated values (wrt xq) of every raw. PS : Don't forget to take the transpose of v matrix.

vq = interp1(y,vq1,yq,'linear','extrap') % This part interpolates the new calculated vq1 for the value yq and gives the final result

Since it is possible to extrapolate by adding ///,'linear','extrap'/// in this method, your task of extrapolating 2d maps will be accomplished. Good by NaNs.

Best

[hang x]

It's a nice way to solve this problem,thanks for sharing it