Using a .mat dataset for a KinArm model

Rauf Iftikhar replied 1 year ago

Thanks for this answer. If I have multiple matrices stored in the .mat file, how will I know how to index into each specific one? How can I see the structure of the converted data?

Koloman Varady Staff replied 1 year ago

Since you create the binary format file, it’ll be up to you what the structure is. In the example, the binary file is just singles and they’re read one at a time per ms. You can have up to 5 files, so if you have rows you could put each in a file. If not, you could do column-major or row-major format.

Rauf Iftikhar replied 1 year ago

Currently I take a .mat file of 3 2 column matrixes and convert it into 3 corresponding binary .dat files with each value stored as a single (float32) so that I can output both values from each row in each byte. In my model I then unpack and convert. If I want the first column to be the xdat/breakpoints for a lookup table and the second column to be the ydat/table data, how can I achieve this? I tried to use a dynamic lookup table, but because the input file only reads a byte at a time we only pass a single value to the xdat and ydat at a time, which causes an error (“breakpoint and table port cannot be scalar”). I want to load the entire data set and then pass the respective vectors to a lookup table, not pass each individual value as it is read in. Is this possible?

Koloman Varady Staff replied 1 year ago

It depends on how big your matrices are. In the example task, the reading doesn’t start until the Go button is pressed in Dexterit-E (this is because it is an enabled subsystem that is taking a signal from Stateflow, which doesn’t start to execute until the Go button is pressed. If you remove that enable, the code will start to execute as soon the task is loaded (when you double click or select the task in Dexterit-E or when it finishes executing, it will also reload). So you could have the task load, wait 2 seconds, and have 6000 elements from your matrices loaded.