Deploy classification leaner model into my C++ arduino code

I am using arduino hardware to collect vibration acceleration from two stepper motor. One stepper motor is healthy the other is faulty which means I have a healthy and fault data. I used these data to do predictive maintenance where I follow the work flow of predictive maintenace with Diagnostic Feature designer where the end point is i train the model in the classifcation learner. Now for the last step is how do I deploy the classifcation learner model into my C++ arduino code to prove it works?

Answer 1

[Meet]

Deploying a classification model trained in MATLAB into an Arduino board can be achieved using MATLAB Coder. MATLAB Coder generates C++ code from MATLAB code, which can be compiled and deployed onto the Arduino board.

To deploy the classification model trained in the Classification Learner app, follow the steps below:

1. Open the MATLAB code that implements the classification model in the Classification Learner app.

2. Export the model to the workspace by clicking "Export Model" in the "Classification" tab.

3. In the MATLAB command window, use the "codegen" function to generate C++ code from the MATLAB code that implements the classification model. For example:

cfg = coder.config('dll');
cfg.TargetLang='C++';
cfg.GenerateReport = true;
codegen -config cfg -args {X} predictModel

In the above example, "X" represents the input data to the model and "predictModel" is the name of the generated C++ function that implements the classification model.

4. In the generated C++ code, replace the input and output data types with the appropriate data types for the Arduino board. This can be done by modifying the function signature in the header file generated by MATLAB Coder.

5. Compile the generated C++ code into a library that can be linked with the Arduino code.

6. In the Arduino code, include the header file generated by MATLAB Coder and link the generated library.

7. Call the "predictModel" function in the Arduino code to classify new data using the deployed model.

Note that deploying a machine learning model onto an Arduino board can be challenging due to the limited memory and processing power of the board. It's important to ensure that the model is optimized for deployment on the board and that the board has enough resources to run the model.

For more details on code generation using codegen, please refer Generate C/C++ code from MATLAB code - MATLAB codegen (mathworks.com).