Model free and model based Reinforcement learning program from Simulink to Arduino

Hi, I have to use Reinforcement learning (RL) to control my hardware. I am using arduino due board with simulink interface to deploy the simulink program in arduino and run the setup.

My concern is, If I use Reinforcement learning to train model, can I run the final program on my arduino board to run setup?

Also, If I want to use modelfree approach for training the agent with hardware, is it possible with arduino interface?

Please give suggestions and methods for doing the above two tasks.

Answer 1

[aditi bagora]

Hi Aakash,

I understand that you have a reinforcement learning model that you want to train and deploy the model on the Arduino board.

You can utilize the "Reinforcement Learning Toolbox" in MATLAB to define an RL agent tailored to your hardware setup within a Simulink model. By simulating this model, you can effectively train the agent. Once trained, the model can be deployed on the Arduino board using the "Simulink Support Package for Arduino Hardware".

I recommend training the reinforcement model in Simulink before deploying it on the board rather than directly training on the hardware as Arduino boards, including the Arduino Due, have limited computational resources compared to a full computer. Running an RL algorithm directly on the board is generally impractical due to these constraints. Moreover, direct hardware interaction during training could lead to wear and tear or potential damage if the agent makes unsafe decisions.

Please refer to the following MathWorks documentation for more information:

• https://www.mathworks.com/help/reinforcement-learning/index.html?s_tid=CRUX_lftnav
• https://www.mathworks.com/help/reinforcement-learning/agents.html?s_tid=CRUX_lftnav
• https://www.mathworks.com/help/reinforcement-learning/ug/train-reinforcement-learning-policy-using-custom-training.html
• https://www.mathworks.com/help/simulink/supportpkg-arduino.html

Hope this helps!