The updated final report that includes an abstract is in now uploaded!
The Github repository is set up in the following way:
Comment 1 - Arduino Code that runs forever reading sensor inputs and sending to the Raspberry Pi through Serial Connection
Comment 2 - DQN RL policies trained from pytorch that were used to run on the raspberry pi
Comment 3 - Main file used to run all of the DQN RL training, sends data into neptune repository (may need to add in a database api key to run this locally) If wanting to just run the training code, the cartpole_RL.ipynb should be able to run for the most part.
Comment 4 - Plots found in paper can be found here as well.
Comment 5 - Please install libraries needed to run scripts.
Comment 6 - Code run on raspberry pi to step through arduino data, read in the pytorch DQN RL policy model, and use action from that model to run the motor.
Videos are included in the folder Videos_Hardware, and can be watched by downloading the video (the Partway_success_180_steps.mp4 is the shortest one, 7 seconds)
Photos of the Hardware are included in the folder named Photos_Hardware, and three main descriptor photos can be found below:
