Repository for the project ML_Robot, a quadruped robot for the implementation of artificial intelligence.

• About
• Folders
  • Model
  • PCB
  • Raspberry Pi Pico
  • Controller
  • Unity Project
  • Results
• Build

This personal project aims to experiment with the principles of reinforcement learning* applied to robotics. After modeling the robot using fusion360, I 3d printed it. See assembly guide (WIP). Physical simulation of the robot inside Unity, and use of the ml-agents library to train a neural network to control the robot.

all 3D files modeled with fusion360 for 3D printing. (image)

gerber file, this printed circuit contains both the power circuit and the connections between the pins of the raspberry pi pico and the servo motors (DS3218) https://www.amazon.ca/dp/B085PVVZNQ

use of the raspberry pi pico because supports the tinyML library of tensorflow, multiprocessor which allows the use of one of the two cores for the execution of the neural network while the second takes care of performing the resulting actions. https://www.youtube.com/watch?v=aIFElaK14V4&ab_channel=LearnEmbeddedSystems (video from Learn Embedded Systems)

all the codes to control the robot using instructions via usb (in python) and the program in C and C++ to program the raspberry pi pico, using .uf2 to flash the memory (video)

the whole project under unity

Unity ml-agents -> https://github.com/Unity-Technologies/ml-agents

https://github.com/EbonGit/ML_Agent-Unity-Robot_Build (google colab .ipynb)

artificial intelligence models and related learning graphs

(WIP)