README.md

The original repository of Self-Balancing-Robot

Implementation of Self Balancing Robot with ML-Agents.

The RolyPoly model is a Newly designed model to reduce complexity in the environment, and I will use this model to implement a balancing robot in the real world.

RolyPoly Model

Requirements

• ml-agents >= 1.4.0

Usage

Training on Simulation

1. import your asset files into the unity workspace.
2. build the RolyPoly scenes and train with default option.

Testing on Simulation

1. import your asset files into the unity workspace.
2. select the RolyPolyModeltest scenes and start the game.

Testing on Real

1. import your asset files into the unity workspace.
2. select the RolyPolyRealtest scenes and start the game (please don't forget to upload the firmware on the ardino and connect to arduino)

Results

training result

Testing Video

References

Kalmanfilter original codes for arduino in HERE

Kalmanfilter original codes for unity in HERE

PID original code in HERE

Citation

If you use this code in your work, please cite our work

@inproceedings{Self-Balancing2021,
    author={Hyeok Yoon and Ji-Hyeong Han},
    title={Training and Implementation of Self-Balancing Robot with ML-Agents},
    booktitle={Proceedings of the 35th Institute of Control, Robotics and Systems Annual Conference (ICROS 2020)},
    pages={448-449},
    year={2021},
}