June 25, 2026
Description
Important Notes This is an advanced electronics/mechanics project, not a simple toy print.
If you need help, or need the .ino files etc, join our discord: https://discord.gg/fp8gyYxzmE
Please test carefully:
- calibrate every servo before walking
- power the servos from a proper high-current power rail
- do not power the servos through the Arduino or small jumper wires
- test with the robot lifted before putting it on the ground
- start with small movements
Parts I used:
A small note about this project:
I know that building such a large quadruped with regular servos does not really make sense from a practical engineering perspective. The servos are limited, the mechanics are not ideal, and there are much better ways to build a serious robot dog.
But that was not really the point of this project.
I wanted to learn more about robotics, motion control, reinforcement learning, and how AI can be used to train robots. This build gave me a lot of hands-on experience with mechanical design, electronics, servo control, walking patterns, and simulation-based training.
For example, I also experimented with training the robot in Godot. This time, I used a different approach: first, I manually designed a walking animation, as shown in video 1. After that, I used PPO training to let the AI improve the movement based on that animation. In video 2, you can see a rendering from the training process.
So even if this robot is not the most practical design, it taught me a lot. My next robot project will combine everything I learned from this build, the training experiments, and all the mistakes along the way. Also for training I will switch to Gazebo…
The robot is open source, so feel free to improve it, remix it, and share your progress. Follow me on TikTok if you like my work!
License:
Standard Digital File License - Community Use