https://github.com/igelbox/lamarr

Quadrupeds learn to conquer mankind
https://github.com/igelbox/lamarr

3d-printing keras machine-learning pybullet quadruped

Last synced: about 2 months ago
JSON representation

Quadrupeds learn to conquer mankind

• Host: GitHub
• URL: https://github.com/igelbox/lamarr
• Owner: igelbox
• License: mit
• Created: 2020-06-06T23:05:11.000Z (about 6 years ago)
• Default Branch: research
• Last Pushed: 2020-08-25T17:39:21.000Z (almost 6 years ago)
• Last Synced: 2025-03-06T13:47:25.569Z (over 1 year ago)
• Topics: 3d-printing, keras, machine-learning, pybullet, quadruped
• Language: Python
• Homepage:
• Size: 13.7 KB
• Stars: 1
• Watchers: 2
• Forks: 0
• Open Issues: 2
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

          
# Lamarr the Headcrab

## A Set of Tools and Algorithms to:

* Train [[train.py](0-train.py)] a virtual robot to walk in a simulated world using [PyBullet](https://pybullet.org), [Keras](https://keras.io/), and a custom genetic-like learning algorithm.

* Apply [[posess.py](3-posess.py)] the trained model to the real-world [3D-printed robot](https://www.thingiverse.com/thing:2901132) over Bluetooth via a simple binary Arduino-powered protocol.

**TL;DR** - just watch it once:

[![The Learning Process](https://i.ytimg.com/vi/UsYsJeZxmWI/hqdefault.jpg)](https://youtu.be/UsYsJeZxmWI)

[![Walking in The Real-World](https://i.ytimg.com/vi/3ROsGh2AWgw/hqdefault.jpg)](https://youtu.be/3ROsGh2AWgw)