Using Reinforcement Learning to Build a Tree

Even though it has been a tough year for many, throughout, I have been inspired by my conversations with people, as well as their publications and contributions to the AI/ML and AGI community, in all aspects, technical and humanitarian. As the year comes to an end, I thought I’d develop a digital Christmas card with a bunch of Reinforcement Learning agents building Christmas trees using a PPO Algorithm with Unitys ML Agents.

WebApp, please try:

AI Learns to Build a Tree

Branch Building School

There are two so called schools or agent academies. The first one is initiating random tree trunks and so the agents can learn how to pick up and attach branches onto the trunk.

Trunk Building School

The second school is an environment where the agent has to pick up trunk pieces and stack trunk pieces to create the trunk of the tree.

1000 Steps
3000 Steps
5000 Steps