Sam Earle

gym-city. An interface with micropolis for city-building agents, packaged as an OpenAI gym environment

control-pcgrl. Train or evolve controllable and diverse level-generators.

script-doctor. Code for PuzzleJAX, a benchmark for reasoning and learning, that reimplements PuzzleScript, a concise and expressive DSL and game engine for grid-based puzzle games, in JAX.

autoverse. Generative cellular automaton-like learning environments for RL.

pcgrl-jax. A JAX-accelerated implementation of the Procedural Content Generation via Reinforcement Learning (PCGRL) framework. We train RL agents to generate grid-based video game levels, optimizing for controllable functional heuristics, and study the out-of-distribution generalization (to new map shapes/sizes) of (multiple) agents with local observations.

picbreeder-vlm. Python

micro-rct. Python

pathfinding-nca. Experiments training neural cellular automata for path-finding.

ScriptDoctor. Automatic LLM-Driven Game Generation in PuzzleScript

pytorch-baselines-micropolis. PyTorch implementation of Advantage Actor Critic (A2C), Proximal Policy Optimization (PPO) and Scalable trust-region method for deep reinforcement learning using Kronecker-factored approximation (ACKTR).

gym-civ. An effort to wrap freeciv in SWIG/Cython, and use it as a learning environment for trainable agents.

particle-world. Simple particle simulation for join environment/agent optimization experiments.

PyTorch-NEAT. Python

micropolis-4bots. Small tweaks for reinforcement learning, or other city-building bots.

LegoTrainingRenderer. Real-time rendering of Lego-building agent in blender via python API.

cellyphus. Bringer of Froglings.

gym-city-slides. Slides from the talk "Fractal Automata" given at EXAG'19