RushHour is an environment for evaluating agents on spatial planning and constraint-based puzzle solving. This environment wraps the RushHour implementation from TextArena, a framework for text-based game environments.
- Spatial reasoning with movement constraints
- Sequential planning to reach goal states
- Understanding blocking relationships between objects
- Multi-step problem decomposition
RushHour does not require a sandbox. It has minimal compute requirements.
MIT.
There are two splits: train (150 tasks) and test (150 tasks). Each split contains 50 tasks across each of 3 variants:
- RushHour-v0
- RushHour-v0-train
- RushHour-v0-raw
Each task is seeded for reproducibility.
This is a sparse reward environment. Rewards are mapped from TextArena's native range of {-1, 0, 1} to {0.0, 0.5, 1.0} via (raw + 1) / 2.
We do not use LLM graders for this environment; reward is determined programmatically.
Game state is generated procedurally by the TextArena engine using seeded randomness. No external data files are required.
Agents are given a single tool:
move_car(car_id, direction): Move a car on the Rush Hour grid. Provide car letter and direction ('+' forward, '-' backward). Free car X to exit!
RushHour is a multi-turn environment.
This environment presents moderate to challenging difficulty, requiring agents to understand spatial constraints and plan sequences of moves to unblock the target vehicle.
There are no further environment requirements; RushHour works out of the box without any secrets or API keys.
Agents in RushHour interact only with a sliding puzzle game and have no access to external systems, the internet, or sensitive data. The environment does not present safety risks.
@software{textarena2024,
author = {Guertler, Leon and Banting, Wilfried and Pignatelli, Eduardo},
title = {TextArena},
year = {2024},
publisher = {GitHub},
url = {https://github.com/LeonGuertler/TextArena}
}