Forest-Fire/gridworld-openenv

🚀 Autonomous Navigation RL Environment (OpenEnv Compatible)

A lightweight, API-driven Reinforcement Learning environment that simulates robot navigation in a 2D GridWorld with obstacles, traps, and dynamic challenges.

🧠 Overview

This project implements a GridWorld environment where an agent must navigate from a start position to a goal while avoiding hazards and penalties.

Environment Elements

• —🟫 Obstacles — Block movement
• —⚠️ Traps — Penalize the agent
• —👾 Enemy (Hard Mode) — Introduces dynamic difficulty

The environment is exposed via a FastAPI interface and is fully Dockerized for easy deployment.

🎯 Features

• —✅ OpenEnv-compliant API
• —✅ Three difficulty levels: easy, medium, hard
• —✅ Reward normalization (0.0 → 1.0)
• —✅ Fully Dockerized setup
• —✅ Lightweight and CPU-friendly
• —✅ Inference script for evaluation
• —✅ Stable API with no validation errors

🎮 Action Space

🧮 Reward System

The reward function is designed to guide efficient navigation:

• —➖ Step penalty (encourages shorter paths)
• —📏 Distance-based reward shaping
• —⚠️ Trap penalty
• —👾 Enemy penalty (hard mode)
• —🏁 Goal reward

👉 Final reward is normalized to `[0, 1]`

🌐 API Endpoints

🔹 Reset Environment

POST /reset?task=easy

Query Params:

• —task: easy | medium | hard

🔹 Take a Step

POST /step

Request Body:

{
  "action": 1
}

Response:

{
  "state": [x, y],
  "reward": 0.85,
  "done": false
}

🔹 Get Current State

GET /state

🧪 Inference

Run the evaluation script:

python inference.py

Output Format

[START]
[STEP] action=...
[STEP] action=...
[END]

🐳 Docker Setup

Build Image

docker build -t env-test .

Run Container

docker run -p 7860:7860 env-test

The API will be available at:

http://localhost:7860

📁 Project Structure

app/
  ├── api.py        # FastAPI routes
  ├── env.py        # Core environment logic
  ├── models.py     # Data models (request/response)
  ├── tasks.py      # Task configurations
  ├── graders.py    # Evaluation logic

inference.py        # Agent interaction script
Dockerfile          # Container configuration
requirements.txt    # Dependencies
openenv.yaml        # OpenEnv specification
LICENSE

📊 Tasks

⚙️ State Representation

• —Format: [x, y]
• —Represents agent position in the grid
• —Fully compliant with OpenEnv expectations

✅ Submission Ready

• —✔ Correct state format
• —✔ Normalized rewards
• —✔ Stable API responses
• —✔ No validation errors

🎉 This environment is fully evaluator-ready.

📜 License

This project is released under the MIT License.

Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference