Skip to content

Run a Multi-Agent Baseline

We implemented several Multi-LLM baselines within CREW Wildfire to demonstrate the state-of-the-art multi-agent intelligence within our environment. This guide shows how to run the baseline CAMON on a simple level.

Prerequisites

Before running any algorithm, ensure you have:

  1. Completed the CREW Wildfire installation
  2. Docker container is running
  3. Navigated to the CREW Wildfire directory and activated the crew conda environment: 
    cd crew-algorithms/crew_algorithms/wildfire_alg
conda activate crew

Procedure

Run the run_tests.sh shell script:

run_tests.sh

or run the following command:

python algorithms\\CAMON\\__main__.py envs.level=Cut_Trees_Sparse_small envs.seed=483  envs.max_steps=20

The script will then open the Unity environment and begin testing.

The terminal should print the following:

Task: Cut all trees at (10, 21), (19, 21), (17, 21), (25, 4), (15, 5), (10, 24)
Agent Count: 3
Max Steps: 20

As the test progresses, POV + Minimap observations, chats, and scores are avaiable under results/logs.

└── logs/CAMON/Cut_Trees_Sparse...          
    ├── Agent_1/
    │   ├── Minimap/ # Minimap Screen Captures
    │   ├── POV/ # POV Screen Captures
    │   └── chats.txt # chat history
    │
    ├── Agent_2/
    │   └── ...
    ├── Agent_3/
    │   └── ...
    │
    ├── Server_Accumulative/ # Team Accumulative Minimap
    ├── Server_Map/ # Ground Truth Map
    └── data.csv # Score + API Calls, Input + Output tokens

When the test is complete, a video will be rendered from all the Minimaps and POVs:

demo