---
library_name: hivex
original_train_name: AerialWildfireSuppression_difficulty_4_task_4_run_id_0_train
tags:
- hivex
- hivex-aerial-wildfire-suppression
- reinforcement-learning
- multi-agent-reinforcement-learning
model-index:
- name: hivex-AWS-PPO-baseline-task-4-difficulty-4
  results:
  - task:
      type: sub-task
      name: protect_village
      task-id: 4
      difficulty-id: 4
    dataset:
      name: hivex-aerial-wildfire-suppression
      type: hivex-aerial-wildfire-suppression
    metrics:
    - type: crash_count
      value: 0.995833334326744 +/- 0.01863389536983029
      name: Crash Count
      verified: true
    - type: extinguishing_trees
      value: 0.0886829849332571 +/- 0.2431451047770778
      name: Extinguishing Trees
      verified: true
    - type: extinguishing_trees_reward
      value: 0.44341491162776947 +/- 1.2157254834823454
      name: Extinguishing Trees Reward
      verified: true
    - type: fire_too_close_to_city
      value: 0.02242424301803112 +/- 0.04967780915490442
      name: Fire too Close to City
      verified: true
    - type: preparing_trees
      value: 253.21703186035157 +/- 35.590698365756126
      name: Preparing Trees
      verified: true
    - type: preparing_trees_reward
      value: 253.21703186035157 +/- 35.590698365756126
      name: Preparing Trees Reward
      verified: true
    - type: water_drop
      value: 1.6389704704284669 +/- 0.33632718660720196
      name: Water Drop
      verified: true
    - type: water_pickup
      value: 1.6389704704284669 +/- 0.33632718660720196
      name: Water Pickup
      verified: true
    - type: cumulative_reward
      value: 145.59471988677979 +/- 48.609159328951485
      name: Cumulative Reward
      verified: true
---

This model serves as the baseline for the **Aerial Wildfire Suppression** environment, trained and tested on task <code>4</code> with difficulty <code>4</code> using the Proximal Policy Optimization (PPO) algorithm.<br><br>

Environment: **Aerial Wildfire Suppression**<br>
Task: <code>4</code><br>
Difficulty: <code>4</code><br>
Algorithm: <code>PPO</code><br>
Episode Length: <code>3000</code><br>
Training <code>max_steps</code>: <code>1800000</code><br>
Testing <code>max_steps</code>: <code>180000</code><br><br>

Train & Test [Scripts](https://github.com/hivex-research/hivex)<br>
Download the [Environment](https://github.com/hivex-research/hivex-environments)