File size: 1,799 Bytes
dd095e8
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
---

library_name: hivex
original_train_name: DroneBasedReforestation_difficulty_5_task_6_run_id_1_train
tags:
- hivex
- hivex-drone-based-reforestation
- reinforcement-learning
- multi-agent-reinforcement-learning
model-index:
- name: hivex-DBR-PPO-baseline-task-6-difficulty-5
  results:
  - task:
      type: sub-task
      name: explore_furthest_distance_and_return_to_base
      task-id: 6
      difficulty-id: 5
    dataset:
      name: hivex-drone-based-reforestation
      type: hivex-drone-based-reforestation
    metrics:
    - type: furthest_distance_explored
      value: 137.37953353881835 +/- 12.615748983046979
      name: Furthest Distance Explored
      verified: true
    - type: out_of_energy_count
      value: 0.6040635073184967 +/- 0.08043410811022636
      name: Out of Energy Count
      verified: true
    - type: recharge_energy_count
      value: 106.3367606653273 +/- 119.63729576848576
      name: Recharge Energy Count
      verified: true
    - type: cumulative_reward
      value: 3.9467455238103866 +/- 4.488707334085729
      name: Cumulative Reward
      verified: true
---


This model serves as the baseline for the **Drone-Based Reforestation** environment, trained and tested on task <code>6</code> with difficulty <code>5</code> using the Proximal Policy Optimization (PPO) algorithm.<br><br>Environment: **Drone-Based Reforestation**<br>Task: <code>6</code><br>Difficulty: <code>5</code><br>Algorithm: <code>PPO</code><br>Episode Length: <code>2000</code><br>Training <code>max_steps</code>: <code>1200000</code><br>Testing <code>max_steps</code>: <code>300000</code><br><br>Train & Test [Scripts](https://github.com/hivex-research/hivex)<br>Download the [Environment](https://github.com/hivex-research/hivex-environments)