File size: 3,088 Bytes
e718d1f 46ed70d e718d1f ea8a32a | 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 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 | ---
library_name: ml-agents
tags:
- SnowballTarget
- deep-reinforcement-learning
- reinforcement-learning
- ML-Agents-SnowballTarget
model-index:
- name: PPO-SnowballTarget
results:
- task:
type: reinforcement-learning
name: reinforcement-learning
dataset:
name: SnowballTarget
type: Unity-MLAgents-Env
metrics:
- type: mean_reward
value: 3.27
name: mean_reward
verified: false
- type: std_reward
value: 1.75
name: std_reward
verified: false
---
# ❄️ **PPO Agent on SnowballTarget**
This repository contains a trained **Proximal Policy Optimization (PPO)** agent that plays the **SnowballTarget** environment using the [Unity ML-Agents Library](https://github.com/Unity-Technologies/ml-agents).
---
## 📊 Model Card
**Model Name:** `ppo-SnowballTarget`
**Environment:** `SnowballTarget` (Unity ML-Agents)
**Algorithm:** PPO (Proximal Policy Optimization)
**Performance Metric:**
- Achieves stable performance in target-hitting tasks
- Demonstrates convergence to an effective policy
---
## 🚀 Usage (with ML-Agents)
Documentation: [ML-Agents Toolkit Docs](https://unity-technologies.github.io/ml-agents/ML-Agents-Toolkit-Documentation/)
```bash
mlagents-learn <your_configuration_file_path.yaml> --run-id=<run_id> --resume
```
```python
# Example: loading the trained PPO model
# (requires Unity ML-Agents setup)
model_id = "KraTUZen/ppo-SnowballTarget"
# Select your .nn or .onnx file from the repo
```
---
## 🧠 Notes
- The agent is trained using **PPO**, a robust on-policy algorithm widely used in Unity ML-Agents.
- The environment involves **throwing snowballs at targets**, requiring precision and timing.
- The trained model is stored as `.nn` or `.onnx` files for direct Unity integration.
---
## 📂 Repository Structure
- `SnowballTarget.nn` / `SnowballTarget.onnx` → Trained PPO policy
- `README.md` → Documentation and usage guide
---
## ✅ Results
- The agent learns to consistently hit targets with snowballs.
- Demonstrates stable training and effective policy convergence using PPO.
---
## 🔎 Environment Overview
- **Observation Space:** Continuous (agent position, target position, environment state)
- **Action Space:** Continuous (throwing angle, force)
- **Objective:** Maximize hits on targets with snowballs
- **Reward:** Positive reward for successful hits, penalties for misses
---
## 📚 Learning Highlights
- **Algorithm:** PPO (Proximal Policy Optimization)
- **Update Rule:** Clipped surrogate objective to ensure stable updates
- **Strengths:** Robust, stable, widely used in Unity ML-Agents
- **Limitations:** Requires careful tuning of hyperparameters (clip ratio, learning rate, batch size)
---
## 🎮 Watch Your Agent Play
You can watch your agent **directly in your browser**:
1. Visit [Unity ML-Agents on Hugging Face](https://huggingface.co/unity)
2. Find your model ID: `KraTUZen/ppo-SnowballTarget`
3. Select your `.nn` or `.onnx` file
4. Click **Watch the agent play 👀**
|