🚢 Autonomous Surface Vehicle Navigation System

한국어 | English

---

📖 Overview

A simulation platform for Autonomous Surface Vessels (ASV) that navigate safely to goal positions while avoiding dynamic and static obstacles in maritime environments

---

✨ Key Features

🤖 Navigation Algorithms

• A* Pathfinding: Grid-based optimal path calculation
• Reinforcement Learning (PPO): Real-time obstacle avoidance with 360° sensors
• RRT* (Reference): Sampling-based path planning (scripts only)

🌊 Simulation Environment

• Realistic maritime environment using Crest Ocean System
• Configurable static (08) and dynamic (04) obstacles
• 6-DOF boat physics simulation

---

🚀 Quick Start

Prerequisites

• Unity: 2021.3 LTS or higher
• Python: 3.8+ (for ML-Agents training)
• ML-Agents Toolkit: 2.0+

Installation

1. Clone repository

git clone https://github.com/your-username/unity-asv-navigation-rl.git
cd unity-asv-navigation-rl

2. Install ML-Agents (for RL training)

pip install mlagents==0.30.0

3. Open project in Unity

   • Launch Unity Hub
   • Click "Open" and select the cloned project folder
   • Wait for Unity to import all assets

4. Verify package dependencies

   • The project will automatically import required packages via Package Manager
   • Check dependencies in Packages/manifest.json

---

🎮 Usage

Method 1: Run in Unity Scenes

A* Algorithm

1. Open Assets/Scenes/Train-AStar.unity
2. Click Play button in Unity Editor
3. The boat will automatically navigate using A* pathfinding
4. Results are logged to the console

Reinforcement Learning

1. Open Assets/Scenes/Train-RL.unity
2. Ensure a trained model is loaded in the BoatAgent_RL component
3. Click Play button
4. The boat will navigate using the trained RL policy

Method 2: Use Training Area Prefabs

1. Create a new scene or open an existing scene
2. Drag the desired prefab from Assets/Prefabs/:
   • TrainArea-AStar.prefab - A* navigation setup
   • TrainArea_RL.prefab - RL agent setup
3. Configure parameters in the Inspector
4. Press Play to start the simulation

Train New RL Model

1. Configure training parameters

   • Edit config/boat_rl_curriculum.yaml (for curriculum learning)

2. Start training

mlagents-learn config/boat_rl_curriculum.yaml --run-id=ASV_Navigation_v1

3. Monitor training progress

   • Open TensorBoard: tensorboard --logdir results/
   • View learning curves, rewards, and episode statistics

4. Export trained model

   • Model will be saved in results/ASV_Navigation_v1/
   • Copy .onnx file to Assets/ML-Agents/ for inference

---

📁 Project Structure

autonomous-usv-navigation/
├── Assets/
│   ├── Scenes/                      # Unity scenes
│   │   ├── Train-AStar.unity        # A* navigation scene
│   │   └── Train-RL.unity           # RL training/test scene
│   │
│   ├── Scripts/                     # C# source code
│   │   ├── BoatProbesBase.cs       # Boat physics base class
│   │   ├── BoatProbes_AStar.cs     # A* boat controller
│   │   ├── BoatProbes_RL.cs        # RL boat controller
│   │   ├── BoatProbesAI.cs         # AI boat controller
│   │   ├── BoatAgent.cs            # Basic ML-Agents agent
│   │   ├── BoatAgent_RL.cs         # Advanced RL agent
│   │   ├── BoatAIController_AStar.cs  # A* pathfinding logic
│   │   ├── BoatAIController_RRT.cs    # RRT* pathfinding logic
│   │   ├── BoatAIAdapter.cs        # AI adapter interface
│   │   ├── BoatMovementEnhancer.cs # Boat movement enhancement utility
│   │   ├── ObstacleController.cs   # Obstacle control script
│   │   ├── MoveComponent.cs        # Movement component
│   │   ├── PlaytimeLogger.cs       # Runtime logging utility
│   │   └── DataScript/             # Data collection utilities
│   │       ├── AStar/              # A* experiment data
│   │       │   └── Experimentmanager.cs  # Experiment manager
│   │       └── Testmanager.cs      # Test manager
│   │
│   ├── Prefabs/                     # Reusable game objects
│   │   ├── TrainArea-AStar.prefab
│   │   ├── TrainArea_RL.prefab
│   │   └── [Obstacle prefabs]
│   │
│   ├── Materials/                   # 3D materials
│   ├── Models/                      # 3D models
│   └── ML-Agents/                   # Trained ML models
│
├── config/                          # ML-Agents configs
│   ├── boat_rl_curriculum.yaml      # Curriculum learning config
│   └── boat_rl_test.yaml            # Test config
│
├── Packages/                        # Unity packages (auto-managed)
├── ProjectSettings/                 # Unity project settings
│
├── README.md                        # English README (this file)
└── README_KR.md                     # Korean README

---

🛠️ Unity Packages

Unity Packages

• Crest Ocean System 4
• A* Pathfinding Project 4.2.17
• Unity ML-Agents 2.0+

👥 Authors

• 김영준 - GitHub
• 김민재 - GitHub
• 정민용 - GitHub

---

📚 References

• Unity ML-Agents Documentation
• Crest Ocean Docs
• A* Pathfinding Docs

---

_{Built with ❤️ for maritime autonomous research}