AI-Based Autonomous Navigation System

An AI-powered autonomous navigation system built in Python — featuring A* path planning, obstacle detection, and a real-time 2D simulation. No hardware or GPU required.

---

Demo

Agent (blue circle) navigating from start (green) to goal (red), avoiding obstacles (black), following the A* planned path (yellow).

---

Problem Statement

How can a robot or self-driving vehicle navigate from point A to point B in an unknown environment, avoiding obstacles — without human intervention?

This project solves exactly that. It implements the core AI pipeline used in real autonomous systems: perceive → plan → navigate → visualize.

---

Industry Relevance

Domain	Company	Application
Self-driving cars	Tesla, Waymo, Cruise	Lane keeping, collision avoidance
Warehouse robots	Amazon Robotics, Fetch	Autonomous goods transport
Delivery robots	Starship, Nuro	Last-mile autonomous delivery
Drones	DJI, Zipline	Obstacle-free flight paths
Industrial robots	ABB, KUKA, Fanuc	Safe arm navigation

---

Tech Stack

Tool	Purpose
Python 3.10+	Core programming language
Pygame 2.5	2D simulation and real-time visualization
NumPy	Grid matrix operations
A* Algorithm	Optimal shortest-path planning
Matplotlib	Path analysis plots (optional)

---

System Architecture

INPUT
  └─ 2D Grid Map (rows × cols)
     ├─ Start position (S)
     ├─ Goal position  (G)
     └─ Obstacle positions (1 = blocked)

PERCEPTION MODULE          (src/perception.py)
  └─ Scans grid → identifies obstacle cells

PATH PLANNING MODULE       (src/path_planning.py)
  └─ A* algorithm with Manhattan heuristic
  └─ Returns optimal waypoint list

NAVIGATION MODULE          (src/navigation.py)
  └─ Agent follows waypoints step by step
  └─ Reports progress and goal status

VISUALIZATION MODULE       (src/visualization.py)
  └─ Pygame renders grid, agent, path
  └─ Auto-saves screenshots for proof

OUTPUT
  └─ Animated navigation window
  └─ Terminal logs
  └─ Screenshot proof in outputs/

---

Folder Structure

AI-Autonomous-Navigation-System/
│
├── src/
│   ├── perception.py       # reads grid, identifies obstacles
│   ├── path_planning.py    # A* and Dijkstra algorithms
│   ├── navigation.py       # agent movement logic
│   └── visualization.py    # Pygame renderer + screenshot
│
├── simulation/
│   ├── environments.py     # open / maze / dense map generators
│   └── grid_maps/          # saved .npy map files
│
├── outputs/
│   └── screenshots/        # auto-saved proof images
│
├── notebooks/
│   └── demo.ipynb          # Jupyter walkthrough
│
├── docs/
│   └── images/             # README assets
│
├── main.py                 # ← run this to start
├── config.py               # grid size, colors, FPS
├── requirements.txt
├── .gitignore
└── README.md

---

Installation

1. Clone the repository

git clone https://github.com/YOUR_USERNAME/AI-Autonomous-Navigation-System.git
cd AI-Autonomous-Navigation-System

2. Create virtual environment

# Windows
python -m venv venv
venv\Scripts\activate

# Mac / Linux
python3 -m venv venv
source venv/bin/activate

3. Install dependencies

pip install -r requirements.txt

4. Verify installation

python -c "import pygame; import numpy; print('All dependencies OK')"

---

How to Run

python main.py

Change the map type in main.py line 52:

MAP_TYPE = "maze"    # "open" | "maze" | "dense"

Expected output

==================================================
  AI Autonomous Navigation System
  Map type : maze
==================================================

[Perception] Environment scan complete
  Total cells  : 400
  Obstacles    : 87
  Free cells   : 313
  Density      : 21.8%

[PathPlanning] Running A* algorithm...
[PathPlanning] Path found!
  Steps        : 34
  Path cost    : 33
  Cells visited: 156

[Navigation] Path loaded — 34 waypoints
[Navigation] Goal reached!
[Simulation] Navigation complete!

Screenshots are auto-saved to outputs/screenshots/.

---

Simulation Workflow

1. Environment loads — 20×20 grid with obstacles appears
2. Perception scans — obstacle positions logged to terminal
3. A* plans path — yellow cells show the optimal route
4. Agent navigates — blue circle follows path step by step
5. Goal reached — final screenshot saved, window stays open 2 seconds
6. Check proof — outputs/screenshots/ has 3 timestamped images

---

Results

Map Type	Path Length	Cells Explored	Time
Open	~28 steps	~90 cells	< 1s
Maze	~34 steps	~156 cells	< 1s
Dense	~48 steps	~220 cells	< 1s

Screenshots

Start	Midway	Goal Reached

---

Learning Outcomes

• Implemented A* search algorithm from scratch in Python
• Designed a modular perception → planning → navigation pipeline
• Built real-time 2D simulation with Pygame
• Understood how autonomous systems perceive and react to environments
• Practiced industry-style project structure and GitHub documentation

---

Future Improvements

• Real camera input with OpenCV lane detection
• YOLO object detection overlay
• CARLA 3D simulator integration
• Reinforcement learning agent (replace A*)
• ROS2 integration for real robot control
• SLAM — simultaneous localization and mapping
• Multi-agent simulation
• Web dashboard with Flask

---