🤖 Autonomous Navigation Using Traditional and AI-Based Approaches

Welcome to the Autonomous Navigation project! This repository demonstrates and compares traditional and AI-enhanced navigation methods for robotic systems in simulated environments using ROS Noetic and Gazebo. 🚀

---

Demo Clips :

Traditional Navigation Demo :

AI Based Approach Demo :

---

📝 Table of Contents

1. Introduction
2. Setup Instructions
3. Usage
4. Results
5. Acknowledgements
6. Contributions

---

🧐 Introduction

This project explores two navigation paradigms:

• Traditional Navigation: Using SLAM and A* for pathfinding.
• AI-Based Navigation: Leveraging semantic segmentation with YOLOv8 to classify terrain into safe (green), challenging (yellow), and restricted (red) zones.

The Jackal Robot was employed in a simulated Gazebo environment, equipped with:

• 🛠 Sensors: LiDAR, Intel RealSense D455, and IMU.
• 🌐 Simulation Environment: Clearpath Robotics' Inspection World.

The environment in Gazebo :

---

🛠 Setup Instructions

Prerequisites

Ensure the following are installed:

• ROS Noetic
• Gazebo
• RVIZ
• Python 3.9
• Dependencies listed here

Clone the Repository

git clone https://github.com/AdharshKan42/AI_Based_Auton_Nav
cd AI_Based_Auton_Nav

Install Dependencies

rosdep update
rosdep install --from-paths src --ignore-src -r

Build the Workspace

catkin_make
source devel/setup.bash

---

🚀 Usage

1️⃣ Launching the Simulation and Moving around

To start the Gazebo environment with the Jackal robot:

roslaunch cpr_inspection_gazebo inspection_world.launch platform:=jackal

Launching RVIZ

roslaunch jackal_viz view_robot.launch

To move the Bot around

rosrun teleop_twist_keyboard teleop_twist_keyboard.py cmd_vel:=/cmd_vel

To Launch the GUI Interface controller :

rosrun rqt_gui rqt_gui -s rqt_robot_steering

To View Camera Output

rosrun rqt_image_view rqt_image_view

RGB Camera Output :

Depth Camera Output :

2️⃣ Traditional Navigation Approach

Launching Gazebo

roslaunch cpr_inspection_gazebo inspection_world.launch platform:=jackal

Path Planning and Navigation (Launching gmapping with custom scan topic:)

roslaunch jackal_navigation gmapping_demo.launch scan_topic:=lidar/scan

roslaunch jackal_viz view_robot.launch config:=gmapping

Gmapping output :

Saving the map :

rosrun map_server map_saver -f mymap

To localize with AMCL :

roslaunch jackal_navigation amcl_demo.launch map_file:=/path/to/my/map.yaml

To start navigation :

roslaunch jackal_viz view_robot.launch config:=localization

Camera Output

• View camera: rosrun rqt_image_view rqt_image_view

3️⃣ AI-Based Navigation

Create a goal waypoint based on the robot's current position.

rosrun utm_to_robot waypoint_creation.py

Start YOLOv8 Inference on the Realsense Camera

rosrun utm_to_robot yolo_inference.py

Inference Ouput on the Realsense Camera :

Run the navigation algorithm to move towards the goal position

rosrun utm_to_robot mock_movebase_combined.py

---

📊 Results

Metric	Traditional (A*)	AI-Based (YOLOv8)
Path Length (meters)	18.6	17.3
Traversal Time (secs)	45.2	39.8
Collision Rate	1	0
Computational Load	100% CPU	65% CPU, 30% GPU
Adaptability	Low	High

Key Insights

• Traditional Approach: Reliable for static environments, struggles with dynamic changes.
• AI-Based Approach: Adapts well to unstructured terrains with efficient navigation.

---

---

🤝 Contributions

This project was a collaborative effort by:

• @AdharshKan42 Adharsh Kandula: Environment setup and AI model integration.
• @lowwhit Lohith Venkat Chamakura: Implementation of Traditional navigation algorithms.
• @nishitpopat Nishit Popat: Mapping algorithms Pathfinding optimizations.
• @rrrraghav Raghav Mathur: Testing of multiple models and finetuning.
• @SwordAndTea Wei Xiang: Data collection and preprocessing, AI model training and evaluation.

---

Feel free to contribute! Submit issues or pull requests to improve this repository. 🎉

---