/autonomous_ground_vehicle

Primary LanguageJupyter Notebook

Autonomous Vehicle Deep Learning Project

Overview

This project involves developing a deep learning model to predict steering angles for an autonomous vehicle. The model is integrated with a Raspberry Pi 4 Model B for real-time operation and control of the vehicle. here

Features

  • Deep Learning Model: Predicts steering angles from images captured by a front-mounted camera.
  • Real-Time Operation: Deployed on a Raspberry Pi 4 Model B.
  • Integration: Connects with hardware components such as a servo motor for steering adjustments.
  • Preprocessing: Images are resized and normalized for improved model accuracy.
  • Road Segmentation: Utilizes a pretrained model to assist with identifying drivable areas.

Getting Started

Prerequisites

  • Hardware:

    • Raspberry Pi 4 Model B
    • Front-mounted camera
    • Servo motor

  • Software:

    • Python
    • TensorFlow
    • OpenCV
    • CARLA simulator (for data collection)

Installation

  1. Clone the Repository: 
    git clone <repository_url>
cd <repository_directory>

Challenges

  • Data Discrepancies: Bridging the gap between simulation and real-world data to improve accuracy.
  • Track Design: Ensuring the turning radius in simulations matches the real-world vehicle.
  • Signal Noise: Addressing noise in the servo motor signal to reduce instability.

Future Improvements

  • Enhanced Data Collection: More real-world data for better model performance.
  • Track Accuracy: Refining simulation tracks to match real-world dynamics.
  • Noise Reduction: Implementing advanced filtering techniques for smoother control.

License

This project is licensed under the MIT License. See the LICENSE file for details.

Acknowledgements

  • CARLA Simulator
  • TensorFlow and Keras
  • OpenCV