Averera

Self-Driving Test Vehicle

The Autonomous Self-Driving Test Vehicle project combines computer vision, perception algorithms, and advanced control systems to achieve autonomous navigation. It integrates perception, localization, path planning, actuators, and real-time hardware communication to create a comprehensive self-driving solution.

Project Overview

This project aims to build a self-driving vehicle capable of perceiving its environment, planning paths, and executing actions autonomously. It incorporates several key components:

• Computer Vision and Perception: The project uses advanced computer vision techniques for road segmentation, obstacle detection, and environment understanding.

• Localization: Precise localization is ensured by sensors and camera data to determine the vehicle's position accurately.

• Path Planning: Path planning algorithms generate optimal trajectories based on environment perception and destination, creating waypoints for navigation.

• Control Systems: PID controllers offer real-time control for accurate steering, speed regulation, and overall stability.

• Arduino Communication: ROS and rosserial enable seamless communication with Arduino microcontrollers, allowing physical interactions and environment changes.

Features and Highlights

• Camera Feed Integration: Real-time camera feeds from Logitech USB cameras and Oak-D Lite cameras provide critical visual data for perception.

• Joystick Control: ROS-enabled joystick control allows manual interaction and controlled testing of vehicle functions.

• Deep Learning Framework: TensorFlow integration facilitates deep learning models for advanced perception tasks and object detection.

Usage

1. Install necessary ROS packages, including USB camera drivers and joystick support.
2. Launch camera nodes to feed video data into the perception pipeline.
3. Execute the perception node (vision.py) for road segmentation and object detection.
4. Generate waypoints and path plans using control systems and algorithms.
5. Utilize PID controllers for precise steering and speed management.
6. Communicate with Arduino microcontrollers for physical actuation.

Results

The project presents two critical demonstrations. The first showcases road segmentation and perception through a GIF, while the second displays autonomous vehicle navigation using perception, control, and planning.

run.mp4

Testing_only_on_vision.mp4

vision_testing.mp4

Future Development

As autonomous technology advances, this project serves as a foundation for further exploration. Future enhancements may involve advanced machine learning, semantic mapping, obstacle avoidance, and integration with larger-scale autonomous systems.

The Autonomous Self-Driving Test Vehicle project demonstrates the limitless potential of modern robotics and automation.

Running

1. roslaunch usb_cam usb_cam-test.launch publishes camera frame to "/usb_cam/image_raw" topic.
2. vision.py subscribes to "/usb_cam/image_raw" and publishes error to "error" topic.
3. controller.py subscribes to "error" and publishes to "steer" topic.
4. steer.py subscribes to "steer_arduino" topic.
5. throttle.py gives constant throttle and publishes to "arduino" topic.

serial_node_steer.py and serial_node_throttle.py used for communication with arduino using rosserial.

steer_joy.py is to be used for joystick control only.

Download model weights from here. Download Manual control ROS package from here. Download Autonomous ROS package from here. Download Oakd-Lite ROS package from here.

Setting up

1. Run sudo apt-get install ros-noetic-usb-cam to install camera node for video feed

2. Run sudo apt-get install ros-noetic-joy on terminal to install joystick package.

3. Run rosrun joy joy_node to start the node.

4. Run rostopic echo joy to see the topic result.

5. Run rosrun usb_cam usb_cam_node for the camera feed for logitech usb cam feed.

6. Run roslaunch depthai_examples mobile_publisher.launch for oak-d lite cam feed.

7. Run python3 -m pip install tensorflow and pip install tensorflow-addons on terminal to install tensorflow for vision.py.

8. Run sudo apt-get install ros-noetic-rosserial-arduino to install rosserial for communication with arduino.

Setting up Oak-D Lite:

1. Follow steps from here.
2. Follow every step till Docker(excluded), then go to 'Executing a file'.
3. There a 8 packages , do not use catkin_build_isolated but use catkin build.
4. After build, simply launch file given above at point 6 in Setting up.
5. Multiple topics will appear and desired image will be published on some topic.

General Structure of Autonomy