Digital Twin of an Autonomous Scale Vehicle

Author: Adwita Kashyap
Supervisor: Dr. Istvan David

Introduction

This research project aims to build a digital twin of a sensor-equipped robot car using real-time data streaming. The digital twin is a virtual representation of the physical system's sensor outputs while also enabling control of the system. In this research project, we offload computation-intensive tasks such as object detection to the digital twin for faster processing.

Software

• Language: Python
• ROS2 distribution: Humble
• Object detection: YOLOv11 custom trained model
• Communication Frameworks:
  • ZeroMQ: Used for communication of controls between the physical twin and the digital twin
  • Image ZeroMQ: Used for communication of sensor data between the physical twin and the digital twin

Hardware

• Robot platform: Hiwonder JetAcker
• Chassis: Ackerman
• Onboard computer: Raspberry Pi 5 (8GB)
• Sensors:
  • SLAMTEC A1 LiDAR
  • Orbbec Astra Pro Plus (RGB + depth camera)

Setup

This project requires code to be run on both the vehicle's onboard computer and the user's remote system. Below is a breakdown of the relevant libraries and files.

Physical Twin:

Install requirements listed in requirements/car.txt Put the following files, located in main, on the physical twin's computer:

• CarController.py
• CarPublisher.py
• CarRoutePlanner.py

Digital Twin:

Install requirements via pip install -r requirements/user.txt. Run the following file, located in main, on the digital twin:

• UserObjectDetector.py