Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Low Fidelity Digital Twin for Automated Driving Systems: Use Cases and Automatic Generation

Jiri Vlasak, Jaroslav Klapálek, Adam Kollarčík, Michal Sojka, Zdeněk Hanzálek

TL;DR

The paper tackles the reality gap in automated driving system (ADS) development by proposing a low-fidelity digital twin (DT) generator that automates virtual environment and vehicle model creation. It advocates prioritizing rapid DT generation over high fidelity, while enabling real-time bidirectional communication with the real vehicle via ROS, demonstrated through a Gazebo-based workflow. The authors provide a practical DT generator, validated by replaying real-vehicle data, and illustrate use cases such as online identification, reality-gap observation, and intersection traversal to show how DT can streamline development and testing. This approach offers a scalable path to accelerate ADS verification and validation, with future work including handling non-reproducible events and leveraging reinforcement learning with multiple vehicle models.

Abstract

Automated driving systems are an integral part of the automotive industry. Tools such as Robot Operating System and simulators support their development. However, in the end, the developers must test their algorithms on a real vehicle. To better observe the difference between reality and simulation--the reality gap--digital twin technology offers real-time communication between the real vehicle and its model. We present low fidelity digital twin generator and describe situations where automatic generation is preferable to high fidelity simulation. We validated our approach of generating a virtual environment with a vehicle model by replaying the data recorded from the real vehicle.

Low Fidelity Digital Twin for Automated Driving Systems: Use Cases and Automatic Generation

TL;DR

The paper tackles the reality gap in automated driving system (ADS) development by proposing a low-fidelity digital twin (DT) generator that automates virtual environment and vehicle model creation. It advocates prioritizing rapid DT generation over high fidelity, while enabling real-time bidirectional communication with the real vehicle via ROS, demonstrated through a Gazebo-based workflow. The authors provide a practical DT generator, validated by replaying real-vehicle data, and illustrate use cases such as online identification, reality-gap observation, and intersection traversal to show how DT can streamline development and testing. This approach offers a scalable path to accelerate ADS verification and validation, with future work including handling non-reproducible events and leveraging reinforcement learning with multiple vehicle models.

Abstract

Automated driving systems are an integral part of the automotive industry. Tools such as Robot Operating System and simulators support their development. However, in the end, the developers must test their algorithms on a real vehicle. To better observe the difference between reality and simulation--the reality gap--digital twin technology offers real-time communication between the real vehicle and its model. We present low fidelity digital twin generator and describe situations where automatic generation is preferable to high fidelity simulation. We validated our approach of generating a virtual environment with a vehicle model by replaying the data recorded from the real vehicle.
Paper Structure (11 sections, 3 figures)

This paper contains 11 sections, 3 figures.

Table of Contents

  1. Introduction
  2. Related work
  3. Simulators
  4. Digital twin technology
  5. Development and testing
  6. Digital twin generator
  7. Digital twin use cases
  8. Online identification and tuning
  9. Observing the reality gap
  10. Driving through an intersection
  11. Conclusion

Figures (3)

  • Figure 1: Connection between the topics of work on automated driving systems from the point of view of digital twin technology. The solid lines represent a close relevance. The dashed lines represent the connection between virtual and physical entities.
  • Figure 2: Diagram of the steps required to use OpenStreetMap data and vehicle parameters (boxes in the top row) in different simulators (boxes in the bottom row). The solid lines represent automatic conversions and the loading and saving of files in different formats (rounded boxes). The boxes for the Unreal Engine editor and the Unity editor represent non-trivial manual work. The dashed lines represent our generator.
  • Figure 3: The screenshot of the Gazebo Classic simulator with automatically generated virtual entities reflecting the test track in Weissach, Germany, and the ADS-equipped vehicle under test. Buildings and roads © OpenStreetMap contributors.