A Containerized Microservice Architecture for a ROS 2 Autonomous Driving Software: An End-to-End Latency Evaluation

Overview

• The automotive industry is transitioning from traditional ECU-based systems to software-defined vehicles.
• Containers, lightweight virtualization technologies, play a central role in this revolution by enabling the flexible consolidation of complex software applications on a common hardware platform.
• Despite widespread adoption, the impact of containerization on fundamental real-time metrics such as end-to-end latency, communication jitter, and resource utilization has remained largely unexplored.

Plain English Explanation

The way cars are built is changing. Instead of relying on dedicated electronic control units (ECUs) for different functions, cars are now being designed as "software-defined vehicles." This means the car's software and hardware are more tightly integrated, allowing for more flexibility and customization.

A key technology driving this change is containers. Containers are a type of lightweight virtualization that allows multiple software applications to run on the same hardware, without interfering with each other. This makes it easier for carmakers to develop and deploy new features and updates.

However, the impact of using containers in cars is not well understood. Factors like the time it takes for data to travel end-to-end, the consistency of data transmission, and how much memory and processing power the containers use have not been thoroughly studied.

Technical Explanation

This paper presents a microservice architecture for a real-world autonomous driving application, where each software component is isolated in its own container. The researchers comprehensively evaluated this architecture, comparing it to a standard Linux-based deployment.

The results show that the container-based approach can improve the mean end-to-end latency by 5-8% compared to the Linux baseline. Additionally, the maximum latencies were significantly reduced when using containers.

These findings suggest that containers can provide benefits in terms of real-time performance for autonomous driving applications, even beyond what can be achieved with a traditional Linux setup. This aligns with research on simulating cloud environments for connected vehicle anomaly detection and practical approaches to resource management in these types of applications.

Critical Analysis

The paper provides a thorough evaluation of containerization for a real-world autonomous driving application, addressing a gap in the existing literature. However, the researchers acknowledge that their study is limited to a specific use case and architecture.

Further research is needed to understand the broader implications of containerization in the automotive domain, including its impact on AI-augmented automation for real driving prediction and long-term autonomous industrial inspection. Factors such as the scalability of container-based systems, their resilience to failures, and their compatibility with safety-critical automotive requirements should also be explored in future research.

Conclusion

This paper demonstrates the potential benefits of using containers in autonomous driving applications, particularly in terms of improving end-to-end latency and reducing maximum latencies. As the automotive industry continues its transition to software-defined vehicles, these findings suggest that containerization could play a crucial role in enabling the development and deployment of complex, real-time automotive systems.