A Real-World Latency Study of Microservice Architectures in Autonomous Driving

The study investigates how containerization impacts real-time performance in software-defined vehicles (SDVs). Using a microservice architecture built on Autoware and ROS 2, researchers deployed an autonomous driving stack on multiple hardware platforms. Their analysis revealed that containerized deployments can outperform bare-metal systems, improving end-to-end latency by 5–8% while enhancing CPU and memory utilization. The findings challenge long-held assumptions about virtualization overhead, highlighting containerization as a key enabler for efficient, scalable, and real-time automotive software architectures.