Autonomous-driving systems must process large volumes of sensor data and generate vehicle-control commands within strict time limits. As vehicles increasingly become software-defined, their development often involves multiple software platforms with different characteristics. The AUTOSAR Adaptive Platform (AUTOSAR AP) is widely used in the automotive industry, while the open-source Robot Operating System 2 (ROS 2) has developed a rich ecosystem for research and early-stage development. Combining the strengths of these platforms is therefore becoming increasingly important. However, when AUTOSAR AP and ROS 2 coexist within the same autonomous-driving system, tracing the complete flow of data and identifying where processing delays occur across the platform boundary remain challenging.
A tracing framework called CART (Combined AUTOSAR AP and ROS 2 Tracing Framework) was previously developed to integrate trace information from the two platforms. However, its evaluation had been limited to a small-scale application configuration. Actual autonomous-driving systems involve many software components working together while processing large amounts of sensor data, and it had not been established whether CART could provide useful end-to-end latency analysis under workloads approaching the scale and complexity of practical autonomous-driving software. Establishing an evaluation environment closer to real development conditions is important for identifying performance bottlenecks during rapid prototyping before software is deployed on physical vehicles.
A joint research team led by Professor Takuya Azumi of the Graduate School of Science and Engineering at Saitama University, in collaboration with Astemo, Ltd., aimed to determine whether unified end-to-end tracing could be applied to an autonomous-driving system of practical scale and complexity. The team constructed a cloud-based evaluation platform combining CART with the high-fidelity CARLA autonomous-driving simulator, the Autoware autonomous-driving software stack based on ROS 2, and AUTOSAR AP. Point-cloud sensor data generated by CARLA were passed through ROS 2 and AUTOSAR AP for object detection, with the detection results returned to Autoware to generate vehicle-control commands. Using this environment, the researchers reconstructed the expected end-to-end communication topology with 100% coverage, observed a point-cloud reception callback rate of about 33 Hz, and analyzed latency across the ROS 2–AUTOSAR AP boundary with low tracing overhead, successfully demonstrating unified end-to-end performance analysis in a practical-scale mixed autonomous-driving software environment. The study, titled “Evaluation Platform for Tracing of Autonomous Driving System Combined AUTOSAR AP and ROS 2” was published in the IEEE Open Journal of the Industrial Electronics Society on August 6, 2026. The DOI is 10.1109/OJIES.2026.3721135.
Key findings of the study include:
“This study is important because autonomous-driving software is becoming increasingly complex, with different software platforms working together within the same system,” says Shunsuke Ito, a master’s student at the Graduate School of Science and Engineering, Saitama University, and corresponding author of the study. “By integrating trace information from AUTOSAR AP and ROS 2, our platform makes it possible to follow the processing path from sensor input to the generation of a control command as a single end-to-end sequence. This allows developers to see more clearly where latency occurs across platform boundaries.”
The study also demonstrates that such analysis can be performed under a workload much closer to practical autonomous-driving development than in the team’s previous evaluation. “It is not enough to confirm that a tracing method works with a small test application,” Ito explains. “Autonomous-driving systems process large amounts of sensor data through many interconnected software components. By combining CARLA, Autoware, AUTOSAR AP, and ROS 2 in a cloud environment, we were able to test CART under more realistic conditions and show that detailed latency analysis remains possible without imposing a large additional processing burden.”
The results could help improve the way increasingly complex Software-Defined Vehicle (SDV) systems are developed and evaluated. By providing a unified view of processing across heterogeneous software platforms, the approach could make it easier for engineers to identify bottlenecks during the Design-and-Test cycle before software is transferred to actual vehicle hardware. The study also identifies several steps needed before practical deployment, including validation with real electronic control units or hardware-in-the-loop systems, clock synchronization across multiple computing systems, larger-scale testing, and further automation of cross-platform message correlation.
Looking ahead, the researchers plan to extend the platform to lighter-weight simulation environments, incorporate OpenSCENARIO-based systematic testing, automate message linking using standard SOME/IP identifiers, and ultimately evaluate CART in real hardware environments.
“As vehicle software grows in scale and diversity, tracing interactions across platforms becomes more important,” Ito says. “If developers can trace the behavior of those systems across different platforms and identify performance problems earlier, they may be able to shorten development and validation cycles while making complex vehicle software easier to evaluate systematically.”
“In the future, we hope to extend CART from cloud-based simulation to real ECUs and further automate the tracing process,” Ito explains. “If these capabilities can be incorporated into practical vehicle-development workflows, unified tracing could become a useful tool for developing autonomous-driving and software-defined vehicle functions more efficiently and reliably.”
IEEE Open Journal of the Industrial Electronics Society
Evaluation Platform for Tracing of Autonomous Driving System Combined AUTOSAR AP and ROS 2
6-Aug-2026