Scheduling Performance Evaluation Framework for ROS 2 Applications
Bo Peng, Atsushi Hasegawa, Takuya Azumi · 2022
Recently, performance analysis research on self-driving software has attracted considerable attention because the development of self-driving software has increased performance needs. ROS 2 is an open-source application framework for robotics. Simultaneously, as the Executor is responsible for scheduling callbacks (the minimal schedulable entity) in ROS 2, the Executor's scheduling performance is one of the most crucial factors affecting the real-time performance of ROS 2. A performance evaluation tool for Executor scheduling is required to evaluate its performance. This paper proposes a tracing and performance analysis framework with instrumentation and tools for ROS 2 Applications, with the visualization of Executor scheduling results and analysis of Executor deadline misses. The proposed framework can help researchers better understand and develop ROS 2 Executor by providing deadline miss analysis of the end-to-end path and the scheduling data. Furthermore, this paper verifies the proposed approach to the Autoware Reference System, which can simulate a real-world scenario to evaluate the performance of Executor.