Modeling and Property Analysis of the Message Synchronization Policy in ROS

Abstract

Sensor fusion plays a significant role in autonomous driving (AD) systems. In reality, the sensor data sent to the fusion algorithm may have substantially different sampling times, especially when different sensors are deployed in a distributed way (e.g., in V2X systems). Without proper management, this could lead to poor sensor fusion quality. ROS is the most popular robotic software framework, which provides a sophisticated message synchronization component to manage the temporal inconsistency in sensor fusion. However, although widely used, there is little information about how the ROS synchronization policy works exactly, and people have to use it as a blackbox. In this paper, we formally model the message synchronization policy in ROS and analyze its important properties, including the uniqueness property, disjunction property, continuity property, optimum property, and delay-dependent property, which were discussed on the ROS website but without formal proofs. Our analysis reveals that some of these properties indeed hold but some only hold under certain conditions. We conducted experiments to validate our formal model’s correctness and evaluate the synchronization policy’s performance in terms of time disparity.