SEAM: An Optimal Message Synchronizer in ROS with Well-Bounded Time Disparity

摘要

Autonomous machines are commonly subject to real-time constraints. ROS 2, a widely-used robotics framework, considers real-time capabilities as a critical factor and is constantly evolving to address these challenges, e.g., the end-to-end timing guarantee and the real-time data fusion, etc. This paper studies the ROS message synchronizer, an integral component for multi-sensor data fusion, and provides a potential direction for the synchronizer's evolution in future versions of ROS 2. For effective data fusion, input data from different sensors must be sampled at time points that align within a specific range. This paper proposes a novel message synchronization policy to meet this requirement, called the SEAM, which Synchronizes the Earliest Arrival Messages once they fall within the specified range. Unlike traditional ROS synchronizers, the SEAM does not rely on prediction information for complex optimization. Instead, it uses information from already-arrived messages to construct a feasible synchronization scheme. We demonstrate the optimality of the SEAM by proving that it always finds a feasible scheme if one indeed exists. We incorporate the SEAM into ROS 2 and conduct experiments to evaluate its effectiveness compared to traditional ROS synchronizers.