ROSfs: A User-Level File System for ROS

TL;DR

The paper tackles the data-query bottleneck and cross-robot collaboration challenge in multi-robot systems using ROS. It introduces ROSfs, a user-level file system that stores ROS messages in labeled sub-files organized by a time-indexed structure, enabling both fast offline queries and real-time online access across robots. The authors implement a C++ core with Python interfaces and demonstrate substantial performance gains—up to 129x offline speedups and up to 7x latency reductions in online, wireless settings—while preserving ROS compatibility. This approach advances practical data management for MRS by enabling prompt data acquisition, analysis, and collaboration across robots and remote servers.

Abstract

We present ROSfs, a novel user-level file system for the Robot Operating System (ROS). ROSfs interprets a robot file as a group of sub-files, with each having a distinct label. ROSfs applies a time index structure to enhance the flexible data query while the data file is under modification. It provides multi-robot systems (MRS) with prompt cross-robot data acquisition and collaboration. We implemented a ROSfs prototype and integrated it into a mainstream ROS platform. We then applied and evaluated ROSfs on real-world UAVs and data servers. Evaluation results show that compared with traditional ROS storage methods, ROSfs improves the offline query performance by up to 129x and reduces inter-robot online data query latency under a wireless network by up to 7x.

Figures (16)

• Figure 1: ROSfs Overview.
• Figure 2: Bag file format.
• Figure 3: ROS publish/subscribe model zhang2020bora.
• Figure 4: Query performance on a 26.5GB dataset.
• Figure 5: rosbag record with split bag files