Robotics

Momentedge's clipper extracts event-triggered robot data clips from disk without RAM bounds

Capture robot event data with seconds of context before the trigger, using disk not RAM.

Deep Dive

Momentedge has released an open-source tool called momentedge-clipper that solves a common robotics problem: capturing the few seconds of data around critical events like disengagements or perception errors. Traditional continuous recording wastes disk space, while ROS 2's snapshot mode buffers messages in RAM, limiting preroll to available memory and missing data written before the buffer. momentedge-clipper runs alongside an ordinary ros2 bag record, tails the growing MCAP file on disk, and on each trigger (via a custom msg type) cuts a standalone MCAP clip covering [trigger_time - preroll, trigger_time + postroll]. This design means the preroll window is bounded only by disk capacity, not RAM, and the clip includes data written well before the trigger fired.

The tool is decode-free: it copies MCAP message bytes straight through without deserializing them, making it agnostic to message types and efficient. The resulting clips are standard, complete MCAP files that open in Foxglove, the mcap CLI, and ros2 bag replay. Triggers are published to a topic, so any fault detector, watchdog, operator button, or perception stack can drive it. Written in Rust using the r2r library, it passes end-to-end tests on ROS 2 Humble and Jazzy and ships as Debian packages per distro. A synthetic camera demo is included for testing without hardware. Licensed Apache-2.0, the project is early and seeks feedback on real-world usage.

Key Points
  • Tails continuous on-disk MCAP recording, not RAM buffer, so preroll is limited by disk space not memory.
  • Decode-free: copies message bytes without deserialization, works with any ROS 2 message types.
  • Triggered via any topic publishing a Momentedge Trigger message; produces standard MCAP files for immediate use.

Why It Matters

Enables reliable capture of rare but critical robot events for debugging, safety analysis, and postmortem.

📬 Get the top 10 AI stories daily