ros2-perception-pipeline
Use when adding or debugging a perception node (camera, LiDAR, depth) in ROS 2 — verify the TF frame tree and time sync first, then wire sensor topics through PCL/Open3D processing to a documented out
SKILL.mdproject-trajectory/skills/ros2-perception-pipeline/SKILL.md
GitHubWhen to use. Any perception change. Why: a wrong sensor→base_link transform or bad time sync
makes everything downstream silently wrong — this is the #1 perception bug, so check it before logic.
Procedure.
- Inspect the TF tree (
ros2 run tf2_tools view_frames); confirm every sensor frame is connected and correct. - Confirm timestamps/
use_sim_timealignment across fused topics. - Process (PCL/Open3D) and publish results in a named, documented frame.
- Done when: TF tree verified (paste it), output visualized in RViz in the right frame, latency measured.
Knowledge: KNOWLEDGE-LIBRARY.md §B1.