---

name: trajectory-data description: "Routes evo_traj workflows, trajectory file formats, converters, synchronization, alignment, export, and bag/MCAP loading." disable-model-invocation: true metadata: disco-role: operating license: GPL 3.0

trajectory-data

Use this sub-skill for trajectory files, conversion helpers, synchronization, export, and the evo_traj CLI.

Route here when the user asks for

• evo_traj or trajectory inspection / export
• TUM, KITTI, EuRoC, ROS bag, ROS2 bag, or MCAP trajectories
• PosePath3D, PoseTrajectory3D, TrajectoryBundle, or load_transform
• timestamp association, --ref, --sync, --merge, --downsample, --motion_filter
• converting KITTI pose/timestamp pairs to TUM or scaling TUM timestamps
• duplicate timestamp checks or trajectory-file validation failures

Do not route here when the task is mainly about

• APE/RPE metric math or result zip creation
• saved-result comparison, tables, or pandas exports
• package settings, logs, or IPython shell setup
• notebook plotting or custom Python embedding

Start with

1. references/data-formats.md
2. references/cli-reference.md
3. references/workflows.md
4. references/troubleshooting.md
5. scripts/trajectory_io_smoke.py

Rules of thumb

• TUM rows are timestamp + xyz + quaternion; KITTI rows are the first 3 rows of a 4x4 pose matrix; EuRoC CSV timestamps are in nanoseconds and become seconds.
• PoseTrajectory3D requires timestamps and they must be ascending and unique.
• bag2 and mcap are the same route in evo; bag-based routes require topics or --all_topics.
• --sync and --ref matter for timestamped workflows; motion_filter on path-only data is invalid in metrics workflows.
• load_transform accepts JSON, .npy, or text 4x4 matrices and rejects anything that is not a valid SE(3) or Sim(3) matrix.
• The safe conversion helpers in this sub-skill write to an explicit output path instead of silently mutating source files in place.