Skip to content

Coordinate frames

Pose Engine generates pose estimates expressed as transforms between coordinate frames. Understanding which frames are used in each output is important for integrating pose results into your autonomy stack.

Local map frame

The local map frame is the world frame referenced in *.local.csv output files and for the online map pose output.

The origin and orientation of the local map frame are set during map building. Robots working in the mapped area will all localize relative to the same local map frame, providing consistent pose references.

ECEF frame

The ECEF (Earth-Centered Earth-Fixed) frame is an absolute global reference frame. Offline *.ecef.csv outputs are expressed in the ECEF frame.

ECEF output is only generated when GNSS or INS data is present in the input sensor data and integrated into the map during map building.

Sensor coordinate frames

Each offline CSV output file corresponds to a specific sensor coordinate frame attached to the robot. The <sensor-cf> component of the filename identifies which frame the pose is relative to.

For example, acme_log1_front_lidar.local.csv contains the world-from-front_lidar transform at each timestamp.

The sensor frame names are defined during platform setup and match the frame names in Calibration Anywhere extrinsics outputs. To obtain pose in a frame that is not directly output, apply the known rigid transform between sensor frames (from extrinsics.yaml).

Online output frames

Online outputs use the following frames:

Output Frame
nav_msgs/Odometry Arbitrary; configured during deployment
geometry_msgs/PoseWithCovarianceStamped Local map frame
Map-relative pose Named coordinate frame registered to the map; configured during deplyment

Relationship between frames used offline and online with the same map

The local map frame is consistent between offline and online operation. Offline *.local.csv poses and online geometry_msgs/PoseWithCovarianceStamped messages reference the same map origin, enabling direct comparison between offline ground truth and online localization output.