cnspy-relative-pose-evaluation 0.1.3

Evaluation of relative pose measurements.

cnspy_relative_pose_evaluation

A python3 package for evaluating relative pose measurements between two spatial frames in order to assess the accuracy. The baseline (ground truth) relative pose can be computed from two recorded 3D trajectories of the moving bodies and known extrinsics to the sensors. These can be specified in a single configuration file, see config.yaml

The following evaluations can be conducted:

Describtion	Images
Pose Plot (left: measured, right: gt)
Pose Error Plot (left: position, right: orientation, top: meas, middle: error, bottom: nees)
Range (gt vs. measured) outliers removed
Range Sorted (gt vs. measured) outliers removed
Range Error (measured-gt) outliers removed
Range Error Histogram (filtered) and distribution
Angle (gt vs. measured) outliers set to zero
Angle Error (measured-gt) outliers removed
Angle Error Histogram (filtered) and distribution
Statistics	statistics.yaml

Installation

Python 3.6 or greater is required. Inside this repo's directory, you may run

pip3 install .

or pip3 install -e . which installs the package in-place, allowing you make changes to the code without having to reinstall every time.

This package is still in development. Once stable it should be sufficient to run:

pip3 install cnspy_relative_pose_evaluation

Run from terminal

• RelPoseMeasEvaluationTool
• RelPoseMeasEvaluation
• RelPose_ROSBag2CSV
• ROSBag_TrueRelPoses
• ROSBag_Poses2RelPose

YAML configuration file

YAML configuration file is in the form of:

# relative pose of the moving sensors
sensor_positions: {0:[0, 0, 0], 1:[0, 0, 0], 2:[0, 0, 0]}
sensor_orientations: {0:[1.0, 0, 0, 0], 1:[1.0, 0, 0, 0], 2:[1.0, 0, 0, 0]}
# true pose of the body
sensor_topics: {0: "/uav10/vrpn_client/raw_pose", 1: "/uav11/vrpn_client/raw_pose", 2: "/uav12/vrpn_client/raw_pose"}
# topcis of the relative pose measurement
relpose_topics: {0: "/uav10/data_handler/uvdar_fcu", 1: "/uav11/data_handler/uvdar_fcu", 2: "/uav12/data_handler/uvdar_fcu"}

Important notes for the NEES computation

The covariance of the measurements are assumed to represent the position uncertainty in meters and in the origin of reference frame of the observer, e.g. the camera. The covariance of the measured orientation is assumed to be in radian representing the uncertainty about the roll-pitch-yaw angles about the x-y-z-axes in zyx order (Rz(y)*Ry(p)*Rx(r)) of the body frame (local), thus EstimationErrorType.type5:

• TRUE = EST + ERR for positions R(3);
• TRUE = EST \oplus ERR for rotations SO(3);
• global position error, local orientation error;.

Please check out: ErrorRepresentationType and EstimationErrorType

License

Software License Agreement (GNU GPLv3 License), refer to the LICENSE file.

Sharing is caring! - Roland Jung