Example Data

Some of the dependencies installed as part of the Installation Guide are various pntos-python-datasets-* packages. These Python modules each provide one dataset, a script to play the data, and a variable in the Python module which provides the path on the disk to the file.

Example Data Package Info

Package	Dataset	Script	Variable
pntos-python-datasets-lcm	cobra_gps_ins_example_data.log	play-lcm-dataset	EXAMPLE_LCM_LOG
pntos-python-datasets-ros	cobra_gps_ins_example_data_0.db3	play-ros-dataset	EXAMPLE_ROS_LOG

Tip

The above datasets are provided for convenience, but pntOS-Python is in no way limited to the above types! An app could use any version of ASPN and any type of transport, so long as there is a compatible transport plugin to handle the conversion.

ASPN23-LCM

For simplicity, most example apps use the ASPN23-LCM example data. See below for some more detailed info about each channel in this log, the data rate, the corresponding ASPN-LCM type, and the error model:

ASPN23-LCM Example Dataset Info

Channel	Rate (Hz)	ASPN Message	About	Error Model*
/sensor/bmp388/baro_pressure	0.2	measurement_barometer	Can be used to bound altitude error	FOGM with \(\sigma=100m\) and \(\tau=3600s\)
/sensor/ins-d/pva	200	measurement_position_velocity_attitude	Reference system (truth)	N/A
/sensor/simulated/directiontoknownfeature	1	measurement_direction_3d_to_points	Simulated 3D direction to points	N/A
/sensor/simulated/velocity	1	measurement_velocity	Simulated 3D platform-frame velocity	N/A
/sensor/ublox-ZED-F9T/position	1	measurement_position	Latitude, longitude, and altitude from a GPS receiver	Horizontal FOGM: \(\sigma=1.5m\) and \(\tau=300s\), Vertical FOGM: \(\sigma=2m\) and \(\tau=200s\). Timestamps are biased about 0.15 seconds into the future.
/sensor/ublox-ZED-F9T/pva	1	measurement_position_velocity_attitude	PVA from a GPS receiver, used for PosVel update	See /sensor/ublox-ZED-F9T/position
/sensor/ublox-ZED-F9T/velocity	1	measurement_velocity	Velocity from a GPS receiver	Timestamps are biased about 0.15 seconds into the future.
/sensor/vn-100/imu	100	measurement_IMU	A tactical-grade IMU	See Appendix

*In addition to any variances or covariances included in the ASPN message.

Appendix: IMU Error Model

This appendix documents the IMU error model which, while more simple than many inertial error models, is significantly more complex than the error models for the other sensors.

Both the accelerometers and the gyroscopes model two sources of error:

• A FOGM bias

• White noise (which manifests as a random walk in the integrated states)

Unless otherwise noted, all three orthogonal sensors have the same error model. Occasionally, we’ll increase the vertical error as part of the filter tuning process, but this is more of a judgement call.

Accelerometer Error Model

The FOGM bias has the following parameters:

• \(\sigma=2.4\text{e-}3, \frac{m}{\text{s}^2}\)

• \(\sigma_{t=0}=0.072, \frac{m}{\text{s}^2}\)

• \(\tau=300, \text{s}\)

And the white noise is:

• \(\sigma=3.887\text{e-}6, \frac{m}{\text{s}^\frac{3}{2}}\)

Gyroscope Error Model

The FOGM bias has the following parameters:

• \(\sigma=2\text{e-}4, \frac{\text{rad}}{\text{s}}\)

• \(\sigma_{t=0}=0.003, \frac{\text{rad}}{\text{s}}\)

• \(\tau=500, \text{s}\)

And the white noise is:

• \(\sigma=9.9\text{e-}4, \frac{\text{rad}}{\text{s}^\frac{1}{2}}\), for the first two axes

• \(\sigma=6.7\text{e-}5, \frac{\text{rad}}{\text{s}^\frac{1}{2}}\), for the third axis

Orientation

The rotation from the sensor frame to the platform frame is a constant:

\[\begin{split} C_\text{IMU}^\text{platform} = \begin{bmatrix} 0.99802515, 0.01772605, 0.06026269\\ -0.01742059, 0.99983262, -0.00559042\\ -0.0603517, 0.00452957, 0.9981669 \end{bmatrix} \end{split}\]