IMU/GNSS to ground truth extrinsics

[DanieleMarchisotti]

I could not find extrinsics between IMU and ground truth. They would be very helpful to make a comparison between the estimated trajectory and the ground truth one. Could you provide this information or is it available somewhere?

Thank you!

[DarrenWong]

Please refer to https://github.com/IPNL-POLYU/UrbanNavDataset/blob/73596f3fbe45b24bfd22b9e5b952e27d2ec84c43/UrbanNav-HK-Medium-Urban-1/extrinsic.yaml#L73

[DanieleMarchisotti]

Thank you very much for your answer!

I have another question. Which is the GNSS receiver that I should use in your dataset? Because in the extrinsics there is only one transformation matrix between GNSS and IMU (ANTENNA_T_IMU). Which is the GNSS receiver I should refer to in order to use that extrinsic matrix?

[DarrenWong]

Hi, the ANTENNA_T_IMU is for M8T and F9P which is located in the car front, which is named "m8t.obs", and"f9p.obs" in the dropbox link. https://github.com/IPNL-POLYU/UrbanNavDataset/blob/c95e9bce7cbf444a2e95765ac59409c66e7acb09/UrbanNav-HK-Medium-Urban-1/extrinsic.yaml#L60

And we also provide a measurement from ublox f9p which shares the same antenna (rear of the vehicle) with Ground truth with the name " f9p.splitter", the approximate lever arm (GNSS->Reference) is x=0.000, y=-0.560, z=-0.070 m (x-right, y-fwd, z-up)

Be noted that the extrinsic between the antenna and the reference frame is measured manually with an accuracy of within 5 centimeters. You can optimize the translation by comparing the RTK fixed results and the output from the ground truth.

[DanieleMarchisotti]

Thank you very much for your answer!

One last question. The IMU data file contains also orientation data (field.orientation.x, field.orientation.y, field.orientation.z, field.orientation.w). Is it the orientation computed directly from data obtained from IMU by considering only the magnetometer? If not, does the dataset contains magnetomer measurements?

[DarrenWong]

Yes, a Mti 10 AHRS sensor with 3-axis magnetometers in UrbanNav. For more details, you can refer to https://base.xsens.com/s/article/Magnetic-immunity-of-the-MTi-AHRS-1605869708053?language=en_US.

[DanieleMarchisotti]

Thank you for your answers.

I have seen that in the IMU csv files there is the covariance matrix for accelerometer, gyroscope and magnetometer. However, noise models are present also in the file xsense_imu_param.yaml. Which is the difference between the two noise models? Which one should I use?

[DarrenWong]

Plese use the xsens_imu_param for the noise term. The covariance matrix in the IMU files is a constant variance within the data collection according to here