vehicle_tunnel00.bag in FusionportableV2

[RingerXu]

Thanks for your dataset! I found that the trajectory of vehicle_tunnel00.bag generated by Fast-lio2 in FusionportableV2 is different with the ground truth.

[fusionportable]

Thank you for your feedback on the use of our FusionPortablev2 dataset. Sorry for the inconvenience. Our vehicle data is collected in complex urban areas, and the trajectory truth is provided by the GNSS-IMU integrated navigation system. However, the tunnel environment is a GNSS-denied environment. In our vehicle_tunnel00 sequence, only reliable truth data when entering and exiting the tunnel is provided. As written in our paper, for ease of use, a unified evaluation standard is established, and poses with excessive uncertainty are manually deleted. We updated the truth data of our vehicle platform recently, please try to download and use it. The original ins data can still be obtained in rosbag.

Regarding the test of fastlio2 you mentioned, I gave an evaluation test based on the latest truth data. It degraded severely after entering the tunnel. Please check the attachment. In addition, in our test, coin-lio performed well in vehicle_tunnel00, and you can try a similar method.

[zbqq]

where is the latest ground truth data of vehicle_tunnel00.bag? @fusionportable

[gogojjh]

@zbqq the GT data is here: https://drive.google.com/file/d/1pJJQR0i-8Yppvh1ilwf6Zil7n0Kk4i00/view?usp=drive_link