Issue with vlp 16

[aditdoshi333]

Hello, Thanks for sharing amazing work. I am able smoothly with livox and pixhawk imu. But whenever I connect vlp 16 it's takes data in 1 2 secs and starts drifting during online refinement phase.

Please let me know if I had to do some code changes to make it run.

Thank you

[zfc-zfc]

starts

If the lidar odometry during Data Accumulation is correct, but refinement is bad, the reason might be there is not sufficient excitation. Please give really sufficient excitation just as our video shows. Besides, you can run the MATLAB code in folder "matlab_code" to see if the angular velocity and linear acceleration curves align well. The following pictures are ideal results:

[rahulsharma11]

Hi, Thanks for referring me here. I would like to add-

1. The calibration happened is so unexpected that it only happened once. After that i am not able to calibrate it.
2. I tried different orientation of Pixhawk also but no result.
3. As soon i start rotating along any axis slightly, all the excitations in x,y and z got quickly filled and it starts online refinement and the calibration gets screwed.

That way i am not confident about VLP16 with external IMU.

Is there any way we can use Odometry (like from realsense T265) directly with the Velodyne in Fast-Lio? or any other algorithm you can suggest.

Thanks.

[zfc-zfc]

Hi, Thanks for referring me here. I would like to add-

1. The calibration happened is so unexpected that it only happened once. After that i am not able to calibrate it.
2. I tried different orientation of Pixhawk also but no result.
3. As soon i start rotating along any axis slightly, all the excitations in x,y and z got quickly filled and it starts online refinement and the calibration gets screwed.

That way i am not confident about VLP16 with external IMU.

Is there any way we can use Odometry (like from realsense T265) directly with the Velodyne in Fast-Lio? or any other algorithm you can suggest.

Thanks.

Here are some suggestions that might be helpful for VLP-16:

1. For the 3rd comment you mentioned, you may increase "data_accum_length" to 800~1000, so that more data will be used for initialization.
2. During data accumulation, please make sure the lidar-only odometry runs correctly, i.e. the pointcloud is clear.
3. Since the vertical resolution of VLP-16 is too low, the lidar-only odometry is fragile. So you need to try to move your device constantly, to suffice the constant velocity model.
4. For VLP-16, try to modify cut_frame_num to 3.

FAST-LIO needs external IMU and accurate extrinsic transformation and synchronization. For Velodyne only lidar odometry, VoxelMap is recommended: https://github.com/hku-mars/VoxelMap

[rahulsharma11]

Hi, Thanks for the suggestions.

I tried all your above mentioned points. But no Luck. Also i tried VoxleMap, Over there either i use IMU or not, movement of Lidar does not produce Odoemtry. Also point cloud is not that good as FastLio(Even using Identity Matrix as Extrinsic Calibration).

My query is still there - To do SLAM with VLP16 , i need external IMU. Also for longer run i need good Lidar-IMU Calibration and that is not happening correctly.

[Taeyoung96]

@rahulsharma11
Hi, did you solve the problem?

[supertiechui]

I have the same question, how it works for 16 scans lidar

[af-doom]

Hello, have you solved the calibration problem of velodyne16

[af-doom]

@rahulsharma11 Hi, did you solve the problem?

Hello, have you solved the calibration problem of velodyne16

[af-doom]

Hello, have you solved the calibration problem of velodyne16

Hello, have you solved the calibration problem of velodyne16

[zfc-zfc]

Please see an example of successfully calibrating VLP16 and external IMU in #35