Some issues with IMU prediction

[Chaoyue-cpu]

Hi Dr. Koide,

When using hdl_400.bag, I found that the results with IMU are not ideal—the errors are actually larger compared to using LiDAR only. Even though I have set invert_imu_acc=true and invert_imu_gyro=true, the performance did not improve.

I noticed that in Issue #11（https://github.com/koide3/hdl_localization/issues/11#issuecomment-787610553）, the IMU results you presented were very good. Could you please share what specific settings or steps you took to achieve that?

Below are my parameter settings; the rest are the default values provided in the package. Additionally, in my own dataset, after transforming the LiDAR point cloud into the IMU frame, I encountered the same issue—IMU fusion even failed in areas where LiDAR-only localization worked well.

Thank you for your efforts. I look forward to your reply！