SLAM becomes unstable and map freezes when Unitree 3D LiDAR is mounted at an inclined angle (~45°)

[Epik-Robotik]

Hello,

I am using a Unitree 3D LiDAR with ROS Noetic, running LiDAR-based SLAM.
An IMU is available and properly connected, and the LiDAR–IMU TFs are correctly defined.

The system works stable only when the LiDAR is mounted vertically (Z-axis pointing upward).
However, when the LiDAR is mounted at an inclined angle (~45°), serious SLAM instability occurs.
Observed Issues

• When the LiDAR is mounted at an angle:
• SLAM starts but map frame stops publishing after a short time
• RViz map visualization freezes
• /tf is published briefly and then stops
• /map or odom → map transform is published only once
• Point cloud appears to drift below the ground

SLAM repeatedly prints warnings such as:

[WARNING] The pose does not comply with the motion limitations. Lidar SLAM may have failed. [WARNING] No intersecting voxels found with current scan [WARN] Skipping Pose as Localization step was not valid.

Checks Already Done

• -TF tree verified (base_link → unilidar_lidar)
• IMU data confirmed valid and synchronized
• Undistortion enabled / disabled (no effect)
• Crop volume and Z-range tuned
• Scan rate, timestamp unit, and frame IDs verified
• Multiple SLAM packages tested (similar behavior observed)

Questions / Request

1. Is there a recommended SLAM algorithm for Unitree 3D LiDAR when mounted at an inclined angle?
2. Are there specific preprocessing or configuration parameters required for tilted installation?
3. Is this behavior a known limitation of the current driver / SLAM pipeline?
4. Are there IMU–LiDAR fusion best practices specific to Unitree sensors?
5. Any guidance, example configurations, or confirmation would be greatly appreciated.

Thank you.

[markgol]

If you are using 2D workmode and the function parseFromPacketPointCloud2D()
In unitree_lidar_utilities.h
the array resize should be 1800 not 300.