lidar-filtering

Use the laser_filters package, which applies a chain of configurable filter plugins to sensor_msgs/msg/LaserScan messages. Define filters in a YAML file, then run the scan_to_scan_filter_chain node with topic remappings from your raw scan to a filtered output topic.

Use LaserScanAngularBoundsFilter to set readings within the chassis angular range to NaN, or LaserScanBoxFilter to remove points inside a 3D box defined in the base_link frame. Multiple filter instances can exclude several chassis regions.

LaserScanShadowsFilter removes artifacts at object edges where grazing beams create false intermediate-range points, detected via angle changes between adjacent readings. LaserScanSpeckleFilter removes isolated noise points whose range differs from all neighbors beyond a threshold.

Yes. Apply cheap angular and range filters first to remove large chunks of data, then run shadow and speckle filters on the remaining valid points. Placing speckle filtering before range filtering wastes computation on out-of-range readings.

The filter chain should publish at the same rate as the input; verify with ros2 topic hz on the filtered topic. A rate drop indicates excessive filter load, so reorder filters to discard invalid points early or reduce filter window sizes.