A proposal for a LidarScan sensor message
New standardized format promises 100% synchronized data channels and zero-copy compatibility for modern high-resolution sensors.
A significant proposal within the Robot Operating System (ROS) community aims to overhaul how lidar data is handled. Currently, most ROS applications use the sensor_msgs/PointCloud2 message, which represents lidar data as a list of 3D points. However, this format is increasingly problematic as lidar sensor resolution rapidly increases, creating hefty data transport burdens that many DDS implementations struggle with at high frame rates. More critically, PointCloud2 discards the native 2D grid structure in which modern lidar sensors natively capture data, losing valuable adjacency information that makes image processing algorithms efficient.
The proposed sensor_msgs/LidarScan message directly mirrors the native format of spinning and solid-state lidar sensors. It acts as a multi-spectral image, storing synchronized channels like range, intensity, and reflectivity in a memory-aligned, structured 2D array. This design eliminates the latency and CPU overhead of using ROS's `ApproximateTime` filter to sync separate data topics. The draft message includes geometry metadata for 3D projection, scaling fields to handle vendor-specific units (e.g., Ouster uses mm, Velodyne uses m), and a `uint8[] data` buffer designed for zero-copy compatibility, offering a more efficient and performant foundation for autonomous systems.
- Replaces inefficient PointCloud2 format that struggles with high-resolution, high-frame-rate data transport in ROS/DDS.
- Preserves native 2D lidar grid structure, enabling efficient image processing and direct application of 2D neural networks like YOLO.
- Provides 100% synchronized, multi-channel data (range, intensity, reflectivity) in one message, removing software sync latency and overhead.
Why It Matters
This standardization is critical for robotics and AV developers to efficiently process data from next-generation, high-resolution lidar sensors in real-time.