The system framework of the LiDAR service is shown in the figure.

• The Unilidar SDK obtains raw measurement data packets from LiDAR hardware through a serial port and parses them into point cloud data and IMU data
• The Unilidar Node module obtains the parsed point cloud data and IMU data through the Unilidar SDK, and preprocesses the point cloud using robot state information obtained from the robot dog motion program to remove distortion. Finally, it publishes the point cloud and robot odometer information in the world coordinate system after removing motion distortion.
• The Height Mapping module utilizes point cloud and odometer information to construct a 2.5D height map and provides it to the motion planning module for path planning, obstacle avoidance, and foot landing control functions.
• The Voxel Mapping module utilizes point cloud and odometer information to construct a 3D voxel map and transmit it to the mobile app, enabling real-time visualization of the 3D environment map around the robot on the mobile end, facilitating remote operation of the robot.

Unitree L1 LiDAR

Unitree 4D LiDAR-L1 is a safe and reliable 4D LiDAR developed by Unitree (3D position+1D grayscale). It has the ability to achieve high-speed laser ranging sampling every 21600 times and excellent ultra wide angle scanning ability. The field of view (FOV) is extended to 360 ° horizontally and 90 ° vertically, enabling three-dimensional spatial detection of hemispherical field of view angles; IMU module with built-in 3-axis acceleration and 3-axis gyroscope, supporting a push frequency of 250Hz. Possessing strong obstacle detection and avoidance capabilities, capable of handling various complex road conditions, achieving obstacle avoidance in the movement of GO2 quadruped robots.

L1 radar mainly includes laser emission and ranging core, reflector, high-speed rotating motor, and low-speed rotating motor. In the working state, according to the perspective shown in the diagram, the rotation direction of the high-speed rotating motor and the low-speed rotating motor is shown in the following figure.

High speed motor rotation direction

Low speed motor rotation direction

L1 adopts laser time-of-flight ranging technology, combined with high-speed laser acquisition and processing mechanisms, to achieve 21600 ranging actions per second. For each ranging action, L1 emits a ns level narrow pulse infrared laser signal. The reflected light after the laser signal is irradiated on the target object will be received by the radar's laser acquisition system. After analysis and processing by the processor, the distance value between the irradiated target object and L1, as well as the current angle, will be output from the communication interface.

The definition of the Cartesian coordinate system O-XYZ for L1 is shown in the following figure. Point O is the origin (IMU position), and O-XYZ is the point cloud coordinate system of L1 (the specific location of the origin can refer to the origin in the L1 3D model, which can be referenced when using Unilidar point cloud software).

LiDAR point cloud

The Go2 machine dog will automatically start the relevant programs of the LiDAR when turned on, and publish the point cloud collected by the LiDAR to the DDS topic.

By default, LiDAR posts two point cloud topics:

• Original point cloud of LiDAR, Its coordinate system is the original LiDAR coordinate system;
• Point cloud after removing motion distortion ,Its coordinate system is the world coordinate system that is consistent with the robot odometer.