Create livox Solid-state Lidar sensor in Carla (including carla-ros-bridge adaptation)

System version Ubuntu 20.04LTS

ROS version Noetic

CARLA version 0.9.12

Reference

CARLA-0.9.12 & livox_laser_simulation

Release version CARLA_0.9.12_release ExtractionCode 1234

Major contribution

• Simulate the livox lidar, including the line number and intensity of livox.
• New attribute float decay_time. Control the reading frequency of lidar csv files. The larger the value, the more points will be read.
• New attribute float lidar_type. Simulation of different types of livox lidar.

  0.0-Horizon 1.0-Mid40 2.0-Avia 3.0-Tele 4.0-Mid360 5.0-HAP

• Original attribute dropoff_intensity_limit and dropoff_zero_intensitywill not be available (because of the livox-intensity)
• Original attribute dropoff_general_rate still available. General proportion of points that are randomy dropped.

Configure the main CPP files

• Use the files in the Unreal folder to replace or modify the files in the carla home directory

Configure the serializer/data files

• Use the files in the LibCarla folder to replace or modify the files in the carla home directory

Configure PythonAPI

• Use the files in the PythonAPI folder to replace or modify the files in the carla home directory

Configure ROS-BRIDGE

• Update the files in carla-ros-bridge

Configure livox_csv files

• Put the LivoxCsv folder in the /home directory and change its permission to user
• This file contains the exit angle (azimuth / zenith angle) and timestamp of different lidar at each point in 4 seconds officially given by Livox

Launch

cd ~/carla
make clean
make PythonAPI
make launch

Analysis of real point cloud and simulated point cloud (taking Horizon lidar as an example)

• Horizon lidar will generate 6 scanning points each time, and its line number is 0 to 5
• When the scanning of point 5 is completed, the next group of scanning will be started next to position 0 again
• Scanning diagram of real point cloud:
• Scanning diagram of simulation point cloud:
• point cloud distribution in CSV file:

Comparison between simulation point cloud and actual point cloud

• Create a custom map in Carla using real scene data
• At the same location, collect the real-world point cloud and the simulation point cloud in the carla map respectively
• Real horizon point cloud scanning(rviz):
• Simulation horizon point cloud scanning(rviz):
• Real lidar point cloud with 5 decay time：
• Simulation lidar point cloud with 5 decay time：

SLAM by simulation point cloud

• Using livox_horizon_loam to build map in Town_03, the effect is as follows: