Awesome

PCSim: LiDAR Point Cloud Simulation and Sensor Placement

Overview

• LiDAR Simulation Library
• LiDAR Placement Evaluation in V2X Scenario
• LiDARPlacementOptimizing
• RainyPCSim
• ReSimAD
• Technical Papers

LiDAR Simulation Library

https://user-images.githubusercontent.com/21357317/192964768-1b8711a5-742b-4b5b-bf59-c318e2e5c464.mp4

• Carla (>=0.9.12)

LiDAR List

1. Surround LiDAR
   1. pandar64
   2. ruby128
   3. pandar128
   4. vlp16
   5. hdl64
   6. hdl32
   7. pandar_qt
   8. bpearl
   9. pandar_40m
   10. pandar_40p
   11. pandar_xt
   12. vlp32
   13. os1_64_gen2
   14. waymo_top
2. Solid State LiDAR
   1. mems_m1
3. Risley Prism LiDAR
   1. horizon
   2. mid40
   3. mid70
   4. avia
   5. tele
4. Default Carla LiDAR
   1. default LiDARs

Usage

1. There are two paths should be modified:

LiDARSimLib/Unreal/CarlaUE4/Plugins/Carla/Source/Carla/Sensor/livox.cpp
line9: modify depend on your own carla path
LiDARSimLib/PythonAPI/lidar_lib_example/lidar_display.py
line30: modify depend on your own carla path

2. Use this file to overwrite the original carla directory

3. Rebuild carla

make clean
make PythonAPI
make launch

4. Create LiDAR

lidar_bp = world.get_blueprint_library().find('sensor.lidar.ray_cast') # choose ray_cast or ray_cast_semantic

lidar_bp.set_attribute("lidar_type", "Surround") # set lidar_type as Surround, Solid_state, Risley_prism

lidar_bp.set_attribute("name","pandar64") # set name as any one in LiDAR List

5. Set LiDAR ghosting object effect

lidar_bp.set_attribute("enable_ghost", "true") # enable ghosting object effect

6. Motion distortion simulation

• Import the LidarMotonDistortion module

import sys
sys.path.append("/your/path/of/carla/LibCustomFunction") # append the LibCustomFunction
from enable_motion_distortion import LidarMotonDistortion

• Init LidarMotonDistortion with file_path and distortion delay_time

lidar_motion_distort = LidarMotonDistortion("./horizon/", 10) # file_path is where the data you want to save. delay_time is the ratio of simulator frequency to lidar frequency.

• Enable motion distortion

ego_lidar.listen(lambda data: lidar_motion_distort.enable_motion_distortion(data, True))

Results

LiDAR beam simulation

Motion distortion simulation

Ghosting object effect simulation

LiDAR Placement Evaluation in V2X Scenario

Usage

1. Follow the instruction in the readme page in v2x-vit and install the v2x-vit project.
2. Download the pretrained model from here.
3. Use the script in Placement-Evaluation/v2xvit/tools/inference.py for evaluation.

LiDARPlacementOptimizing

Code and usage will be release soon.

RainyPCSim

Simulate rain effect for LiDAR and collect synthetic LiDAR data in CARLA. We have released the RainyPCSim Dataset obtained using the this method.

https://github.com/puffyyy/PCSim/assets/63792704/2fb22c7c-d94f-4bee-a2b6-1b3830efe28c

Requirements

• Carla (>=0.9.12)

Usage

1. Following LiDAR Simulation Library Copy and modify files from RainyPCSIM/carla/* to $CARLA_ROOT
2. Rebuild carla

   make clean
   make PythonAPI
   make launch
   

3. Launch CARLA, and open $CARLA_ROOT/Unreal/CarlaUE4/Content/Carla/Blueprints/Vehicles/BaseVehiclePawn.uasset in the blueprint editor.
4. Add component ProxyParticleSpawn.uasset. Click Add Component botton, search and add ProxyParticleSpawn component.   <p align="center"> <img src="pic/modify_base_vehicle_pawn.png" width="50%"> </p>
5. Create waymo_top LiDAR

   lidar_bp = world.get_blueprint_library().find('sensor.lidar.ray_cast')
   lidar_bp.set_attribute("lidar_type", "Surround")
   lidar_bp.set_attribute("name","waymo_top")
   

Data collection

We build our synthetic dataset using OpenCDA simulation tool.

Weather-based Intensity predictor

Please refer to Intensity_predictor.md to learn more usage about the predictor module.

ReSimAD

Please refer to ReSimAD/README.md to learn more usage about the project.

Technical Papers

@inproceedings{cai2023analyzing,
  title={Analyzing Infrastructure LiDAR Placement with Realistic LiDAR Simulation Library},
  author={Cai, Xinyu and Jiang, Wentao and Xu, Runsheng and Zhao, Wenquan and Ma, Jiaqi and Liu, Si and Li, Yikang},
  booktitle={2023 IEEE International Conference on Robotics and Automation (ICRA)},
  pages={5581--5587},
  year={2023},
  organization={IEEE}
}

@inproceedings{jiang2023optimizing,
  title={Optimizing the Placement of Roadside LiDARs for Autonomous Driving},
  author={Jiang, Wentao and Xiang, Hao and Cai, Xinyu and Xu, Runsheng and Ma, Jiaqi and Li, Yikang and Lee, Gim Hee and Liu, Si},
  booktitle={Proceedings of the IEEE/CVF International Conference on Computer Vision},
  pages={18381--18390},
  year={2023}
}

@article{yang2023rainy,
  title={Realistic Rainy Weather Simulation for LiDARs in CARLA Simulator},
  author={Donglin Yang, Zhenfeng Liu, Wentao Jiang, Guohang Yan, Xing Gao, Botian Shi, Si Liu, Xinyu Cai},
  journal={under review},
  year={2023}
}

@article{zhang2023resimad,
  title={ReSimAD: Zero-Shot 3D Domain Transfer for Autonomous Driving with Source Reconstruction and Target Simulation},
  author={Zhang, Bo and Cai, Xinyu and Yuan, Jiakang and Yang, Donglin and Guo, Jianfei and Xia, Renqiu and Shi, Botian and Dou, Min and Chen, Tao and Liu, Si and others},
  journal={arXiv preprint arXiv:2309.05527},
  year={2023}
}