Session: 14-04-01: Reliability and Safety in Transportation Systems

Paper Number: 73770

Start Time: Tuesday, 10:45 AM

73770 - Effect of Weather on the Performance of Autonomous Vehicle LiDAR Sensors 

With great progress accomplished over the past few years, autonomous vehicles are gradually entering the industrialization stage. Instant environments are built to be used for autonomous vehicle navigation by the use of algorithms to process data captured by sensors. The deployment of LIDAR imaging systems for autonomous vehicles is expanding, however, the final technology implementation is still undetermined as major automotive manufacturers are starting to select providers for data collection units to introduce them in commercial vehicles. Currently, testing for autonomous vehicles is mostly done in sunny environments where it is hard to have a good quality detection range in extreme conditions such as fog, rain, and snow. This introduces a large number of false detection alarms from the backscattered intensity, reducing the reliability of the sensor. 

In this work, LIDAR sensors were experimented in adverse weather to investigate how extreme weather affects data collection and LiDAR sensors performance. A portable testing setup was developed and utilized to investigate the effect of fog, rain, and snow on the performance of two commercial LiDAR sensors. These sensors are the Velodyne VLP-32 and Ouster OS1-32. Both sensors communicate over an ethernet connection making it easy to interface with them.  The LIDAR sensors have similar specifications, and 32 channels with approximately 120 meters of range in ideal circumstances.

 

Two types of tests were performed. The first set of tests were stationary tests. With stationary tests, we drove the car forward and stop every 25 feet. Data were collected using both LiDAR sensors at every 25 feet for 10 seconds. The second set of tests were conducted while the car moving toward the LIDAR sensors at varying speeds.

To interface with the LiDAR sensors, we utilized ROS (Robot Operating System). ROS is an industry-leading platform and was used to record the data. This includes the raw data sent to the driver and the point cloud that the driver sends out. In every test, about 1.55 gigabytes of data for each minute were recording. Recording the data allowed us to analyze it at a later time. A program to interface with ROS was developed. The program allowed us to capture and analyze the data, quantify the effect of each weather condition, and count the number of points in a region of space.

Testing results were used to provide a technology certainty for utilizing LIDAR in the commercial deployment of automated vehicles. Effective range metric technique is used to help estimate the useful range of the two LiDAR sensors in adverse weather conditions mainly in fog, rain, and snow. We consider that the metric range will fall short at communicating if there are not enough points for the algorithms to be effective. Results showed that fog severely affected the LIDARs performance followed by a limited effect of the rain. Results also showed that there was no effect of snow on the LIDARs’ performance.    

Presenting Author: Jamil Abdo Frostburg State University

Authors:

Jamil Abdo Frostburg State University
Spencer Hamblin Frostburg State University
Genshe Chen Intelligent Fusion Technology, Inc.