Jeonnam National University Develops Core Technology to Address Autonomous Vehicle Safety

Jeonnam National University Campus View.

A core technology to address the biggest challenge of autonomous vehicles-safety-has been developed at Jeonnam National University. The research team led by Professor Kim Chansu of the Department of Future Mobility at Jeonnam National University, in collaboration with Hanyang University, has developed a LiDAR-based dynamic object segmentation technology. This enables more accurate detection and response to moving objects such as people, bicycles, and vehicles, even in complex road environments.

According to Jeonnam National University on September 15, the joint research team of Professor Kim Chansu from the Department of Future Mobility at Jeonnam National University and Professor Cho Gichun from the Department of Automotive Engineering at Hanyang University has developed a core technology to enhance the safety of autonomous vehicles. This research was published in the internationally recognized journal IEEE Transactions on Intelligent Vehicles (Impact Factor: 14.3), which ranks in the top 1.8% worldwide in the field of electrical and electronic engineering, demonstrating its global significance.

The newly developed technology is a LiDAR (Light Detection and Ranging)-based dynamic object segmentation technology. LiDAR is a distance and shape measurement sensor mounted on autonomous vehicles that creates a three-dimensional map by emitting laser beams at surrounding objects and measuring the time it takes for the beams to reflect back.

The research team advanced a method to distinguish between moving objects (dynamic objects) and stationary objects (static objects) among the numerous items on the road using this technology. This is a core technology essential for autonomous vehicles to predict the movements of pedestrians, bicycles, and other vehicles in real time, and to accurately map fixed environments such as roads, buildings, and streetlights.

The research team combined LiDAR and an Inertial Measurement Unit (IMU) sensor to develop the 'AWV-MOS-LIO' algorithm. This algorithm introduced an analysis technique that considers the uncertainty (data reliability) of point data to reduce potential positional errors from sensors and errors caused by the incident angle of the LiDAR laser (errors occurring when the laser enters at an angle). Additionally, by utilizing keyframes (data from major time points), the system integrates information obtained from various angles and takes into account the size of objects to further reduce recognition errors.

Experimental results showed that this system improved the accuracy of distinguishing moving objects by 6.3% compared to existing technologies, and enhanced the odometry (vehicle position and distance estimation) performance of autonomous vehicles by 14.4%.

Professor Kim Chansu stated, "This research is significant in that it lays the foundation for autonomous vehicles to drive more safely even in complex environments. Moving forward, we plan to further develop this technology through the Jeonnam National University Regional Innovation-led University Support System (RISE) project and apply it to various automation systems."