Research

Establishment of Testbed Environment for Lv.4 Automated Vehicle

To accelerate the deployment of Lv.4 and cooperative autonomous driving, we are establishing a realistic test environment by utilizing MORAI SIM to simulate virtual traffic scenarios based on real vehicle data. Our lab designs and validates these scenarios through Vehicle-in-the-Loop (VILS) testing conducted within K-City to ensure the safety and performance of autonomous services under realistic conditions. This allows for iterative validation and feedback in a real-world living lab environment, supporting the reliable deployment of Lv.4/4+ autonomous and cooperative driving technologies. 

Development of a Technology to Enhance Automated Driving Using Infrastructure Guidance

This research project develops cooperative autonomous driving technology to improve traffic flow and safety in mixed road environments. By leveraging real-time perception between vehicles and infrastructure, it enables behavior analysis and optimal driving guidance. Our lab focuses on deep learning-based maneuver prediction and trajectory prediction with robust, real time multi-task models.

Development of safety of the intended functionality from insufficiency of perception and decision making

Safety of the Intended Functionality (SOTIF) aims to minimize risks associated with unknown or undefined hazardous situations. This involves identifying potential hazards, evaluating risks, and implementing safeguards so that the system operates safely under all intended conditions. Our lab is conducting a project with the goal of building a system and service that can automate SOTIF testing based on various use cases.

Development of AI-based Collision Mitigation and Avoidance System for Urban VRUs and Micro-mobilities

This study aims to develop an AI-based collision avoidance and mitigation system for urban autonomous driving, capable of predicting potential accidents involving pedestrians, cyclists, and micromobility users. The system goes beyond automatic emergency braking (AEB) by executing collision avoidance strategies to either prevent collisions or reduce the severity of injuries. This enables autonomous vehicles to respond effectively to a wide range of traffic situations in real-world urban environments. 

Scenario for Developing a Redundant Architecture for Collision Risk Minimization and Scalable Learning

This research is an industry-academia collaboration with Hyundai Motor Company, aiming to develop a deep learning-based algorithm for collision mode classification and collision avoidance decision-making. We define and implement a framework to enhance the performance. A learning-based method is applied to improve algorithm accuracy, and an additional strategy is introduced to reduce false predictions. 

Development of End-to-End Autonomous Driving Technology for
Mass-Produced Vehicles Equipped with All-Directional Sensors

Soon