Domestic researchers have embarked on the development of next-generation autonomous driving technology in which artificial intelligence learns driving strategies based on actual road driving data. As automakers, research institutions, and specialized autonomous driving firms collaborate to pursue technology development with the goal of applying it to actual vehicles, the industrial applicability of domestic autonomous driving software is being put to the test.

The Electronics and Telecommunications Research Institute (ETRI) announced on the 17th that it has entered into cooperation with KG Mobility and Sodis, a company specializing in autonomous driving, to develop next-generation end-to-end (E2E) autonomous driving artificial intelligence technology. A signing ceremony for a letter of intent was held at ETRI in Daejeon on the 16th, and the three organizations agreed to jointly pursue technology development, demonstration, and industrial application.

The core of this research lies in implementing 'driving intelligence' in which artificial intelligence comprehensively understands the road environment and independently determines the vehicle's steering, acceleration, and deceleration. It is designed to determine driving strategies in a human-like manner by utilizing large-scale driving data and vehicle movement information collected from actual roads for learning.

Conventional autonomous driving technology typically features a structure where perception, decision-making, and control functions are processed step-by-step by separate systems. In contrast, this research applies an end-to-end approach where a single integrated AI model interprets road conditions and simultaneously determines vehicle steering and speed control. Furthermore, methods are being explored to enhance decision-making accuracy even in complex road environments by incorporating a multimodal AI structure that understands both visual information and situational context.

The research team is also pursuing an approach that reduces reliance on expensive LiDAR sensors and combines camera-centric visual information with artificial intelligence decision-making capabilities. Through this, the goal is to develop an intelligent driving model capable of stable autonomous driving with a minimal sensor configuration.

The three institutions plan to proceed in stages with data acquisition in real-world road environments, AI model training, and demonstration of application in actual vehicles. The research team envisions expanding this technology into a general-purpose mobile intelligence technology applicable not only to automobiles but also to various mobile devices such as robots and drones.