South Korea's Ministry of Science and ICT said on June 19 it published the "Guidelines and Specification Definition for Building Autonomous Driving End-to-End (E2E) Data".

E2E autonomous driving is a method in which a single artificial intelligence (AI) model performs integrated functions from sensor input through perception and decision-making to vehicle control. It can respond more flexibly to various unexpected situations than existing methods that operate based on pre-set rules, but requires large-scale training data.

In South Korea, companies and institutions have built training data separately, but differences in sensor location and configuration by vehicle type have made data sharing and joint use difficult.

The ministry developed the guidelines to maximise the sharing and use of training data in South Korea and to respond quickly to changes in technology trends. They cover the full process of building autonomous driving E2E training data, from collection to processing, alignment and correction, and labeling. They define collection system requirements such as data-building procedures, sensor configuration and storage formats, and present methods for validating raw data.

They also include data processing methods, including driving scenario selection, location correction, spatial alignment methods, labeling items and specifications for training data sets. They also provide cases of training data built by applying the standards.

The guidelines were prepared through the "Autonomous Driving Technology Development Innovation Project" jointly 추진 by the Ministry of Science and ICT, the Ministry of Trade, Industry and Energy, the Ministry of Land, Infrastructure and Transport, and the National Police Agency. The Institute of Information & Communications Technology Planning & Evaluation (IITP), the Korea Autonomous Driving Innovation Foundation (KADIF) and the Electronics and Telecommunications Research Institute (ETRI) participated, and ETRI led the development.

After preparing a draft at the end of last year, the ministry gathered opinions from the industry and experts through a special session of the Korea ITS Society and a meeting between industry, academia and research on autonomous driving. Working with the Ministry of Land, Infrastructure and Transport, it will apply the guidelines to an autonomous driving demonstration city that builds large-scale, city-level E2E training data and continuously improve them. It will also link them with the "Autonomous Driving AI Challenge" held every year from 2024.

The guidelines can be viewed on the website of the Autonomous Driving Technology Development Innovation Project group.

Park Tae-wan (박태완), director general for information and communications industry policy at the ministry, said a foundation has been 마련 for building a shared autonomous driving data system and securing technological capabilities based on high-quality data. He said the ministry would work with related ministries to support the use of the guidelines in the field and ensure they lead to results.