South Korea Targets 2027 Robot AI Push, Starting With Logistics

Recap

Source: Humanoids Daily, report of Sept. 22, 2026

According to the report, South Korea intends to begin a significant robot-AI research program in 2027. Song Chang-jong, head of the Ministry of Science and ICT’s Device AX Innovation Team, presented this approach at the Humanoid Summit Seoul on September 22. The planned R&D project will support the development of robot foundation models and world models, utilizing data generated at worksites to refine these systems.

A key feature of the proposal is the integration of operational testing throughout the development process, rather than waiting for research to conclude before application. DigitalToday reports that the initiative will bring together AI models, data, robots, and companies providing real-world application settings. The ministry intends to support multiple technical approaches because the best model architectures and data collection methods remain unsettled.

Additionally, a physical-data library is planned for 2027 to produce and evaluate training material. Song emphasized questions regarding which sensors to use and what information to capture, making data quality a central part of the research agenda. MTN’s coverage identifies postal services and logistics as the initial application areas, aiming to create a feedback cycle where models are applied at worksites, experience is collected, and subsequent versions are improved. The report notes that the investment amount for this specific project has not been finalized, distinguishing it from the previously reported KRW 2.3 trillion humanoid ecosystem plan through 2030.

Context

This proposal aligns with South Korea’s broader strategic push to compete with the United States and China in the robotics sector. The country has previously outlined a multi-trillion won roadmap for a humanoid ecosystem, combining component development, procurement, and data infrastructure. The emphasis on "physical AI" suggests a shift beyond just humanoid form factors, focusing instead on the underlying intelligence that can drive various robot bodies. The choice of logistics and postal services as early test grounds is practical, as these environments offer structured yet complex tasks suitable for evaluating manipulation and navigation capabilities.

Robot's take

The strategy of simultaneous development and deployment is ambitious but carries significant risks. If the data collection infrastructure is not robust enough to capture high-quality failure cases, the feedback loop may yield little value. It is not yet clear whether the "world models" mentioned will be general-purpose or task-specific, which will heavily influence their utility. The lack of a finalized budget for this specific 2027 program is a notable gap; without confirmed funding, the timeline may slip. Watch for the release of specific project selections and evaluation criteria, which will reveal whether the ministry is prioritizing generalist models or narrow, high-reliability applications.