[Edaily Reporter Yun Junghoon ] Korea Aerospace Industries (KAI)KOREA AEROSPACE INDUSTRIES(047810) is working to enhance the autonomous flight capabilities of unmanned aerial vehicles (UAVs) using AI pilot technology and is pushing to establish a future air combat system in which multiple UAVs operate alongside manned fighter jets. In addition, the company has begun developing artificial intelligence (AI)-powered UAVs capable of determining their own positions and carrying out missions even when the Global Positioning System (GPS) is disrupted by enemy electronic warfare attacks.

Lim Seong-shin, Head of KAI’s AI Development Division, stated on the 13th at the “5th Conan Defense AI Summit” held at Nine Tree Locaus in Yongsan, Seoul, “We are also conducting in-house flight tests on technology for navigation in environments where GPS is blocked.”

In future aerial combat, the autonomous flight capabilities of unmanned aerial vehicles (UAVs) are considered a key competitive advantage. This is because if GPS and communication networks are disrupted by enemy electronic warfare attacks, it may become difficult to perform missions reliably using conventional navigation systems alone. To address such battlefield conditions, KAI is developing AI-based situational awareness and autonomous flight technologies.

KAI is collaborating with the domestic AI company KONAN TECHNOLOGY INCORPORATION(402030)on various projects to develop the AI pilot “KAiLOT.”

Director Lim explained, “We are focusing our efforts on securing technologies not only for threat assessment, decision-making, and mission assignment, but also for resource management, mission monitoring, and mission automation,” adding, “Through these projects, we are working to elevate our capabilities beyond simple AI application to world-class technology.”

The future combat system envisioned by KAI goes beyond a single unmanned aircraft performing a mission; instead, it involves multiple types of unmanned aircraft dividing roles to carry out operations. First, expendable unmanned aircraft penetrate enemy air defense networks to perform reconnaissance, electronic warfare, deception, and strike missions, after which unmanned combat aircraft take over to carry out air-to-air missions and suppress the air defense network. In the rear, manned fighter jets such as the FA-50 will conduct operations in coordination with the UAVs.

Previously, KAI successfully conducted its first formation flight between the domestically produced FA-50 light attack aircraft and the AAP (Adaptable Aerial Platform), a multipurpose UAV developed in-house. During a test conducted on October 8 over Goheung, South Jeolla Province, the FA-50 and AAP flew for 20 minutes at an altitude of approximately 2.1 km, maintaining a distance of 9 to 30 meters between them. Based on the data collected from this test, KAI plans to further advance its AI-based Manned-Unmanned Teaming (MUM-T) technology and accelerate the development of unmanned combat aircraft.

Director Lim explained, “To establish a manned-unmanned collaborative system, the level of autonomy of unmanned aircraft must first be improved,” adding, “Since responsibility lies with the pilot of the manned aircraft when manned and unmanned aircraft fly together, low autonomy in unmanned aircraft inevitably increases the pilot’s workload.”

He also proposed a plan to deploy multiple low-cost UAVs using multi-purpose transport aircraft. The vision is that by leveraging 6G and low-Earth orbit satellite communications, it will be possible to create an operational environment that connects the entire battlefield—beyond land, sea, and air—into a single network. Director Lim predicted that this multi-domain network-based operational concept could become a reality by the 2030s.

Technology to ensure AI safety is also being developed. In particular, a backup system is being established that switches to a conventional rule-based control method if a reinforcement learning-based AI agent exhibits abnormal behavior.

Since standards for verifying and certifying AI safety in the aviation sector have not yet been sufficiently established, the military believes that safety measures to respond to abnormal behavior are just as important as the AI’s autonomy.

A data management system capable of continuously improving AI models was also identified as a key challenge. The explanation noted the need for a Machine Learning Operations (MLOps) system—which uses actual military operational data to retrain AI and enhance performance—as well as high-fidelity simulators that closely replicate real battlefield conditions.

Director Lim stated, “Development should not end with a one-time effort but must take the form of continuous improvement,” adding, “Ultimately, the goal is to increase the level of autonomy within a system where humans and AI collaborate.”