ETRI and Geotwo Co., Ltd. introduce ‘APPS: AI Platform for Personal Safety'

People under protective orders will soon benefit from a new technology developed by Electronics and Telecommunications Research Institute (ETRI). To address the limitations in the existing emergency response system, ETRI has unveiled APPS (AI Platform for Personal Safety), which can detect potential danger and automatically send SOS alerts, providing initial 3D location data via hybrid positioning and support drone search in large or inaccessible areas.

In recent years, requests for police protection has increased steadily, highlighting the growing need to safeguard victims of stalking and other violent crimes. Under Korea’s Anti-Stalking Act, restraining orders prohibit the restrained person from approaching within a specified distance (e.g., 100 meters) of a protected person. However, conventional monitoring devices depend primarily on GPS-only positioning, which occasionally experiences large errors (100-500 meters) in indoor, subway, or high-density urban environments. As a result, determining restraining order violations based on the GPS data not only raises privacy concerns but also suffers accuracy limitations, making reliable enforcement difficult. They also require the user to manually press an SOS button, which may not be feasible in a real emergency.

On-site search operation also present significant challenges. When officers attempt to locate a protected individual near their last known position, they currently rely on handheld close-range search devices. Yet in wide or difficult terrain, such as mountainous areas, riverbanks, or other remote environments, ground searches demand substantial time and manpower, delaying critical first-response actions.

To address this limitations, ETRI developed an integrated AI platform for personal safety. The platform combines three key technologies:

1) Al-powered threat recognition

2) Al-powered hybrid/multi-source positioning

3) AI-powered drone search

AI-powered Threat Recognition

The personal safety smartwatch, issued to individuals under police protection, continuously monitors biometric and motion data via heart-rate and inertial sensors. An AI model analyzes these signals to detect signs of danger or abnormal behavior and automatically initiates an SOS alert. To assess restraining-order violations, the system employs multi-signal similarity analysis rather than distance-base GPS data, enabling accurate assessment even in non-line-of-sight indoor environments. Trained on a dataset of more than 10,000 samples, the AI model achieved over 90% accuracy in anomaly detection and automatic SOS triggering.

AI-powered Hybrid/Multi-source Positioning

After an SOS is received, the smartwatch’s location is estimated using a combination of cellular, satellite, and short-range signals. In addition to the measurements obtained from the user’s carrier network, the system also integrates AI-inferred neighboring 4G/5G cell signals (defined as “AI-pseudo Cells”), improving positioning accuracy by a factor of three. By further fusing GNSS, Wi-Fi, BLE, and barometer data, the system achieves horizontal accuracy within 30 meters and vertical accuracy within 3 meters, maintaining precision in indoor, underground, or dense urban areas where GPS traditionally fails. This technology has been applied to the smartwatch currently in service for police protection programs in Korea and has been successfully field-tested in three metropolitan cities; Incheon, Daejeon, and Busan. The barometric sensor-based vertical positioning also enables the police to determine the user’s floor level in multi-story buildings.

AI-powered Drone Search

When the estimated location of a protected individual falls in a mountainous, riverside, or otherwise hard-to-reach area, the police can dispatch a rescue drone. The drone-mounted transmitter periodically emits Wi-Fi or BLE signals through a directional antenna, and upon request from the emergency response agencies, the smartwatch scans nearby Wi-Fi or BLE signals and transmits the measurement data via 4G/5G cellular networks. Based on the signal strength and signal direction between the drone and smartwatch, the system then uses reinforcement learning to autonomously generate optimal flight paths for rapid rescue.

With the enhanced drone capabilities including Al-based flight path generation, automatic navigation, and precision search, the system significantly reduce search time and lower the workload of patrol officers. Additionally, by combining wireless signal data with EO/IR camera imagery, it also increases overall rescue success rates. This approach is particularly effective in forested or visually obstructed environments, where visibility is limited.

The AI-powered drone search technology has been field-tested at two local police stations using police drones. The results demonstrated significant improvements in search efficiency and response time compared to conventional ground or camera-only searches.

Planned Deployment and Expansion Across Public and Private Sectors

The platform will first roll out through the police emergency response system, and will later extend to the disaster safety network, the Ministry of Justice’s electronic monitoring system, and private safety sectors. Its core technologies, including AI-pseudo cell low-power hybrid positioning, AI-powered threat detection and restraining-order violation assessment, are also applicable to healthcare, smart cities, IoT, and smart logistics, opening up strong commercial potential in addition to public benefits.

ETRI aims to advance public safety by creating an ecosystem that connects government-led public safety services with private-sector technologies. Although initially developed for individuals under police protection, the platform is built to scale to all vulnerable groups and provide broader protection across society. Its design reflects core principles of human rights, ethics, and public responsibility. To protect privacy, collection of sensitive data, such as biometric information is minimized, and only signal measurement data are stored instead of precise location information. All data are encrypted and anonymized in compliance with the Personal Information Protection Act and strictly managed within trusted national control systems.

(Overview)AI Platform for Personal Safety => https://youtu.be/PEwpO_Ga7L8?si=X7L_O7qsWf-mFGTL

###

This work supported by Protection Technology for Socially vulnerable individuals Program(www.kipot.or.kr) funded by Korean National Police Agency(KNPA, Korea) [Project Name: Development of an Integrated Control Platform for Location Tracking of Crime Victim based on Low-Power Hybrid Positioning and Proximity Search Technology/ Project number : RS-2023-00236101].

About Electronics and Telecommunications Research Institute (ETRI)

ETRI is a non-profit government-funded research institute. Since its foundation in 1976, ETRI, a global ICT research institute, has been making its immense effort to provide Korea a remarkable growth in the field of ICT industry. ETRI delivers Korea as one of the top ICT nations in the World, by unceasingly developing world’s first and best technologies.