AI Will Search for Missing People Where There's No Cell Service
Around 200,000 people go missing in Russia every year. Some head out fishing or into the woods and get lost; others fall victim to crime. In the first nine months of this year, the LizaAlert search-and-rescue organization received 32,500 reports from across Russia. When a human life is at stake, every minute counts.
Researchers at Novosibirsk State University (NSU) have unveiled a system for coordinating groups of drones that could change how search-and-rescue operations are conducted. Several drones can search hard-to-reach areas simultaneously and autonomously, analyzing images onboard without an internet connection. They then return with the coordinates of any people they have found.
An Intelligent Search Team
The system combines artificial intelligence, computer vision and autonomous navigation. Using an app, an operator assigns tasks to an entire group of drones, specifying which areas each one should search. After that, the drones work independently. Each drone's camera takes a picture every second. An onboard neural network immediately analyzes each image, looking for people or vehicles. If it detects either, it records the coordinates.
When a connection is available, the drones exchange data. If communication is lost, they can still complete their missions independently and return to base with the information they have collected. A drone can also drop a small package, such as a first-aid kit or a bottle of water, to someone it has found but rescuers have not yet reached.
A Swarm of Technologies
Russia has other projects tackling a similar challenge: getting groups of drones to operate autonomously without GPS or a constant connection to an operator. Researchers at the St. Petersburg Federal Research Center of the Russian Academy of Sciences have developed a prototype system for autonomous visual navigation by drone swarms. A drone uses its camera to capture its surroundings, identifying roads, buildings and bodies of water. It then compares the image with a satellite photo loaded onto the drone in advance. By matching visible landmarks, it calculates its own coordinates. No GPS or internet connection is needed – just a camera and a preloaded image of the area.
Students at the Moscow Aviation Institute are building their own drone team. These aircraft do more than follow a preset route: they assign roles among themselves. Some are scout drones equipped with thermal imaging cameras and regular cameras to search for people. Others act as relay drones, maintaining communications within the group, while a third type serves as a command center, collecting data and deciding what to do next. If one drone goes missing or breaks down, the others reorganize and continue the mission.
The LizaAlert search-and-rescue team also has fixed-wing drones equipped with video cameras and thermal imagers, supplied by Geoscan. Beeline has provided the organization with technology support for years. Its Beeline AI neural network scans thousands of images captured by drones and automatically flags those that appear to show a person. Volunteers then only need to review the flagged images. In the past, a team of 30 people could spend five to seven hours reviewing images from a single flight. The neural network does the job in minutes, taking just two to three seconds per image.
The Future of Search and Rescue
According to LizaAlert, drones captured nearly 390,000 images in 2023 alone. The neural network processed more than 309,000 of them, saving specialists approximately 22,700 hours of work. The new system developed at NSU could speed up searches even further. Multiple drones can survey different areas at the same time, expanding the territory covered without requiring a proportional increase in the number of operators.
Beyond finding missing people, the system could be adapted to monitor forest fires, inspect extensive infrastructure and deliver supplies in emergencies. Last year, NSU had already demonstrated an algorithm in which one drone locates an object and a second delivers a payload to its coordinates.
To verify how well the system performs in real-world conditions, the researchers will conduct joint field trials with rescue services. If successful, the technology could give Russian rescue teams a powerful new ally in the effort to save lives. And that is the ultimate measure of any technology's success.
