Researchers at the University of Glasgow have developed an innovative computer system that could revolutionize search and rescue missions in remote areas. By analyzing patterns in how real people behave when lost, this new method predicts the most likely locations where missing individuals might be found—without relying on guesswork or chance.

The technology, led by PhD candidate Jan-Hendrik Ewers from the James Watt School of Engineering, uses artificial intelligence to simulate the actions of lost individuals. These “simulated agents” behave according to psychological models and real-world data, factoring in needs like finding water, shelter, paths, or roads. The result is a detailed heat map showing high-probability search areas across a given landscape.

Drawing from historical case studies and behavioral research, the team developed algorithms that mimic how lost persons typically move through unfamiliar terrain. These virtual agents were released in a digital replica of Scotland’s Isle of Arran to test the model. Impressively, the AI-generated probability maps aligned closely with actual data from previous search and rescue operations on the island.

The model’s accuracy opens the door to faster, more focused rescue missions, particularly when time is critical. This system can integrate with sensor-equipped drones, offering rescue teams a bird’s-eye view of areas where someone is most likely to be found. Such an approach could dramatically improve the efficiency of operations in mountains, forests, deserts, and even jungles worldwide.

Ewers, who grew up in the Scottish Highlands and is an experienced hiker himself, emphasized the importance of supporting search and rescue teams, often made up of under-resourced volunteers. “The idea started as part of my PhD—to explore whether machine learning could assist these teams. We didn’t have access to large datasets because rescuers prioritize lives over data, so we turned to existing studies to guide our AI.”

Professor David Anderson, co-author of the study, highlighted the broader potential of this system. Because the psychological models are adaptable, the same technology could be used in a variety of landscapes globally—not just rugged highlands.

Though further testing and development are needed before deploying this in real-world emergencies, the early results are promising. The team now hopes to integrate the technology more deeply with drone systems, offering a high-tech boost to traditional search and rescue methods.

This pioneering work offers a glimpse into a future where AI, behavioral science, and drone technology unite to save lives—helping rescuers find the lost faster, smarter, and safer.

By Impact Lab