Scottish researchers have developed a sophisticated computer system to model the actions of simulated people lost in outdoor environments.
Illustration by Cortland Johnson | Pacific Northwest National Laboratory
Researchers in Scotland have come up with a unique computer method to predict where missing people are likely to be found in the wilderness, based on patterns of their behavior in similar situations.
The sophisticated computer system, developed by scientists at the University of Glasgow, uses real-world data on how people acted when lost outdoors.
It further operates by creating a heat map that shows the likelihood of where they may be found in any given landscape.
The innovative system not only helps search and mountain rescue teams focus their recovery efforts by incorporating sensor-equipped drones to scan the landscape, but it could also pave the way for the development of new practices in emergency situations.
Behind the study
Jan-Hendrik Ewers, a PhD candidate at the University of Glasgow’s James Watt School of Engineering and lead researcher of the project, reveals how the team went through many historic studies of how lost people behaved in real-world situations.
This helped them create so-called simulated agents, whose behavior – based on various psychological states – is driven by algorithms and guided by distinct sub-models, each designed with a specific goal in mind, whether looking for water, trees, buildings, paths or roads.
“Search and rescue teams perform vitally important lifesaving work, despite being frequently under-funded and often being crewed by volunteers,” Ewers says. “I grew up in the rural Highlands, and I’m a keen hillwalker, so I’m very conscious of both how dangerous hiking can be and what incredible work search and rescue teams do.”
The simulated agents decide where to go based on factors such as their current location and whether their preferred terrain is within sight. To help inform their actions, the system used data on where missing people are most likely to be found and how far they typically travel from their last known location.
“Initially, as part of my PhD, I set out to see whether it would be possible to use machine learning to train a new type of search and rescue system to predict where lost hikers might be found,” Ewers explains adding that that the process needs large amounts of data to reach results.
However, with little data available, since search teams prioritize saving lives over data collection, the researchers turned to existing studies on missing people to understand where they went and why. “Using that as the basis for these simulated agents has given us some really encouraging results,” Ewers says.
A statistical heat map
To validate their model, the team released AI agents from various points across a digital recreation of the Isle of Arran, located off Scotland’s west coast.
To their surprise, the probability distribution map closely matched real-world data on where missing people are usually found. This suggests the AI agents acted similarly to real people, making the system a useful tool for future search and rescue missions.
“One of the advantages of this kind of psychological modelling approach to locating missing people is that it could potentially be applied to any landscape,” David Anderson, PhD, a professor at the James Watt School of Engineering and study co-author, says. “That means it could help search and rescue teams around the world, no matter if they’re working in the mountains, jungles or deserts.”
A map of the Isle of Arran, where the team validated their model by setting AI agents loose from various locations across a digital recreation of the island. Credit: University of Glasgow
“We’re keen to explore the possibility of applying this technique to our ongoing efforts to realise the full potential of drones for search and rescue missions,” Ewers concludes in a press release. “Further development work and validation will be required before it could be used in real-world situations, but this is a promising early demonstration of the effectiveness of this kind of modelling and mapping.”
The study has been published in the journal IEEE Access.
Georgina Jedikovska Georgina Jedikovska, journalist, plant engineer, oenophile and foodie. Based in Skopje, North Macedonia. Holds an MSc. degree in Horticultural Engineering, with a specialization in viticulture and oenology. Loves travelling, exploring new cultures, a good read, great food and flavorful wines. Enjoys writing about archaeology, history, and environmental sciences.
The Blueprint Daily
Stay up-to-date on engineering, tech, space, and science news with The Blueprint.
By clicking sign up, you confirm that you accept this site's Terms of Use and Privacy Policy
