Ewers’ algorithm outperformed both the lawnmower and existing algorithmic approaches in virtual testing, showing superior results in two key metrics: the distance a drone had to fly to find the missing person, and the percentage of time the person was located. While the lawnmower and existing algorithm found the person 8% and 12% of the time, respectively, Ewers’ approach successfully located them 19% of the time. If this success translates to real rescue situations, the new system could significantly reduce response times and save more lives in critical scenarios.
“The search and rescue operations in Scotland are diverse and often hazardous,” Ewers explains. Emergencies can occur in dense forests on the Isle of Arran, the rugged mountains and slopes around the Cairngorm Plateau, or the challenging terrain of Ben Nevis, a popular but perilous rock climbing location. “Using drones for efficient search operations could potentially be life-saving.”
Experts in search and rescue suggest that employing deep learning to optimize drone routes could expedite the process of locating missing individuals in various wilderness environments, depending on the suitability of the terrain for drone exploration (dense canopy poses more challenges compared to open brush, for instance).
“This strategy seems promising in the Scottish Highlands, especially during the initial stages of a search when waiting for additional resources,” notes David Kovar, a director at the US National Association for Search and Rescue in Virginia, who has utilized drones for disaster response in California and wilderness search missions in New Hampshire’s White Mountains.
However, there are potential drawbacks. The efficacy of the planning algorithm will rely heavily on the accuracy of the probability maps. Over-reliance on these maps could lead to drone operators spending excessive time searching in the wrong areas.
