MIT Develops Autonomous Drone Fleet That Works Together Without the Need for GPS

It’s not uncommon for hikers or campers to get lost while in a forested or an unfamiliar area, and while it’s not hard to become lost, it’s difficult for rescuers to search areas with difficult terrain or those with thick tree canopies. According to a 2009 report from the Wilderness Medical Society, there were 65,439 SAR (Search and Rescue) incidents in US national parks from 1992 to 2007. Out of those operations, there were 2,659 fatalities, 24,288 that were ill or injured, and 13,212 saved.

To help reduce the number of people that go missing or become lost, researchers from MIT are developing a fleet of autonomous drones that can search large areas in remote locations, even if they’re heavily forested. Traditional drones rely on GPS and other systems for navigation, which have difficulty receiving signals while under thick tree cover, like trying to get s cell signal in rural areas. With that in mind, the MIT team is designing a fleet of autonomous drones that overcome that issue by taking GPS out of the equation.

Each drone is outfitted with laser-range finders, which is used for position estimation, localization, and route planning for navigation. As they fly, a mounted LIDAR system creates a 2D scan of the environment — identifying trees, boulders, and other objects in that area.

The drones generally fly in loops in any given area, and send the 2D scans they have taken to a ground-based station via Wi-Fi, where is uses SLAM (Simultaneous Localization and Mapping) to piece together the images to create a 3D rendering of the area. It’s also used to keep track of drone positions inside the area being mapped. If the drones locate a missing person through its mapping process, it will geo-tag that person’s location and display it to rescuers on the ground.

The researchers have tested multiple drones in both simulations of randomly generated forests, as well as those at NASA’s Langley Research Center and found that the drones were capable of mapping an area of about 20-square meters in about 2 to 5-minutes while stitching the maps together in real-time.