MIT’s Computer Science and Artificial Intelligence Laboratory has developed new technology that helps drones dodge collisions as they fly — an AI “Artful Dodger”.
As you travel — whether human or machine — the journey is filled with uncertainties: sudden stops by vehicles, birds flying close by, road closures, sudden storms, low flying planes, rocks on the road.
Machines aren’t equipped to deal with that kind of uncertainty as they navigate with relatively static maps. But, MIT researchers have innovated a new system NanoMap that models in uncertainty — changing conditions as the drone flies.
The system includes depth sensors that constantly measure the distance between the drone and objects around it, creating an image for the machine of where it has been and where it’s going.
According to MIT Researcher Pete Florence: “It’s kind of like saving all the images you’ve seen in the world as a big tape in your head. For the drone to plan motion, it essentially goes back into time to think individually of all the different places it has been.”
The system has been successfully tested. Small drones flew through forests and warehouses at 20MPH and avoided collisions.
The project was funded in part by DARPA. NanoMap has tremendous, obvious implications for safer drone deliveries and for military missions, including surveillance.