Drone Swarm Obstacle Avoidance
This new methodology enables drones to predict neighbors' movements to avoid gridlock.
By flying in a swarm, drones can cover larger areas than one alone and can be equipped with different types of sensors. When flying in a group, however, there is the risk of gridlock with one drone “piling up” against another. While this wouldn't seem extremely problematic in an open air environment, in an urban or forested area, this can pose a real obstacle.
Enrica Soria, a PhD student at EPFL’s (École polytechnique fédérale de Lausanne in Lausanne, Switzerland) Laboratory of Intelligent Systems has come up with a novel predictive control model to avoid this sort of gridlock. Here instead of simply reacting to other drones in a swarm, this new method helps each drone anticipate when its neighbor is about to slow down. This prediction is combined with an internal knowledge of its own flight dynamics, allowing group intelligence and navigation beyond a centralized control system.
The idea is explained by Soria, along with Professor Dario Floreano, in the video below as being analogous to the way birds are able to maintain a seemingly ad hoc, but beautifully coordinate flight path as well. The video also shows a demonstration of several drones navigating a simulated forest together. Perhaps search-and-rescue operation and other such aerial tasks will see this kind of multi-drone swarm behavior implemented in the future.