This swarm of drones uses virtual force fields to avoid crashing into each other

The idea of a swarm of autonomous drones flying in formation is pretty darn cool. You know what is less cool? The idea of a swarm of autonomous drones all crashing into one another by accident.

That is a challenge that a new project carried out by researchers at the Georgia Institute of Technology has attempted to solve: Opening up the possibility of multiple quadcopters working together in close proximity in a way that has not been quite so straightforward previously.

“What we developed is a strategy that lets a team of quadrotors fly aggressively and efficiently while ensuring that they remain safe at all times,” Wang Li, a graduate research assistant at Georgia Tech, told Digital Trends. “In particular, we have developed algorithms that let the team stay as close to the planned, collaborative behaviors as possible, while only deviating from the planned paths as little as is mathematically possible.”

swarm drones virtual forcefield dscn3734

As described in a new paper, titled “Control Barrier Certificates for Safe Swarm Behavior,” the work involves setting up virtual barriers around each drone. If one drone enters another’s airspace, it moves away automatically. In conception, it is not all that different from computer scientist Craig Reynolds‘ classic “boids” flocking algorithm — which has been around since the early 1990s — although it also has to account for the airflow problem that comes with one drone flying too closely below another.

Li told us that the technology could be crucial to a range of drone-based applications, such as precision agriculture (in which farmers want to send teams of quadcopters to inspect a crop’s growth condition) or search, rescue, and environmental monitoring tasks (where a team of quadcopters needs to quickly respond to changing task requirements.)

The task now, he said, will be taking the drones out of the lab and into the real world. “Currently, experiments are performed in a well-structured lab environment,” Li said. “The next significant step is to implement this algorithm in an outdoor environment in combination with advanced perception technologies, similar to those used by autonomous cars.”