One of the memorable sequences in Flight: The Genius of Birds is the starling story, where half a million birds are shown flying in close formation without bumping into each other. For more on the design implication of starling murmurations, see Dr. Timothy Standish’s comments here. It’s a stunt you can’t appreciate fully without trying it. A Hungarian team succeeded — sort of. According to Nature:

A Hungarian team has created the first drones that can fly as a coordinated flock. The researchers watched as the ten autonomous robots took to the air in a field outside Budapest, zipping through the open sky, flying in formation or even following a leader, all without any central control. (Emphasis added.)

Ten? That’s not quite 500,000, but it is a start. As we explained earlier in the context of those "termite robots," there really was central control in the engineer’s experimental design. Even so, this project, involving "biologically inspired rules," took a lot of practice. Patience, too. Earlier tests by a French team were plagued by collisions. A paper in the IEEE’s journal Intelligent Robots and Systems identified numerous challenges and tradeoffs:

The success of swarm behaviors often depends on the range at which robots can communicate and the speed at which they change their behavior. Challenges arise when the communication range is too small with respect to the dynamics of the robot, preventing interactions from lasting long enough to achieve coherent swarming. To alleviate this dependency, most swarm experiments done in laboratory environments rely on communication hardware that is relatively long range and wheeled robotic platforms that have omnidirectional motion. Instead, we focus on deploying a swarm of small fixed-wing flying robots. Such platforms have limited payload, resulting in the use of short-range communication hardware. Furthermore, they are required to maintain forward motion to avoid stalling and typically adopt low turn rates because of physical or energy constraints. The tradeoff between communication range and flight dynamics is exhaustively studied in simulation in the scope of Reynolds flocking and demonstrated with up to 10 robots in outdoor experiments.

The Hungarian team, lead by Tam�s Vicsek, modified existing software to train helicopter drones to obey three rules: (1) match the average direction of your neighbors, (2) move towards them, and (3) keep a minimum distance.

These rules — alignment, attraction, repulsion — were enough to produce a computer simulation of a bird-like flock, but real fliers face other problems. "The big enemies are noise and delay," says Vicsek. The GPS (Global Positioning System) signals are very noisy, so it is hard for the copters to accurately discern their position. The fliers also need time to receive and process those signals, and these lags mean the drones often get too close to one another or overshoot their mark.

Once they got the reaction time down, things worked better. They could make the drones fly in formation and even queue up to fly in line when faced with obstacles. A video clip in the article shows some of the successful flights.

Starlings also obey simple rules — but they move at high speeds, unlike slowly hovering helicopter drones. They also fly much closer together, and must respond within milliseconds to keep from colliding. �Additionally, they fly in 3D, with murmurations moving in several directions at once.

The robotics team surely learned to appreciate birds by trying to imitate them. Next, they need to add some eyes:

For now, the drones communicate among themselves via radio, but that sometimes leads to jammed signals. Fitting them with cameras might provide a workaround. "It’s not by chance that birds have very good vision," says Vicsek. "The big next step is to make the copters see each other."

"Not by chance?" It sounds like these designers recognize intelligent design when they see it.