Engineers created a paper plane-throwing bot to learn more about flight
The bot made and launched more than 500 planes with dozens of designs. Here’s what happened.
How you fold a paper airplane can determine how fast or how far it goes. A lot of people arrive at the best designs through trial, error, and perhaps a little bit of serendipity. The paper plane can be modeled after the structure of a real aircraft, or something like a dart. But this question is no child’s play for engineers at the Swiss Federal Institute of Technology Lausanne (EPFL).
A new paper out in Scientific Reports this week proposes a rigorous, technical approach for testing how the folding geometry can impact the trajectory and behavior of these fine flying objects.
“Outwardly a simple ‘toy,’ they show complex aerodynamic behaviors which are most often overlooked,” the authors write. “When launched, there are resulting complex physical interactions between the deformable paper structure and the surrounding fluid [the air] leading to a particular flight behavior.”
To dissect the relationship between a folding pattern and flight, the team developed a robotic system that can fabricate, test, analyze, and model the flight behavior of paper planes. This robot paper plane designer (really a robot arm fashioned with silicone grippers) can run through this whole process without human feedback.
[Related: How to make the world’s best paper airplane]
In this experiment, the bot arm made and launched over 500 paper airplanes with 50 different designs. Then it used footage from a camera that recorded the flights to obtain stats on how far each design flew and the characteristics of that flight.
During the study, while the paper planes did not always fly the same, the researchers found that different shapes could be sorted into three broad types of “behavioral groups.” Some designs follow a nose dive path, which as you imagine, means a short flight distance before plunging to the ground. Others did a glide, where it descends at a consistent and relatively controlled rate, and covers a longer distance than the nose dive. The third type is a recovery glide, where the paper creation descends steadily before leveling off and staying at a certain height above the ground.
“Exploiting the precise and automated nature of the robotic setup, large scale experiments can be performed to enable design optimization,” the researchers noted. “The robot designer we propose can advance our understanding and exploration of design problems that may be highly probabilistic, and could otherwise be challenging to observe any trends.”
When they say that the problem is probabilistic, they are referring to the fact that every design iteration can vary in flight across different launches. In other words, just because you fold a paper plane the same way each time doesn’t guarantee that it’s going to fly the exact way. This insight can also apply to the changeable flight paths of small flying vehicles. “Developing these models can be used to accelerate real-world robotic optimization of a design—to identify wing shapes that fly a given distance,” they wrote.