Countless fields, from military operations to environmental research, rely heavily on drones. Executing a safe landing is still one of the biggest challenges, especially when conditions are far from perfect.
Researchers have already come up with some pretty creative solutions. Learning from the animal world, they developed drones that can land on steep, icy surfaces using technology inspired by cat claws, or that slow down for a landing by extending flying-squirrel-inspired wing membranes. But landing on tree branches is even more complex, so it has remained an unresolved challenge.
Wouldn’t it be easier to just keep the drone hovering in the air, then? Well, sometimes that’s exactly what researchers need, but some other times it can get in the way, mainly for two reasons: hovering drains the battery pretty quickly and produces a constant, noticeable noise. When it comes to observing wildlife, researchers need something closer to a spy camera: unobtrusive and quiet enough that animals and birds don’t get scared away.
Some drones already use claw-like grippers to grab onto tree branches. But gripping a branch is only half the challenge. The drone also needs to know exactly where the branch is, and that’s when things get tricky.
When landing, drones typically rely on cameras to determine where and how to land. But in a forest, branches and leaves can obstruct the camera’s view. That’s why researchers at Delft University of Technology (TU Delft) in the Netherlands took a different approach: what if a drone could use tactile sensing instead of relying entirely on its cameras? They drew inspiration from birds, which use their feet to feel a branch before landing on it.
The new prototype features three flexible “fingers,” each made up of three sections – or phalanges – just like a human finger. This design provides flexibility to wrap around a branch and also creates a larger contact area.
At first, the drone still uses its camera to estimate the location of the branch. But then, instead of completely trusting the camera, it reaches toward the branch with its “fingers” and literally begins feeling for it until physical contact is made.
Each finger has three tactile sensors in its soft silicone surface. Once one of the sensors touches the branch, it sends an electrical signal to the drone, essentially providing a simple yes-or-no answer to the question, “Am I touching the object?”.
Because the drone knows which of its nine sensors has been triggered, it can improve an estimate of the branch’s location, size, and position. It’s pretty much like trying to find and recognize an object with your hands while keeping your eyes closed.
“Each sensor only tells us 'something is here.' But if you know the shape of your own hand, that is enough to reconstruct where the branch is and how it is oriented. The drone builds up an increasingly accurate 'picture' of the target purely by bumping into it,” explains Anton Bredenbeck, one of the authors of the project.
Once contact is made, the drone adjusts its grip, wraps its fingers around the branch, locks them together, and holds its position securely. Gripping onto the branch doesn’t consume any additional energy thanks to the special springs built into the fingers.
Tests of the system showed impressive results: the drone achieved a landing success rate of more than 99%. Even when its initial visual estimate of the branch location was off by as much as 60 cm (about 24 in), the drone was still able to correct itself and complete a safe landing.
The ultimate goal of the project is not to completely replace cameras when landing drones, but rather to complement them with tactile sensing, especially when conditions are challenging and visual information is less reliable.
A paper on the study, which was led by Assoc. Prof. Salua Hamaza, was recently published in the journal Nature npj Robotics.
Source: TU Delft
