Interesting Engineering (TU Delft) | 2026-09-01 13:56 UTC
Each finger has a copper tactile sensor beneath its silicone surface that detects contact through changes in electrical capacitance.TU Delft
Researchers at TU Delft have developed a drone that can perch on branches using touch rather than relying solely on cameras.
The drone uses tactile sensors to detect and understand a branch as it approaches, allowing it to adjust its position before gripping the surface.
Once securely attached, the drone can shut down its motors and remain perched, reducing noise and conserving battery power.
The approach could help drones operate for longer periods in cluttered environments such as forests, where conventional vision-based systems can struggle to locate and grasp suitable perching points.
Drone That Feels
The drone uses touch sensing to locate, approach, and grip branches, allowing it to perch without depending entirely on cameras.
The system combines a lightweight anthropomorphic robotic hand with soft capacitive sensors and a tactile control system that lets the drone detect a branch through physical contact. Rather than requiring an accurate visual model of the target, the drone can begin with only a rough estimate of where the branch is and use touch to progressively refine its position and orientation.
At the center of the system is a three-fingered robotic hand designed to provide both mechanical compliance and tactile feedback. Each finger consists of three rounded phalanges connected by revolute joints, giving it a structure similar to a human finger. Torsional springs at the joints provide passive stiffness and naturally return the fingers toward a closed position, reducing the amount of active control and power required to hold a branch.
The fingers are also built from a combination of rigid and soft materials. The structural backbones of the phalanges are 3D-printed from PLA, while soft silicone interfaces provide friction and compliance when the fingers contact a branch. Embedded capacitive sensing pads are positioned across the phalanges, allowing contact to be detected at multiple points rather than relying on a single sensor.
The tactile sensors use copper foil embedded beneath the silicone surface. When a sensorized phalanx touches a conductive object, the electrical capacitance changes. An MPR121 capacitive sensing controller processes these changes and converts them into binary contact signals, giving the drone real-time information about whether and where its fingers have made contact.
The researchers use these signals as part of a closed-loop flight-control system. Instead of simply flying toward a predetermined landing point, the drone actively searches the area around its estimated target. It follows a sinusoidal figure-eight pattern while simultaneously opening and closing its fingers. The search pattern is also stepped vertically, allowing the hand to cover a larger volume around the estimated branch position.
Touch guides drone
The first tactile contact triggers a change in the drone’s behavior. Once a finger detects the branch, the system uses the contact information to determine how the drone is positioned relative to the target. It then adjusts its flight position and orientation while continuing to monitor contact across the fingers.
According to researchers, the control architecture is organized as a sequence of states, including takeoff, searching, touch detection, approach, positioning, rotation, finalization, perching, and abort. This allows the drone to respond dynamically to contact rather than treating the grasp as a single predetermined maneuver.
A particularly important part of the system is its ability to determine whether the grasp is stable. The drone waits for appropriate contact information across the hand before completing the perching maneuver. If the attempt fails, it can retreat to a safe hovering position and restart the approach.
The approach is designed around the limitations of small aerial vehicles, where additional weight and continuous motor operation can quickly drain a battery. By using passive mechanical elements, lightweight materials, and low-power tactile sensing, the hand can support the drone while minimizing the energy required for interaction.
Once the system confirms a secure grasp, the drone can switch off its propulsion motors and remain attached to the branch. This converts touch from a simple collision-detection mechanism into an active source of information for aerial navigation, positioning, and grasp control.