Rescue teams are increasingly relying on ultra-compact drones to locate survivors in collapsed buildings, but these small machines face a critical limitation. When smoke obscures the air or darkness falls, standard cameras and lidar sensors stop working. A research team at Delft University of Technology in the Netherlands has addressed this problem by equipping tiny drones with tactile sensors inspired by the whiskers of nocturnal animals.
Biological Inspiration for Tactile Navigation
Ultra-compact drones weigh less than a stick of butter, which restricts their computing power and sensor capacity. Larger drones can carry heavy lidar units, but these small models cannot. The new system uses delicate wires made of nickel-titanium, each measuring 0.4 millimeters in diameter. These antennae are longer than the drone itself, extending nearly 8 inches from a palm-sized body.
Two whiskers are attached to the drone, each equipped with three pressure sensors at the base. These sensors detect the bending of the wires, allowing the drone to estimate its distance from nearby barriers. All calculations happen on an onboard microcontroller, eliminating the need for external computers. Salwa Hamaza, a professor at Delft University, noted that tactile sensing for drones must be lightweight, low-latency, and low-power to be effective.
Testing Flight Angles and Drift Correction
Mounting long, delicate wires on a flying machine introduces stability challenges. Resistance from the whiskers can cause the drone to rotate and crash when moving along walls. The researchers tested 72 flights across six different mounting angles and three speeds. They found that angles of 30, 45, and 60 degrees allowed for stable flight. The 45-degree angle was selected because it caused the least deviation in orientation.
Processing the sensor data presented another hurdle. Propeller vibrations created noise that the system initially mistook for wall contacts. This drift caused a 38 percent false positive rate. The team developed a correction algorithm that estimates drift trends based on sensor readings when the drone is not touching anything. This filter reduced false positives to zero percent.
In tests with transparent walls, the drone maintained a distance of 60 to 100 millimeters from surfaces. It successfully navigated angled walls, a task that would confuse laser-based sensors. The system also used probabilistic estimation to explore unexplored areas, finding a 2.5-foot-wide exit in 90 percent of tests in a small room with glass walls.
Real-World Application and Memory Efficiency
While the tests used neat geometric shapes, real disaster zones contain rubble and irregular obstacles. The researchers encoded an exploration method that prioritizes moving through unexplored areas. This approach helps the drone avoid getting stuck in corners or missing side barriers. The entire tactile perception pipeline uses only 34 KB of memory, a small fraction of the drone’s 192 KB capacity.
Chao-Shan Ye, a doctoral student at Delft University, stated that the goal was to show that touch does not have to come at the cost of size or computational power. The system allows a tiny drone to sense and respond to its environment in real time. This technology could significantly improve the ability of rescue workers to find survivors in environments where visual sensors are useless.
Source: Wired
