Whisker-based Tactile Flight for Tiny Drones

TL;DR

This paper proposes a lightweight, 3.2-gram, whisker-based tactile sensing apparatus for tiny drones, enabling them to navigate and explore through gentle physical interaction, and redefined vision-free navigation for micro aerial vehicles in extreme environments.

Abstract

Tiny flying robots hold great potential for search-and-rescue, safety inspections, and environmental monitoring, but their small size limits conventional sensing-especially with poor-lighting, smoke, dust or reflective obstacles. Inspired by nature, we propose a lightweight, 3.2-gram, whisker-based tactile sensing apparatus for tiny drones, enabling them to navigate and explore through gentle physical interaction. Just as rats and moles use whiskers to perceive surroundings, our system equips drones with tactile perception in flight, allowing obstacle sensing even in pitch-dark conditions. The apparatus uses barometer-based whisker sensors to detect obstacle locations while minimising destabilisation. To address sensor noise and drift, we develop a tactile depth estimation method achieving sub-6 mm accuracy. This enables drones to navigate, contour obstacles, and explore confined spaces solely through touch-even in total darkness along both soft and rigid surfaces. Running fully onboard a 192-KB RAM microcontroller, the system supports autonomous tactile flight and is validated in both simulation and real-world tests. Our bio-inspired approach redefines vision-free navigation, opening new possibilities for micro aerial vehicles in extreme environments.

Figures (12)

• Figure 1: Design and biological inspiration of the whiskered drone. (a) The whiskered drone platform considered here: a tiny, 44.1-g Crazyflie Brushless drone with 2 artificial whiskers. (b) Structural diagram of the whisker sensor, the STM32F0 is only used as a data acquisition and transmission for the three barometers and for temperature compensation. (c) A simplified structural diagram of the rat whisker sensing mechanism diamond2008and (Colour-Coded to Match (b)). (d) Whiskered drone wall sweeping and rat whisker sensing in darkness.
• Figure 2: Effect of whisker placement angle on drone stability during wall-sweeping interactions. (a) Schematic illustration of forces and moments applying on the drone during whisker sweeping. The whisker generates an normal force $N$, which induces a friction force $f_N$. The placement angle $\alpha$ influences the resulting pitch moment $M_{\text{pitch}}$ and yaw moment $M_{\text{yaw}}$. (b–f) Experimental setup, whisker deflection, and sweeping motion of the drone with whiskers mounted at $-15^\circ$, $0^\circ$, $15^\circ$, $30^\circ$, and $45^\circ$ relative to the yaw plane. Due to gravity, the whisker naturally tilts downward by approximately $10^\circ$ from the preset angle. The top-view inset in the lower right corner illustrates that crashes primarily occur due to the drone's inability to compensate for $M_{\text{yaw}}$. (g) Experimental results showing yaw deflection and the success or failure outcomes for different whisker placement angles at sweeping velocities of 0.1 m/s, 0.2 m/s, and 0.5 m/s. The $45^\circ$ placement exhibited the highest stability, successfully withstanding all speed conditions without inducing excessive $M_{\text{yaw}}$.
• Figure 3: Whisker-based tactile depth estimation. (a) Data collection on a rigid panel (Dataset 1). (b) Prediction error density across models for Dataset 1. (c) Predicted depth from MLP+KF (Full Model) vs. GT and laser on Dataset 1. (d) Data collection on a rigid and transparent panel (Dataset 2). (e) Prediction error density across models for Dataset 2. (f) Predicted depth from MLP+KF (Full Model) vs. GT and laser on Dataset 2.
• Figure 4: Experimental results of aerial tactile navigation through three glass walls. (a) Whiskered drone successfully navigating three nearly-parallel glass baffles. (b) Trajectories of the drone across five independent trials in the first baffle setup, with absolute position and orientation visualized. (c) Comparison of true and estimated contact depths with the corresponding threshold bands in the first setup. (d) Absolute orientation of the drone in the first setup. (e) Whiskered drone successfully navigating three differently-oriented glass baffles. (f) Trajectories of the drone in five trials in the second setup. (g) Comparison of true and estimated contact depths with the corresponding threshold bands in the second setup. (h) Absolute orientation of the drone in the second setup. The GT was derived from the initial baffle positions and the drone’s relative pose, yet its accuracy can be affected by baffle disturbances during interaction.
• Figure 5: Simulation and real-world results of active aerial tactile exploration. (a) Initial unexplored environment with maximum uncertainty. (b) Simulation in ISAAC SIM. (c1, d1, e1, f1) Successive GPIS applications after collecting training data (red: surface; blue: interior), showing reduced uncertainty and improved shape reconstruction. The drone’s trajectory (rainbow line), reconstructed contours (orange), and real-time state (drone icon) are shown. The exit is reached after the third GPIS. (c2, d2, e2, f2) Curvature analysis of extracted contours: brighter colors indicate higher curvature. Identified convex corners (white dots) are penalized, and the next exploration target is selected (blue hollow circle) based on updated uncertainty. (g) Real-world trajectory with time-color-coded path; start and landing points marked. (h) Initial environment map with maximum uncertainty. (i1) Initial exploration in four directions to gather GPIS data, with corresponding trajectory and reconstruction. (i2) Real-world curvature analysis guiding target selection. (j) Final reconstructed map with successful exit navigation; minor deviations are due to odometry drift.