Aerial tactile perching via an anthropomorphic hand with embodied soft tactile receptors.

Aerial robots are widely employed for exploration, inspection, and environmental monitoring, where their agility and maneuverability are strong assets. However, their endurance remains limited, inhibiting their applicability in long-term missions. Perching, the ability to attach to environmental structures and rest with minimal power, offers a solution. Yet existing methods typically rely on bespoke attachment mechanisms tuned to a single, known, and predefined target, as well as vision-based target detection systems, prone to noise and occlusions, forcing reliance on brittle feed-forward control. We introduce a tactile-driven perching strategy for aerial robots that refines pose through touch. The system integrates a compliant anthropomorphic hand with embedded binary tactile sensors, enabling closed-loop alignment and grasp stability assessment through direct physical interaction. In simulation, the method achieves over 99% perching success across diverse geometries and pose errors up to 0.6 m and 50°. Hardware experiments validate robust perching across 26 real-world trials on diverse structures, despite corrupted pose estimates. By embedding tactile feedback into perching, this work advances a new paradigm, enabling micro aerial vehicles to exploit contact as informative feedback rather than relying solely on pre-contact visual estimates, facilitating robust autonomous perching on diverse, previously unseen targets in unstructured environments.