Learning from nature how to land aerial robots

See allHide authors and affiliations

Science  20 May 2016:
Vol. 352, Issue 6288, pp. 895-896
DOI: 10.1126/science.aaf6605

One of the main challenges for aerial robots is the high-energy consumption of powered flight, which limits flight times to typically only tens of minutes for systems below 2 kg in weight (1). This limitation greatly reduces their utility for sensing and inspection tasks, where longer hovering times would be beneficial. Perching onto structures can save energy and maintain a high, stable observation or resting position, but it requires a coordination of flight dynamics and some means of attaching to the structure. Birds and insects have mastered the ability to perch successfully and have inspired perching robots at various sizes. On page 978 of this issue, Graule et al. (2) describe a perching robotic insect that represents the smallest flying robot platform that can autonomously attach to surfaces. At a mass of only 100 mg, it combines advanced flight control with adaptive mechanical dampers and electro-adhesion to perch on a variety of natural and artificial structures.

In nature, many birds and insects are highly skilled in precise perching on complex structures using a variety of techniques (see the top half of the figure). Large birds, for example, use visual feedback to maintain a constant expansion pattern on their retina that allows for a smooth deceleration in air to reach zero velocity just before touching the surface (3). This aerial braking is achieved through a complex deep-stall maneuver that uses morphing wings and airflow sensing (4). Simultaneously, the legs are extended and the feet attach firmly to the structure on contact through a passive grasping system called the “digital tendon locking mechanism” (5).

Smaller animals have a different perching strategy that relies more on mechanical intelligence of their body morphology. Bumblebees first hover near the structure and then use a combination of visual sensing and mechanoreceptors to slowly align their body and attach to the surface (6). Houseflies perch on ceilings without even reducing their flight speed before impact, but instead they simply extend their legs as compliant soft structures to passively damp the impact and then perch (7). Smaller insects, such as ballooning spiders, spin a silk-thread parachute as an aerial support structure to travel long distances using the wind (8). The silk thread can also act as a tensile anchoring element allowing them to perch on suspended surfaces in a completely passive manner (9). Similarly, flying seeds can be seen as ultralightweight perching structures that use exclusively passive mechanisms (for example, interlinking bristles into feathers) to attach themselves to animals for long-distance hitchhiking (10).

The field of perching aerial robotics has shown a similar pattern; higher levels of reliance on sensing and control is needed for larger vehicles, but more attention is paid to mechanical intelligence for smaller systems (see the bottom half of the figure). For example, the Stanford Univ. Scansorial Unmanned Aerial Vehicle (11) uses onboard sensors to detect a wall. It then performs a dynamic flight maneuver to pitch up and perch on the surface using passively aligned microspines on its legs. In many ways, this approach is similar to that of birds because it relies on distance sensing, dynamic flight maneuvers, and locally adaptive claws that attach to the surface via smart mechanical interlinking.

As system scale decreases, the dynamic flight accuracy is generally reduced and computational power is limited. Highly robust perching is achieved by relying on passive features enabled by mechanical processes. For example, the École Polytechnique Fédérale de Lausanne (EPFL) microglider (12) manages to perch by using a smart attachment mechanism without relying on any sensor feedback. A mechanical trigger releases two elastically preloaded spikes that snap forward and attach to the surface. The mechanism is designed such that the mechanical impulse created by the snapping movement is sufficient to decelerate the robot smoothly to zero velocity just before the pins touch the surface. Its perching technique is similar to that of the housefly that uses its legs to dampen impact and attach to the surface without executing a complex flight pattern.

Perching with tensile structures similar to ballooning spiders has been used in (13), albeit for larger vehicles of ~26 g and in combination with a fairly complex force-compensated flight controller. Nevertheless, the same principle of relying on a thread to anchor the vehicle has been used to greatly reduce control requirements and allow for a highly robust perching maneuver.

The size and mass dependence of perching.

In both nature and robotics, perching for smaller flying bodies relies on the mechanical intelligence of the body morphology, whereas larger bodies use predominantly complex control, sensing, and planning. Mass is indicated on a logarithmic scale. In the top panel, perching in biology is shown with typical masses indicated: Bumblebee (Bombus ruderatus); housefly (Musca domestica); ballooning spider (genus Stegodyphus); and Verreaux's eagle (Aquila verreauxii). In the bottom panel, perching of aerial robots is shown from the smallest to largest examples: The robotic insect of Graule et al.; the École Polytechnique Fédérale de Lausanne Perching Microglider; the Imperial College String-Based Percher; and the Stanford Univ. Scansorial Unmanned Aerial Vehicle.


The robotic insect of Graule et al. perches by hovering beneath the target structure and initiating a controlled approach to the surface. It aligns its body such that an electrostatic adhesive pad makes contact with the surface and electrically engages the pad for attachment. Electrostatic adhesive is a highly compatible solution for this scale because it is nondirectional and will work on practically any material under most environmental conditions. The robot also uses a passive polyurethane foam mount that adapts to the surface geometry of the perching substrate and balances the vehicle orientation and approach trajectory before attachment. In a similar manner, bees attach to surfaces by slowly approaching the surface and then using local adaptation with mechanical alignment structures to perch.

Perching can be achieved with complex sensing, planning, and dynamic flight maneuvers as used by birds and large aerial robots. When the system is scaled down, the use of a more mechanical perching technique can be beneficial. The mechanical insect of Graule et al. has demonstrated that mechanical intelligence can successfully be combined with complex flight control to enable robust perching behaviors. The work is also a prime example that demonstrates how engineering can learn from nature to build the next generation of aerial robots. A future direction for the field could be to selectively use environmental vectors such as wind to travel larger distances or, similar to allochory seeds, avoid aerial locomotion capabilities altogether and travel by perching on animals and larger mobile robots at no energetic cost.


View Abstract

Navigate This Article