Crash-perching on vertical poles with a hugging-wing robot

Winged UAVs are efficient for long-distance missions but struggle to land or perch on complex vertical structures for inspection, manipulation, monitoring, or recharging. Prior avian-inspired approaches often rely on aggressive pitch-up and post-stall control maneuvers that demand precise sensing and control and operate near stall, posing risks. Mechanical perching solutions exist mainly for multicopters (e.g., claws, wrapping arms, microspines, compliant grippers, adhesives), with fewer options for fixed-wing platforms. This study addresses the challenge of enabling fixed-wing-like UAVs to passively perch on vertical poles and tree trunks without complex control or specialized feet/claws. Inspired by geckos’ head-first crash landings on trunks and broader animal strategies that encircle trunks with forelimbs, the authors propose a dual-use design: an upturned nose for passive reorientation on impact and preloaded, segmented wings that hug the pole. The research aims to determine impact conditions enabling reliable reorientation and perching, characterize static/dynamic perching performance across pole sizes and materials, and validate the approach on real trees.

Control-oriented perching research focuses on rapid pitch-up and post-stall control to reduce landing speed, sometimes using microspines for walls or hooks for cables, but these techniques require fast, accurate sensing and control and can encounter stall. Mechanical approaches are abundant for hover-capable multicopters (passive claws, wrapping arms, microspines for rough walls, modular landing gears, compliant grippers, dry/fiber adhesives, threaded anchors), but fewer exist for winged UAVs. For fixed-wing systems, adhesive-based impact attachment with tethered hanging is surface-dependent and uncharacterized in performance. Spring-loaded needles for microgliders suit extremely light robots. A passive perching claw mitigates kinetic energy for perching on small horizontal bars up to 55 mm diameter but does not scale well to larger vertical poles. Biological observations highlight crash-landing geckos and perching/climbing animals using forelimb encirclement, tails, and body orientation to manage impact and grip. These insights motivate a dual-use morphing design that avoids dedicated perching feet, leverages wing wrapping for encirclement, and uses an upturned nose to convert impact energy into inertial reorientation.

Platform design (PercHug): The UAV integrates perching functionality into flight structures without added feet or claws. The airframe is an EPP foam body (ready-to-perch weight 550 g, wingspan 960 mm). An upturned nose generates a moment about the center of gravity (COG) on head-on impact to induce pitch-up reorientation; nose tip placement relative to COG is critical. Detachable elastic nose extensions (flat carbon bars) of varying flexural rigidity (D) were evaluated to adjust moment arm and impact load distribution. Wings are foldable with three hinged segments per side; two outer segments bend ventrally to wrap the pole. Three parallel torsion springs (combined stiffness 3.45 N·mm−1) preload the joints, releasing on impact to clamp the pole. Removable hooks on the outermost segments aid engagement on rough surfaces. A tensioning wire imposes ~5° dihedral for flight and connects to a fuselage latch. The latch passively releases on impact; wall height (5 mm vs 10 mm) sets release timing (primary vs secondary impact). A backup bistable trigger beneath the latch releases at secondary impact if required by pushing a switching pad linked to linear springs and a pushing rod.

Reorientation experiments: A 220 g EPP glider (geometry per Fig. 2c) with four nose types (standard upturned foam; three elastic extensions with different flexural rigidity) was launched by a bungee catapult toward a vertical wall at speeds ~3–9 m·s−1 and pitch angles up to 25°. Motion capture (OptiTrack) tracked position and attitude; slow-motion video at 240 Hz recorded impacts. Successful reorientation was defined as achieving vertical attitude with secondary contact. Impact speed, impact angle (β), pitch rate, and impact force were estimated from kinematics; Fi was estimated via peak acceleration times mass after gravity correction.

Static perching model: A 2D planar multibody model of wing wrapping for static perching (hook-less wings) computes forces at contacts for a given pole diameter (Ø) and static friction coefficient (μ). The model assumes all segments contact the pole; torsion springs are linear; Coulomb friction; symmetric wings. It solves geometry to locate contacts, computes hinge angles and spring moments Mi = kiθi, and uses moment equilibria to determine normal (Fn) and tangential in-plane friction forces (Ft) for each segment. Vertical friction components (Fv) provide payload support. The model sweeps friction splits between planar (μt) and vertical (μv) components under the constraint μ2 = μt2 + μv2 to find feasible solutions that both prevent in-plane slipping and support weight, choosing the solution with minimum combined μ. It predicts feasible pole diameters (set by wrapping angle θ, practical at 180°) and payload capacity trends as functions of Ø and μ. Wing segmentation sizing was explored (two equal folding segments per wing) to maximize static payload for a fixed 960 mm span; 195 mm segments were selected.

Static experiments and friction measurement: A 325 g prototype with hook-less wings was placed on poles/trees with different Ø and surface materials. Calibration weights (100 g increments) were added at the COG until sliding; maximum sustained weight was recorded as static payload. Poles included paper-covered, rubber pad, paper towel, concrete (smooth/rough), bamboo tree guard, and six tree species (Ø≈250–360 mm). Static friction coefficients between EPP and surfaces were measured either by pull tests on detachable materials (μs = Fpull/mg) and incline tests (μs = tanθ) or, for non-detachable surfaces (trees), by a custom spring-loaded tool pressing an EPP block with known normal force against the vertical surface and pulling upward to slip (μ = (Fpull − mg)/(kΔl)). Each coefficient was measured multiple times to compute mean±SD.

Dynamic perching experiments: PercHug (550 g) with hooks and reinforced tail/body was hand-launched toward six trees (same as static tests). Two nose configurations were tested: standard upturned and elastic extension with D = 0.233 N·m2. The latch was configured for primary-impact release (5 mm wall); secondary-impact release via the bistable trigger was also evaluated. Trials were recorded at 240 Hz, and speed/pitch were extracted (Tracker). Inclusion criteria required meeting minimum impact angle thresholds for successful reorientation (≥15° standard, ≥8° elastic) and nose-first impact. Metrics included trigger delay from impact, reorientation and wrapping durations, impact speeds, and perching success across trees.

• Passive inertial reorientation: With the standard upturned nose, successful reorientation occurred for impact angles β above about 15° over speeds 3–9 m·s−1. Adding elastic nose extensions increased performance at lower angles; the stiffest extension (D = 0.233 N·m²) enabled success down to ~8°.
• Timing: Reorientation duration from primary impact to 90° pitch averaged 196 ± 59 ms across conditions. In dynamic tests, primary-impact latch release occurred with mean delays of 26(7) ms (standard nose) and 37(15) ms (elastic). Wrapping times in successful trials were 156(30) ms (standard) and 152(25) ms (elastic).
• Impact forces: Estimated peak primary impact forces scaled with impact speed, ranging ~15–120 N across 3–9 m·s−1. Force showed marginal dependence on impact angle for the standard nose.
• Static perching model and validation: The feasible pole diameter range scales with wingspan and is constrained by achieving at least 180° wing-wrapping. For a 960 mm span with two folding segments, predicted practical diameters were ~265–470 mm (≈28–50% of wingspan). The model predicted higher net squeezing force on larger diameters but lower maximum vertical payload as diameter increased due to reduced vertical friction component; higher surface friction (μ) improved payload. Static experiments across materials and diameters showed strong agreement with model predictions and confirmed payload trends (greater payload on smaller Ø and higher μ).
• Dynamic perching performance: Across trees and valid trials (impact speeds Vi ≈ 3–5 m·s−1; average ≈ 4.1(0.7) m·s−1; β ≥ 15°), the standard upturned nose achieved an overall perching success rate of 73%, outperforming the elastic nose (42%). Wider trees generally reduced success rates (except one case), consistent with model trends. Hooks contributed to halting downward sliding in over one-third of successful trials.
• Release strategy and tail role: Primary-impact release was crucial; secondary-impact release led to near-universal failure due to rebound/push-off preventing wing wrapping. The rigid tail reliably halted reorientation near 90°; missing tail contact often led to over-rotation and failure.
• Geometric tuning: Increasing vertical nose tip offset relative to COG enhances pitching moment at low angles, aiding reorientation; the stiffest elastic extension behaved more like a rigid extension, transmitting greater pitch moment but in dynamic perching stored elastic energy could push the robot away, reducing success.

The work demonstrates that a gecko-inspired crash-landing strategy can be realized with a fixed-wing-like UAV by transforming impact energy into a rapid passive reorientation followed by wing wrapping to perch. The upturned nose establishes robust reorientation over practical cruise speeds, reducing reliance on complex pitch-up control near stall. The static wing-wrapping model captures key geometric and frictional constraints underlying the interlocking strategy observed in animals, correctly predicting that feasible pole size is set by wing span/segmentation and that payload capacity improves with smaller diameters and higher friction. Dynamic tests confirm that, provided friction suffices to hold, tree diameter more strongly governs success than μ within the tested range, aligning with the model’s vertical friction component analysis. Timing is critical due to sub-200 ms dynamics; releasing wings at primary impact maximizes wrapping efficacy and reduces rebound. Although stiff elastic nose extensions aid reorientation at lower angles in isolation, in full perching they can store and release energy that increases separation from the trunk, reducing attachment, explaining the lower dynamic success compared to the standard nose. Overall, findings validate the passive, dual-use design approach and provide quantified guidelines (impact angle/speed, geometry, friction) for designing similar systems at different scales.

This study introduces PercHug, a bio-inspired, fully passive crash-perching method for winged UAVs that uses an upturned nose for inertial reorientation and preloaded, segmented wings for pole hugging. The approach eliminates complex near-stall pitch-up maneuvers and dedicated perching feet, achieving robust reorientation for β ≥ 15° (down to ~8° with a stiff elastic extension) at 3–9 m·s−1 and dynamic perching success up to 73% on trees with the standard nose. A validated static model quantifies feasible pole diameters (≈28–50% of wingspan for the configuration tested), shows how payload capacity depends inversely on diameter and positively on friction, and informs wing segmentation. Future work will integrate avionics and control surfaces for autonomous flight and targeting; vision-based pole detection and guidance; and a reversible, powered wing latch for controlled opening/closing. This will enable thrust-assisted climbing, unperching, and recovery maneuvers, potentially improving robustness and expanding application domains such as infrastructure inspection and environmental monitoring.

• Current prototype lacks onboard avionics and autonomous flight/targeting; all dynamic tests were hand-launched toward trees.
• No active unperching or re-flight capability; latch mechanism is one-way and relies on passive triggers.
• Dynamic perching success is sensitive to approach quality: lateral offsets, roll/yaw errors, and low impact angles can prevent nose-first impact, tail contact, or wrapping.
• The static model simplifies to 2D, assumes all segments contact the pole, linear torsion springs, Coulomb friction, and symmetric wings; it does not model hooks or micro-scale adhesion and may be less accurate outside the designed diameter range.
• Elastic nose extensions can improve reorientation but may hinder attachment in dynamic conditions by storing/releasing energy that pushes the robot away.
• Friction measurements, especially on natural bark, have variability; bark non-uniformity can introduce discrepancies between model and experiments.