Landing and Perching on Vertical Surfaces with Microspines for - PowerPoint PPT Presentation

Landing and Perching on Vertical Surfaces with Microspines for Small UAVs Alexis Lussier Desbiens and Mark Cutkosky Biomimetics and Dextrous Manipulation Laboratory Stanford University UAV’09, Reno

Why should we perch? 2 10 • Small airplanes have limited endurance Aerosonde Fixed � wing • Other techniques can only Rotary � wing 1 10 provide limited benefit: Flapping � wing Rotomotion SR200 Endurance (hour) – Energy extraction from Raven B gusts [Patel & Kroo WASP � III 0 Black Dragon Eye 10 2006] Widow – Optimization [Grasmeyer Delfly II Draganflyer UofF Skybotic 2001] CoaX Gator � 1 10 Delfly Micro • Perching would increase mission duration to extended period of time (i.e. days or weeks) � 2 10 0 1 2 3 4 5 10 10 10 10 10 10 Mass (g) 3

Other Advantages of Perching • Stable vantage point while perched vs fast dynamics of small UAVs during flight • Possibility of landing and physically interacting with the landing surface. • Perching combines the best of climbing and flying: – Agile and fast while flying – Can cover long distances – Limited energy consumption while perched – Wait for better weather conditions – Quiet (no motor noise) 4

Why Vertical Surfaces? • Walls provide a large surface to • Walls are common in urban perch on environment • Walls remain relatively free of • Walls are easy to detect (at least debris. easier than a passive wire or pole) 5

Related Work • On agile flight: – J. How et al. (MIT) on indoor flying and [Cory & Tedrake, 2008] hovering – P. Oh et al. (Drexel) on autonomous hovering • On perching aerodynamics & control: – Wickenheiser et al. (Cornell) on vehicle morphing for perching [Wickenheiser, 2007] – Tedrake et al. (MIT) on controllability of fixed- wing plane for perching on a wire • No explicit consideration of the landing system • Slow maneuvers sensitive to disturbances • Use of highly accurate motion capture system/sensors to enable control 6 [Green & Oh, 2006]

Research Goals Allow a small airplane to perch autonomously on a variety of vertical surfaces Keep the system simple and lightweight Maintain the efficiency of conventional airplanes 7

Sonar Our Approach Spines • Quick maneuver to minimize Paparazzi disturbance effects Autopilot Suspension & sensors • Focus on the suspension and spines to simplify sensing and control Modified Flatana • Everything onboard! Airplane 4) Touchdown 2) Wall detection 1) Approach 5) Rest 3) Pitch up Elevator 8

Sticking to the wall • Small spines (10-15 µm tip radius) that catch and hang on asperities • Individual spine suspensions distribute the load Approach volume • Why spines? – They require no power – They work on a wide range of outdoor surfaces – They are relatively unaffected by films of dirt and moisture 1 – They leave no trace of their passage 4 5 – They provide directional adhesion 3 (multiple loading cycles) x 2 y 9 Loading Forces Volume

Spine Limit Surface • Spines require a particular loading cycle to engage asperities on the surface without slipping or failing Shear force Coulomb Compression friction 1/ µ 1. Normal force 2. Pull down Surface failure Normal force � load Loading 3. Pull away Adhesion cycle Different angles of surface asperities 10

Spine Performances • Used on Spinybot and RISE to climb brick, stucco, concrete and rock • Climbing robot spine suspensions take advantage of the robot's control over foot trajectories and forces • With UAVs, the challenge is to provide desired forces and velocities at the instant of contact with the wall 11

Suspension • Tests reveal that vertical rebound was the main failure mode • Goal is to find the optimal components (spring, damper, nonlinear elements) to: – Minimize peak landing force – Minimize suspension travel – Prevent negative force, to stay on the wall (vertical rebound) • Maximum energy dissipation achieved with constant force during the impact 12

g Simple Model K B F M • Vertical model of a spring, v 0 damper and coulomb 0 Position Airplane (m) friction suspension � 0.01 • Damper creates forces � 0.02 dependent on initial � 0.03 velocity 0 0.05 0.1 0.15 0.2 40 • Coulomb friction provides Force on spines(N) Fric = 33 N � = 0.62 constant force 20 � = 0.29, Fric = 11 N • Balance friction and 0 Spine rebound region damper to get desired � 20 0 0.05 0.1 0.15 0.2 time (sec) properties 13

Non-linear properties • Material properties can be used to create 35 Rubber CONFOR foam constant force 30 Viscous Damping (Ns/m) • Damping scaled w.r.t 25 position and velocity: 20 b = F max − kx ( t ) 15 v ( t ) 10 • Urethane foam exhibits reduced damping at 5 0.05 0.1 0.15 0.2 0.25 0.3 Max. Velocity (m/s) high velocity 14

Suspension Designs k* = 871 N/m � * = 0.38 k* = 780 N/m � * = 0 k* = 695 N/m � * = 0.15 15

Leg Structure Attachment Foam Foam points ankle hip Spines Carbon tibia Sorbothane Balsa/Carbon knee femur 16

Planar Model " & ! ! • Simple cartesian suspension + model for now... ,3#$1 4/50640 2 • Wing and control surfaces modeled as flat plates [Cory & ,- !.//0 #$1 !.//0 2 ! '( Tedrake 2008] ! & " ! • Equations of motion generated '! ) !' using Kane � s Dynamics ) *!") • Used to study the effect of: • Incoming velocities • Suspension parameters (foot location, linkage non-linearity, etc.) • Improve landing forces on the " !#$ % ! spines Intermittent contact

Planar Landing Forces on spines during landing 30 Simulation 25 • Loading trajectory is 20 important Adhesion limit surface 15 • Low damping ratio: Normal Force (N) – Ratio Fn/Fs too high 10 – Rebound 5 • High damping ratio: Steady state � = 0.8 – High peak force 0 Rebound • Moderate damping: Initial � = 0.5 Contact � 5 – Ratio Fn/Fs within adhesion limit surface � 10 � = 0.1 0 5 10 15 20 Shear Force (N)

Trajectory/Control • Previous research focuses on low contact velocity: – Low controllability at low velocity – The longer the approach, the riskier it gets (gust, etc) • Spines need normal force to engage, we want forward Pitch up maneuver Low angular velocity ! velocity • Use the dynamics of the plane to reach the Feedforward successful perching envelop kick by the of the suspension elevator Touchdown possible 19

Perching Strategy 1. Fly toward the wall at about 9 m/s 2. Detect the wall with ultrasonic sensor • 20 Hz, 6 m range 3. Pitch up to slow down for landing (take about 2-3m) 4. Touchdown possible for about 1.5 to 2 m before impact 5. Touchdown at about 2 m/s, let the suspension absorb the impact Simulated trajectory of the perching maneuver 0 � 0.2 y (m) � 0.4 � 0.6 � 0.8 Waiting for wall detection Pitching up Successful landing � 6 � 5 � 4 � 3 � 2 � 1 0 x (m) 20

Onboard Sensors • Simple wall detection using the LV-Maxsonar: – Range of 6 m – Update rate of 20 Hz • Onboard accelerometer and Different techniques for measuring pitch 80 gyro are used for data Complementary Filter Rate Gyro Integration Gravity measurement analysis 60 40 • Combined using a second Sensitive to vibrations Pitch (deg) order complementary filter: 20 0 � 2 τ 2 s � τ s + 1 θ ( s ) + 2 τ s + 1 ( τ s + 1) 2 ˙ θ ( s ) = ( τ s + 1) 2 θ ( s ) � 20 τ s + 1 Drifting 65 70 75 80 85 90 • Need something better!!! time (sec) 21

22

Touchdown Elevator possible up 9 m/s Wall 2 m/s detection Pitch up maneuver x y 30 successful landings (10 autonomous, 20 in manual control)... out of 40! • Pitch = 60 to 105 deg • v x = up to 3 m/s • Pitch rate = 0 to 200 deg/s • v y = up to 2.7 m/s (downward) 23

Landing Data 100 Pitch (deg) 50 Pitch up maneuver 0 • Wall detection at 6m 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 time (sec) Wall Distance (m) • Maneuver duration 6 4 of less than 0.7 sec Wall 2 detection 0 • Ready to perch 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 time (sec) starting at t = 1.3 sec Touch Down 8 Throwing v x (m/s) 6 • Lands at 1 m/s 4 Autonomous flight 2 0 • Most of the elevator 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 time (sec) action happens at 1 Moment (Nm) Trim flight 0.5 Elevator up low angle of attack 0 � 0.5 0 20 40 60 80 100 120 140 160 180 Angle of attack (deg)

Improvement and future work • Land on more challenging surfaces • Trajectory optimization – Maximize horizontal distance travelled while ready to perch – Add propulsion • Real conditions landing (windy, side approach, etc.) • Take off from the wall!! 25

Conclusion • A properly tuned mechanical system simplifies the perching maneuver • Suspension is essential for: – Proper spine engagement – Maintaining controllability – Reducing control & sensor requirements • Only 7% (28g) of total airplane mass • Perching is interesting for a wide range of applications • Perching is pretty cool! 26

Questions? http://bdml.stanford.edu - alexisld@stanford.edu 27