Robo Raven:

A Flapping Wing Air Vehicle with Compliant

and Independently Controlled Wings

Satyandra K. Gupta

Director, Advanced Manufacturing Lab

Director, Maryland Robotics Center
Mechanical Engineering Department and
Institute for Systems Research
University of Maryland, College Park
 Research Interests

• Computational Foundations for Automation

─ Computer Aided Design
─ Manufacturing Automation
─ Robotics
• Current Focus
─ Exploiting synergy between robotics and manufacturing
 The Role of Robotics in
Advanced Manufacturing
• Robots have been used to
improve manufacturing on
high volume production
lines
─ Reduce labor cost
─ Increase production rate
─ Increase quality (Image Source: ATACO Steel Products)
• Use of robots is currently
very limited

New opportunities for deploying robots in manufacturing

Robotic Assembly at Small
Size Scales

Assembly at Microscale: Assembly at Mesoscale:

Optical Micromanipulation In-Mold Assembly

Robots look and behave differently

 Human Robot Collaboration in
Bin Picking and Assembly
• Bin-picking precedes
assembly in many low volume
production scenarios
• Challenges
─ Random part postures, overlaps,
occlusions, background clutter,
shadows, poorly lit conditions
• Approach

─ Robot does bin-picking and assembles each part to build the product
─ Human assists robot in critical situations by (1) resolving perception
and/or grasping problems encountered during bin-picking and (2)
performing dexterous manipulation required during assembly
 Learning from Human
Demonstrations
• Approach allows learning from
successful human demonstrations,
errors made by humans, and how
humans recovered from these errors
in subsequent trials
• Classifiers and iterative search to
generate initial task parameters for
robot
• If robot fails, simple rules are learned
to refine them by capturing how
humans change parameters to
transition from failure to success
J. D. Langsfeld, K. N. Kaipa, R. J. Gentili, J. A. Reggia, and S. K. Gupta. Incorporating failure-to-success
transitions in imitation learning for a dynamic pouring task. Workshop on Compliant Manipulation: Challenges
and Control, held at IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2014),
Chicago, IL, September 18, 2014.
 Ensuring Human Safety
in Hybrid Cells
• Real-time replication of human
and robot movements inside a
physics-based simulation of
the work cell
• Multiple Kinects based system
to track and model human
• Roll-out strategy
─ forward-simulate robot’s trajectory and create temporal set of its
postures for next few seconds
─ Check whether any of these postures collide(s) with human model
• Pause robot’s motion whenever imminent collision detected
Morato, C., Kaipa, K.N., Zhao, B., and Gupta, S.K. (2014). Toward safe human robot
collaboration by using multiple Kinects based real-time human tracking. Journal of Computing
and Information Science in Engineering, 14(1):011006-011006-9.
 See Videos
https://www.youtube.com/watch?v=qg0kjzrag1I
https://www.youtube.com/watch?v=X68F9p8DMLg
 The Role of Advanced
Manufacturing in Robotics

• Additive Manufacturing
• Polymer Composites
• In-Mold Assembly
• Microfabrication

Stratasys 3D Printer

Advances in manufacturing can be used to realize

improved robot designs
How to Realize Bio-Inspired
Robots?

NaviGator RoboTerp ScaleBot

R2G2 RoboCrab RoboRaven

 R2G2:
Robot with Rectilinear Gait for Ground Operations
• Goal: Create a limbless robot with
high forward velocity and small cross
section
• Approach
─ New Rectilinear Gait
─ Compact Parallel Mechanism
─ Additive Manufacturing
─ Design Details
 Cross Section: 70 x 70 mm
 Length (Contracted): 1000 mm
 Length (Extended): 1385 mm
 Weight: 2.5 kg
See Video at
https://www.youtube.com/watch?v=t4oQos_LYyc
J.K. Hopkins and S.K. Gupta. Design and modeling of a new drive system and exaggerated rectilinear-gait for
a snake-inspired robot. ASME Journal of Mechanism and Robotics, 6(2):021001, 2014.
 Robo Terp

• Goal: Develop a legged

amphibious robot for splash free
swimming
• Approach
─ Incorporate compliance in legs to
assist swimming
─ Optimize leg design
─ Develop new gaits for walking,
swimming, and transitioning
─ Develop sensors for autonomous gait
transitions
See Video at
https://www.youtube.com/watch?v=X9IV7QDcJRg
A. Vogel, K.N. Kaipa, G. Krummel, H.A. Bruck, and S.K. Gupta. Design of a compliance assisted quadrupedal amphibious robot.
IEEE International Conference on Robotics and Automation (ICRA 2014), Hong Kong, China, May 31-June 7, 2014.
 Robo Crab

• Goal: Develop a self-righting

robot Inspired by horseshoe
crabs
• Approach
─ Built for surf zone traversal
─ Walks on sand, in water
─ Self-rights when tipped over
─ Fully waterproof

See Video at
https://www.youtube.com/watch?v=L-J3NW3sXY8\
G. Krummel, K.N. Kaipa, and S.K. Gupta. A horseshoe crab inspired surf zone robot with righting capabilities.
ASME Mechanism and Robotics Conference, Buffalo, NY, August 2014.
 SCALE Bot

• Goal: Develop a low-cost robot

capable of autonomously
climbing stairs using on-board
sensing and computation
• Approach
─ Develop twelve degree of freedom
legged robot using off the shelf
actuators, sensors, and controllers
─ Develop parameterized gaits to
climb stairs
─ Develop algorithms to process
sensor data to select gait
parameters
See Video at
https://www.youtube.com/watch?v=BDSzO8mhOuY
Flapping Wing Air Vehicles
 Robotic Birds:
Motivation
• Attributes of fixed wing flight
─ High forward speeds required for generating lift
─ Low maneuverability
─ Difficult to operate in confined spaces

• Attributes of rotary wing flight

─ Low forward speeds and hovering possible
─ High frequency leads to noisy operation

• Attributes of flapping wing flight

─ Low frequency flapping leads to quiet flight
─ Low forward speeds lead to high maneuverability
─ Bridges gap between fixed and rotary wing
 First Effort: Small Bird
(2005-2007)
• Goal: Develop a lightweight and
efficient drive mechanism to
transmit power from the motor to First Flight
wings in 2007
• Approach
─ Develop a new compliant
mechanism concept
─ Develop multi-piece molds to
realize the light weight compliant Weight: 9.7 g
mechanism (excluding battery)
─ Optimize shape Wing Span: 34.3 cm
Flapping Frequency: 12.1 Hz
─ Incorporate multi-functional Pay Load Capability: 5.7 g (including battery)
materials for dissipating heat
D. Mueller, H.A. Bruck, and S.K. Gupta. Measurement of thrust and lift forces associated with drag of compliant flapping wing for micro air vehicles using a new test stand
design. Experimental Mechanics, 50(6):725–735, 2010
W. Bejgerowski, A. Ananthanarayanan, D. Mueller, and S.K. Gupta. Integrated product and process design for a flapping wing drive-mechanism. ASME Journal of Mechanical
Design, 131, 2009
W. Bejgerowski, S.K. Gupta, and H.A. Bruck. A systematic approach for designing multi-functional thermally conducting polymer structures with embedded actuators. ASME
Journal of Mechanical Design, 131(11): 111009, 2009
 Big Bird with Folding Wing
(2007-2008)
• Goal: Generate static lift by folding
wings during up-strokes
• Approach
─ Increase the size of the platform to
enhance payload capacity
─ Incorporate on-way joints in wings to
facilitate passive wing folding in up-
stroke Weight: 29.9 g
─ Develop new joint designs based on the (excluding battery)
distributed compliance concept Wing Span: 57.2 cm
─ Optimize wing design Flapping Frequency: 4.5 Hz
Pay Load Capability: 17.0 g
(including battery)
First Flight in 2008
D. Mueller, J. Gerdes, and S.K. Gupta. Incorporation of passive wing folding in flapping wing miniature air vehicles. ASME Mechanism and Robotics Conference, San Diego,
2009
 Jumbo Bird
(2009-2010)
• Goal: Increase payload Weight: 38.0 g
(excluding battery)
capacity of the robotic bird Wing Span: 63.5 cm
First Flight
• Approach in 2010
Flapping Frequency: 6.1 Hz
Pay Load Capability: 33.0 g
─ Develop a new transmission
(including battery)
mechanism based on multi-
material compliant mechanism
concepts
Multi-Material Drive Frame
─ Develop in-mold assembly
process for realizing transmission
mechanism
─ Concurrently optimize product
and process parameters
─ Optimize wing designs
J.W. Gerdes, K.C. Cellon, H.A. Bruck, S.K. Gupta. Characterization of the mechanics of compliant wing designs for flapping-wing miniature air vehicles. Experimental
Mechanics, Accepted 2013
W. Bejgerowski, J.W. Gerdes, S.K. Gupta, and H.A. Bruck. Design and fabrication of miniature compliant hinges for multi-material compliant mechanisms. International Journal
of Advanced Manufacturing Technology, 57(5):437-452, 2011
W. Bejgerowski, J.W. Gerdes, S.K. Gupta, H.A. Bruck, and S. Wilkerson. Design and fabrication of a multi-material compliant flapping wing drive mechanism for miniature air
vehicles. ASME Mechanism and Robotics Conference, Montreal, Canada, August 2010
 See Video at
https://www.youtube.com/watch?v=qJme
FKf0l-g
Robotic Birds: Observations

Wings undergo through significant deformation!

 Main Limitations of
Previous Designs
• Wing velocity has significant influence on wing deformation
• No way for us to control wing deformation by controlling
velocity in previous designs
─ We can only control flapping frequency

Need to control wing shape by controlling wing velocity

 New Direction in Research

• Develop robotic bird with Independently

Controllable and Programmable Wings to
understand bio-inspired flight
─ Understand the influence of wing velocity on lift and
thrust forces
─ Optimize performance
─ Aerobatic maneuvers
 Robo Raven
(Joint Project with Dr. Hugh Bruck)
 Our Inspiration: Raven

Raven Specs
Length: 24 to 26 in (61
to 66 cm)
Wingspan: 45.6 to 56.4 in
(1.2 to 1.4 m)
Weight: 2.3 lbs (1.3 kg)

Flapping 4-6 Hz
frequency:
http://www.birdsource.org/gbbc/gallery/2007/comrav_tu
cker-cr_nm.jpg/image_preview
 Research Challenges

• Independent wing control means two independent actuators

→ heavier platform
• Wing and motor must be properly matched to enable flight
─ Optimal wing design
─ Run motors at optimal operating point
• Difficult problem to model at system level

Flapping Motor Compliant Unsteady

Profile Dynamics Wings Aerodynamics
 Wing Design

• Mylar foil and carbon fiber stiffeners

• Passive deformation in response to loading
• Many iterations to get correct deformation, used high
speed imaging and load cell data to evaluate
 Approach
Additive manufacturing 6DOF Dynamic 3D DIC Strain Programmable Wings
multi-material sheets Load Measurement

Cut Mylar Incorporate

Solar Cell

Incorporate Completed
Carbon Fiber Wing Time (s)

New additive manufacturing Lift and thrust models Analytical multifunctional flight
process for incorporating for compliant flapping performance models based
flexible solar cells in thin wings by performing on energy and payload
complaint structures experimental characterization

Parameterized multifunctional System Level Optimization for

wing model using DIC/CFD/FEA Autonomous Flight

Power Solar Power Predicted Increase Measured Increase

Consumption Generation in Flight Time in Flight Time
W W Time (sec) % Time (sec) %
Regular UAV 35.8 4.10 n/a n/a n/a n/a
22 Module UAV 37.0 7.41 41.9 15.4 38.1 14.1
 Wing Sizing

Robo Raven wing area and wingspan (red)

compared to 33 other birds (blue)
Selected Wing

Para- Value
meter
S 23.85”
C 14.25”
t1 0.125”
t2 0.064”
t3 0.064”
t4 0.064”
θ1 20.5°
θ2 43°

Flapping Range: 70 degrees

Flapping Frequency: 4 Hz
 Result: Robo Raven Flying
Prototype
• Fabricated using additive manufacturing
• Independently controlled wings capable of arbitrary gaits
• New maneuvers possible: flips, dives, gliding
• Vehicle weight = 264.5g
• Flight speed = 6.7 m/s
• Endurance = 4 minutes 45 seconds

31
 Comparison with Ravens

Raven Specs Robo Raven Specs

Length: 24 to 26 in (61 Length: 24 in (61 cm)
to 66 cm)
Wingspan: 45.6 to 56.4 in Wingspan: 44 in (1.1 m)
(1.2 to 1.4 m)
Weight: 2.3 lbs (1.3 kg) Weight (w/battery, (291.6 g, 264.5 g)
w/out):

Flapping 4-6 Hz Flapping 4 Hz

frequency: frequency:
 See Video at
https://www.youtube.com/watch?v=mjOW
pwbnmTw
Flight Power Comparison

J. J. Videler, Avian Flight. Oxford: Oxford University Press 2005.

Flight cost of 33 species of birds

 Robo Raven II
(Work done in Collaboration with ARL)

• Goal: Increase system-level

performance through subsystem
modeling and optimization
• Approach
─ Dynamometer motor testing
─ Expand to a family of wing sizes
─ Select wing size and flapping gait to
improve performance by matching to
the motor
─ Increased payload used to lift
batteries or larger payload
─ Characteristics
o Max weight = 350g Robo Raven Robo Raven II
o Endurance = ~20 minutes
 See Video at
https://www.youtube.com/watch?v=q6ga9
hxm6FY
Robo Raven III
 Multifunctional Wings

• Limitations in flight time are

caused by small batteries due to
payload limitations
• Multifunctional wings with
integrated solar cells that harvest
energy
─ Assist the battery during flight
─ Recharge the battery using solar
power
 Development of Robo Raven III

• A new wing manufacturing

technique was developed to
fabricate the solar cell wings
• Solar panels:
─ Each wing contains a panel of
6 MPT6-75 solar cell modules
from Powerfilm’s©. These
flexible 7.3 x 11.4 cm solar
cells run at 6 V and produce 50
mA of current
─ Each wing is capable of
producing 1.8 Watts for a total
of 3.6 Watts
• A layered manufacturing
process is used to integrate
the panels into the wing
 Power Gains

• The current battery is a two cell Lithium Polymer (LiPo)

battery rated at 25 C.
• Robo Raven depletes the battery in 4.75 minutes using
an average of 34.6 Watts
• The solar cells produce 3.6 Watts
• The solar cells can only be used to extend the flight time
by assisting the battery or recharge the batteries
between flights
• The new maximum flight time recorded with the solar
celled wings was 5.48 minutes, a 15.3% increase
 Performance Characterization

• The integration of solar cells is expected to have an effect

on vehicle performance
• The increase in mass and stiffness of the wings will alter
how the wing deform thus alter the performance
• Load cell testing was conducted to identify the differences
in lift and thrust production by each wing design
• Digital Image Correlation was also used to determine the
differences in wing deformation between the two wing
designs
 Load Cell Testing
• Using a 6 DoF force transducer, mounted on a test
stand, the aerodynamic lift and residual thrust forces
were measured while flapping.
• The UAV was angled at a 20 deg. incline, inside of a
wind tunnel operating at 5 m/s
• The 12 cell Wings produced 5 more grams of residual
thrust and 11 more grams of aerodynamic lift

Wing Residual Lift

Thrust (grams)
(grams)
Regular 113 240
12 Cell 118 251

Regular Wings Solar Cell Wings

 Digital Image Correlation

• The load cell results quantify the

difference in force production of
each wing; however, they do not
tell us how the changes in the
wing generate these different
forces
• Digital Image Correlation (DIC)
uses two cameras to record the
speckled surface of the wing
during flapping
• Using DIC software each frame
is used to determine the
deformation and strains on the
surface of the wing
 DIC Results
• DIC directly calculates the εxx, εyy,
and εxy strains along the surface
• The biaxial strains εxy correlated
well to the lift generated by the
vehicle and the calculated shear
strain correlated best to the thrust
force generated by the vehicle Regular
Wing

Regular Solar Cell Solar Cell

Wing Wing Wing
 DIC Results (Cont.)

• To see just how well these data correlate to each

other, they were plotted in relationship to wing position

Regular Regular
Wing Wing

Solar Cell Solar Cell

Wing Wing
 DIC Results (Cont.)

• Direct comparison of the

wings
• Left: Aerodynamic Forces
• Right: Strains
• Solar cell wings generate
higher forces and higher
strains
Flight Testing Results

Regular Wings Solar Cell Wings

Forwar Climb Forwar Climb

d Rate d Rate
Velocity (m/s) Velocity (m/s)
(m/s) (m/s)
5.75 0.6 5.75 0.32
 See Videos at
https://www.youtube.com/watch?v=t1_mPe8Y0V4
https://www.youtube.com/watch?v=a8x8P5F3qTI
 Summary of Results

• The solar cells extend the flight time of the vehicle by

15.3% and charge the battery in 92 minutes
• DIC results show a correlation between residual thrust and
shear strain and aerodynamic lift and biaxial strain
• A total of 20.2 grams were added to the vehicle due to
solar cell integration but 11 grams were recovered by the
change in wing design
 New Design for Robo Raven III

• More solar cell coverage of

the wings
• Multifunctional model of
integrating solar cells to the
wings
• More solar cell coverage
throughout the body and tail
• Using more efficient solar
cells
• Energy Harvested: 8.4W
Robo Raven IV
 Robo Raven IV Goal

• Basic loitering control

─ Stabilization
─ Navigation
─ Data logging
• Stay within 40 gram payload
 Robo Raven IV: Hardware
Overview
• ArduPilot Mega 2.5
Microcontroller
─ Accelerometer, gyroscope,
compass
─ GPS/Xbee plugins
─ 16 MB DataFlash storage
─ 16 I/O pins
─ 28 grams
• Turnigy 7.4 V 370 mAh LiPo
Battery
• Spektrum receiver
• Spektrum remote
Loitering: Topographical View
 Loitering: Algorithm

• Navigation
─ GPS for localization
─ Compass for heading
─ Compare with desired position/heading
─ Set tail value to stabilize about to initiate
turns to go to center point
─ Initiates when out of set range, stops
when center point is reached (< 3 m)
• Stabilization
─ Tail as actuator controls roll/yaw
─ PID control
Loitering: Algorithm
Loitering: Error Characterization

• GPS 16

─ Rectangle 14

walk 12

─ Average error
North (m)
10
~1m
─ Blue is GPS, 8

red is path 6

traveled
4

0
-10 0 10 20 30 40 50 60 70 80

-2

East (m)
 Simulation Results

• Matlab simulation with

and without
characterized error
─ GPS: add random value
between 𝜎 and −𝜎 4 times
per second
─ Tail: Wait 150 ms once
the turn is commanded
before starting the turn
• Does not take wind into
account
 Flight Results: Log

• GPS log from Data

Flash memory
─ Red dot is the center
point
─ Blue is the bird
position
─ Errors due to
wind/GPS/tail error
 See Video at
https://www.youtube.com/watch?v=nZ0sOFI5suw
 Autonomous Dive

• Goal
─ Characterize and model dive maneuver of a FWAV to enable
autonomous diving for chemical or visual inspection of an area without
disturbing the environment
• Approach
─ Use Robo Raven IV platform (sensor suite, camera, autonomous
stabilization and navigation)
─ Video FWAV dives at varying dihedral angles
─ Develop physics model by applying fixed wing gliding theory to the
FWAV dive
─ Optimize lift and drag coefficients in model
 Dive Characterization

Example of video dive data. Wings held at a 45 degree

dihedral. Frames are 1/15 of a second apart.
Dive Characterization
 Future Directions

• Complex autonomous behaviors

• Improved wing designs
• Mechanisms for autonomous launching, landing, and
perching
• Soaring
• Improved sensing
 Summary

• Advances in robotics continue to improve manufacturing

• Realizing bio-inspired robots is challenging
o Often requires new manufacturing approaches
o We have developed unique capabilities to combine design, modeling
& simulation, and manufacturing to realize novel robot concepts
 Acknowledgements

Students and Postdocs Sponsors

• Arvind Ananthanarayanan
• AFOSR, ARO, ARL, and NSF
• Eli Barnett
• Wojciech Bejgerowski
• Leicester Ehrlich
• John Gerdes
• Adrian Greisinger
• Alex Holness
• James Hopkins
• Krishna Kaipa
• Johannes Kempny
• Gregory Krummel
• Josh Langsfeld
• Carlos Morato
• Dominik Mueller
• Savannah Nolen
• Ariel Perez-Rosado
• Luke Roberts
• Andrew Vogel
Questions?
 References

1. Holness, H.A. Bruck, and S.K. Gupta. Design of propeller-assisted flapping wing air vehicles for enhanced aerodynamic
performance. ASME Mechanism and Robotics Conference, Boston, MA, August 2015.
2. J.W. Gerdes, H.A. Bruck, and S.K. Gupta. A systematic exploration of wing size on flapping wing air vehicle performance. ASME
Mechanism and Robotics Conference, Boston, MA, August 2015.
3. A. Perez-Rosado, H.A. Bruck, and S.K. Gupta. Enhancing the Design of Solar-powered Flapping Wing Air Vehicles using
Multifunctional Structural Components. ASME Mechanism and Robotics Conference, Boston, MA, August 2015.
4. J.W. Gerdes, A. Holness, A. Perez-Rosado, L. Roberts, A. Greisinger, E. Barnett, J. Kempny, D. Lingam, C.H. Yeh, H.A. Bruck, and
S.K. Gupta. Robo Raven: A flapping wing air vehicle with highly compliant and independently controlled wings. Soft Robotics,
1(4):275--288, 2014.
5. L. Roberts, H.A. Bruck, S.K. Gupta. Autonomous loitering control for a flapping wing aerial vehicle with independent wing control.
ASME Mechanism and Robotics Conference, Buffalo, NY, August 2014.
6. Perez-Rosado, A.G.J. Griesinger, H.A. Bruck, and S.K. Gupta. Performance characterization of multifunctional wings with integrated
solar cells for miniature air vehicles. ASME Mechanism and Robotics Conference, Buffalo, NY, August 2014.
7. J.W. Gerdes, L. Roberts, E. Barnett. J. Kempny, A. Perez-Rosado H.A. Bruck, and S.K. Gupta. Wing performance characterization for
flapping wing air vehicles. ASME Mechanism and Robotics Conference, Portland, OR, August 2013.
8. J.W. Gerdes, K.C. Cellon, H.A. Bruck, S.K. Gupta. Characterization of the mechanics of compliant wing designs for flapping-wing
miniature air vehicles. Experimental Mechanics, 53: 1561-1571, 2013.
9. J.W. Gerdes, S.K. Gupta, and S. Wilkerson. A review of bird-inspired flapping wing miniature air vehicle designs. ASME Journal of
Mechanism and Robotics, 4(2), 021003.1-021003.11, 2012.
10. W. Bejgerowski, J.W. Gerdes, S.K. Gupta, H.A. Bruck, and S. Wilkerson. Design and fabrication of a multi-material compliant flapping
wing drive mechanism for miniature air vehicles. ASME Mechanism and Robotics Conference, Montreal, Canada, August 2010.
11. D. Mueller, H.A. Bruck, and S.K. Gupta. Measurement of thrust and lift forces associated with drag of compliant flapping wing for
micro air vehicles using a new test stand design. Experimental Mechanics, 50(6):725–735, 2010.
12. W. Bejgerowski, A. Ananthanarayanan, D. Mueller, and S.K. Gupta. Integrated product and process design for a flapping wing drive-
mechanism. ASME Journal of Mechanical Design, 131: 061006, 2009.
13. D. Mueller, J. Gerdes, and S.K. Gupta. Incorporation of passive wing folding in flapping wing miniature air vehicles. ASME Mechanism
and Robotics Conference, San Diego, August 30-September 2, 2009.