System Identification of Rotorcraft Rebecca Creed Mechanical Engineering

System Identification of Rotorcraft Rebecca Creed, Mechanical Engineering, University of Dayton Andrea Gillis, Aerospace Engineering, University of Cincinnati Urvish Patel, EE-Comp. E Accend, University of Cincinnati Dr. Kelly Cohen, Faculty Mentor, University of Cincinnati Mr. Wei, Graduate Mentor, University of Cincinnati July 11, 2013 Part of NSF Type 1 STEP Grant, Grant ID No. : DUE-0756921 1

Introduction • Natural disasters take thousands of lives every year. • Many first responders perform dangerous rescue missions to save lives. • Technology will allow first responders to assess the situation more quickly and efficiently. 2

2013 Arizona and 2012 Colorado Wildfire • The progression of the fire could not be anticipated due to severe weather conditions. • Accurate situational awareness and fire growth predictive capability can be obtained using a UAV and intelligent software. • An autopilot stabilized UAV (rotorcraft) would be able to collect information using a camera Image courtesy of csmonitor. com 3

UAV Advantages • • • Maneuverability Capable of indoor flight Safer for Crews Endurance Cost Sushi Delivery Image courtesy of http: //www. todaysiphone. com/2013/06/yo-sushidelivering-food-on-ipad-controlled-trays/ 4

Why Autopilot? • Easy to use with simple controls • Increase the range of the rotorcraft – Without autopilot, the rotorcraft must remain in the operator’s line of sight • A dynamic model is necessary to develop an autopilot 5

System Identification • A dynamic model is a representation of the behavior of a system (for this case, rotorcraft) • Two options for creating a dynamic model – System Identification – Wind Tunnel Testing • Placing the rotors in a wind tunnel is complex – Simulation • Based on the moment of inertia 6

Research Goals – Study the characteristics of Aeroquad. – Use System Identification and CIFER software to develop and validate the dynamic model of Aeroquad. – Report the findings so process can be repeated in the future. 7

So, what is System Identification? Inputs System Outputs Given the inputs to a system, a system model can predict the outputs 8

Simple Example: Pushing a Sled Input is the “pushing” force applied to the sled • Push(force) Output is the sled’s movement Sled • • • Acceleration Velocity Displacement 9

System Inputs and Outputs • 4 inputs – – Yaw Pitch Roll Thrust • 9 outputs – 3 attitudes – 3 angular rates – 3 accelerations Aeroquad System 10

System Identification Flowchart Flight Testing Data Processing MATLAB Data Evaluation CIFER System Model Validation System Identified! 11

Flight Test • Inputs given to the rotorcraft by RC Controller from Futaba. • Outputs recorded by the 9 DOF Sensor stick from Open Hardware. 12

How the quad-rotor works Roll Control move right/left Yaw Control spin cw/counter-cw Pitch Control move forward/backward 13

Data Processing Flight Testing Data Processing Data Evaluation System Model Validation Record raw data in MATLAB program Filter recorded data Reformat data for use in CIFER System Identified! 14

Filter Data Sensor stick used in Rotorcraft – 9 DOF Accelerometer ADXL 345 Next Step Filtered Data Noisy Data Filter Picture from: www. sparkfun. com 15

Result from Kalman Filter 5 X axis 4 3 2 1 0 Regular -1 Kalman -2 -3 -4 -5 -6 We designed new and unique Kalman Filter ! 16

Moving average and Kalman 5 X axis 4 Regular 3 2 Kalman 1 Regular 0 Kalman -1 Moving average -2 -3 Moving Average -4 -5 -6 17

Data Evaluation CIFER • Advanced program used for System Identification • Stands for Comprehensive Identification from Frequency Responses • Developed by the U. S. Army and the University of California Santa Cruz • We use CIFER to identify the Aeroquad system CIFER image from: http: //uarc. ucsc. edu/flight-control/cifer/ 18

Data Evaluation First Step – find the frequency response • Frequency response relates the inputs and outputs of our data 100 80 60 40 20 0 -20 -40 -60 -80 -100 0 10 20 30 40 50 Time (seconds) Output Time History Data 60 Pitch Output (degree/second) Pitch Input (Percentage) Input Time History Data 100 80 60 40 20 0 -20 -40 -60 -80 -100 0 10 20 30 40 50 60 Time (seconds) 19

Coherence Phase (Deg) Magnitude (DB) Data Evaluation – Frequency Response Frequency (Hz) A coherence value closest to 1 shows that the inputs and outputs correlate well. 20

Data Evaluation Next step: transfer function fit • CIFER fits a transfer function to the frequency response What is a Transfer Function? The transfer function relates the inputs to the outputs of a system CIFER finds the coefficients of this transfer function 21

Data Evaluation – Stability • CIFER produces transfer functions for three motions • These transfer functions model the Aeroquad system and must be stable(by doing a closed loop system identification, the system is stable) Roots of the denominator should be on this side! Stable Example Negative real roots Unstable Example Positive real roots 22

Results • Transfer Functions of Aeroquad – Roll: roots: -5. 385, -58. 934 – Pitch: roots: -4. 715, -208. 084 – Yaw: roots: -6. 199 *Notice all roots are negative representing a stable system !! 23

Validation • CIFER also finds the state-space representations from the pitch, roll, and yaw frequency responses – Separate sets of data for pitch, roll, and yaw were used – Angular rates of the output were predicted from the input angular rates Inputs (percentages) State-space Model Outputs (angular rates) 24

Pitch Output (radians/second) Validation – Pitch Time (seconds) 25

Roll Output (radians/second) Validation – Roll Time (seconds) 26

Yaw Output (radians/second) Validation – Yaw Time (seconds) 27

Summary • Observed the connection between our research and potential societal impact… saving lives during fires/natural disasters • Learned to fly quad-rotor using RC controls • Data collected from flight tests utilized to develop dynamic model using system identification • A new and unique Kalman filter was developed and its effectiveness was demonstrated. • The model was validated with additional flight test data! 28

Timeline 29

Publications • Conference: 9 th Annual Dayton Engineering Sciences Symposium (DESS 2013), October 29, 2013, Wright State University, Dayton, Ohio • Journal: Computer Application in Engineering Education, Wiley Periodicals, Inc. 30

References • Bestaoui, Y. , and Slim, R. (2007). “Maneuvers for a Quad-Rotor Autonomous Helicopter, ” AIAA Infotech@Aerospace Conference, held at Rohnert Park, California, May 7 -10, pp. 1 -18 • Chen, M. , and Huzmezan, M. (2003). “A Combined MBPC/2 DOF H∞ Controller for a Quad Rotor UAV, ” AIAA Guidance, Navigation, and Control Conference and Exhibit, held at Austin, Texas, August 11 -14, n. p. • Esme, B. (2009). “Kalman Filter For Dummies. ” Biligin’s Blog, <http: //bilgin. esme. org/Bits. Bytes/Kalman. Filterfor. Dummies. aspx> (Mar. 2009). • Guo, W. , and Horn, J. (2006). “Modeling and Simulation For the Development of a Quad-Rotor UAV Capable of Indoor Flight , ” AIAA Modeling and Simulation Technologies Conference, held at Keystone, Colorado, August 21 -24, pp. 1 -11 • Halaas, D. , Bieniawski, S. , Pigg, P. , and Vian, J. (2009). “Control and Management of an Indoor Health Enabled, Heterogenous Fleet, ” AIAA Infotech@Aerospace Conference, held at Seattle, Washington, April 6 -9, pp. 1 -19 31

• References Koehl, A. , Rafaralahy, H. , Martinez, B. , and Boutayeb, M. (2010). “Modeling and Identification of a Launched Micro Air Vehicle: Design and Experimental Results, ” AIAA Modeling and Simulation Technologies Conference, held at Toronto, Ontario Canada, August 2 -5, pp. 1 -18 • Mehra, R. , Prasanth, R. , Bennett, R. , Neckels, D. , and Wasikowski, M. (2001). “Model Predictive Control Design for XV-15 Tilt Rotor Flight Control, ” AIAA Guidance, Navigation, and Control Conference and Exhibit, held at Montreal, Canada, August 6 -9, pp. 1 -11. • Milhim, A. , and Zhang, Y. (2010). “Quad-Rotor UAV: High-Fidelity Modeling and Nonlinear PID Control, ” AIAA Modeling and Simulation Technologies Conference, held at Toronto, Ontario, Canada, August 2 -5, pp. 1 -10. • Salih, A. , Moghavvemi, M. , Mohamed, H. , and Gaeid, K. (2010). “Flight PID controller design for a UAV quadrotor, ” Scientific Research and Essays, ? ? , Vol. 5, No. 23, pp. 3660 -3667. • Tischler, M. B. , and Cauffman, M. G. (2013). “Frequency-Response Method for Rotorcraft System Identification: Flight Applications to BO 105 Coupled Fuselage/Rotor Dynamics, ” University Affiliated Research Center: A Partnership Between UCSC and NASA Ames Research Center, pp. 1 -13. 32

Questions? 33