43
Project #4: Simulation and Experimental Testing of Allocation of UAVs Tim Arnett, Aerospace Engineering, Junior, University of Cincinnati Devon Riddle, Aerospace Engineering, Junior University of Cincinnati ASSISTED BY: Chelsea Sabo, Graduate Research Assistant Dr. Kelly Cohen, Faculty Mentor

Project #4: Simulation and Experimental Testing of Allocation of UAVs Tim Arnett, Aerospace Engineering, Junior, University of Cincinnati Devon Riddle,

Embed Size (px)

Citation preview

Page 1: Project #4: Simulation and Experimental Testing of Allocation of UAVs Tim Arnett, Aerospace Engineering, Junior, University of Cincinnati Devon Riddle,

Project #4: Simulation and Experimental Testing of

Allocation of UAVs

Tim Arnett, Aerospace Engineering, Junior, University of Cincinnati

Devon Riddle, Aerospace Engineering, Junior University of Cincinnati

ASSISTED BY:

Chelsea Sabo, Graduate Research Assistant

Dr. Kelly Cohen, Faculty Mentor

Page 2: Project #4: Simulation and Experimental Testing of Allocation of UAVs Tim Arnett, Aerospace Engineering, Junior, University of Cincinnati Devon Riddle,

Outline• Applications of UAVs• Challenges• Project Goals and Objectives• Vehicle Routing Problems• Experimental Testing

– Experimental Setup– Waypoint Navigation Algorithm

• AMASE– Why use AMASE?– Overview– Features

• Results & Analysis• Acknowledgements• Questions

2

Page 3: Project #4: Simulation and Experimental Testing of Allocation of UAVs Tim Arnett, Aerospace Engineering, Junior, University of Cincinnati Devon Riddle,

Why UAVs?3

• Missions that are “dull, dirty, and dangerous”• Cost and performance

– Do not need pilot life support systems– Removal of human survivability constraints

allows better performance

Page 4: Project #4: Simulation and Experimental Testing of Allocation of UAVs Tim Arnett, Aerospace Engineering, Junior, University of Cincinnati Devon Riddle,

Applications of Surveillance Missions with UAVs

• Search and Rescue• Weather Observation• Forest Fire Monitoring

• Traffic Surveillance• Border Patrol• Military

4

Page 5: Project #4: Simulation and Experimental Testing of Allocation of UAVs Tim Arnett, Aerospace Engineering, Junior, University of Cincinnati Devon Riddle,

Challenges

• Obtaining software and equipment suitable for tests– Systems difficult to obtain and usually

expensive• Verifying solutions on proven systems

– Systems not always well-documented or fully supported

5

Page 6: Project #4: Simulation and Experimental Testing of Allocation of UAVs Tim Arnett, Aerospace Engineering, Junior, University of Cincinnati Devon Riddle,

Project Goals

• Learn to interface equipment for UAV controller development

• Compare two routing solutions for common performance metrics

6

Page 7: Project #4: Simulation and Experimental Testing of Allocation of UAVs Tim Arnett, Aerospace Engineering, Junior, University of Cincinnati Devon Riddle,

Objectives• Objective 1: Interface with cooperative control

development systems– Interface and run algorithms on AR Drones– Interface and run algorithms on AMASE

• Objective 2: Validate task allocation algorithm both in simulation and experimentally

• Objective 3: Test and compare cooperative control strategies for UAVs– Distance travelled– Delivery time for time critical targets

7

Page 8: Project #4: Simulation and Experimental Testing of Allocation of UAVs Tim Arnett, Aerospace Engineering, Junior, University of Cincinnati Devon Riddle,

Vehicle Routing Problems8

DepotTargets

Targets

Targets

• Multiple routing solutions exist depending on desired operational goals

• Which UAV services a target and in what order are the targets visited?

Page 9: Project #4: Simulation and Experimental Testing of Allocation of UAVs Tim Arnett, Aerospace Engineering, Junior, University of Cincinnati Devon Riddle,

Vehicle Routing Problems:Minimum Distance Route

9

Minimum distance solution is useful for minimizing total mission time, fuel consumption, etc.

Page 10: Project #4: Simulation and Experimental Testing of Allocation of UAVs Tim Arnett, Aerospace Engineering, Junior, University of Cincinnati Devon Riddle,

Vehicle Routing Problems:Minimum Delivery Latency Route

10

Often desirable to deliver data to a high-bandwidth connection or “depot”

For this case, the delivery time is often of interest due to missions being time critical

Page 11: Project #4: Simulation and Experimental Testing of Allocation of UAVs Tim Arnett, Aerospace Engineering, Junior, University of Cincinnati Devon Riddle,

Test Cases• 3 different tests performed

– Differing difficulty and number of targets– Both Minimum Distance and Minimum Delivery

Latency solutions implemented for each test• Tests done both experimentally and in simulation

– Experiments done in IMAGE Lab with AR Drone UAVs– Simulations created in AMASE – an Air Force flight

simulation environment• Compared distance travelled and delivery time for

each test

11

Page 12: Project #4: Simulation and Experimental Testing of Allocation of UAVs Tim Arnett, Aerospace Engineering, Junior, University of Cincinnati Devon Riddle,

Experimental Setup12

AR Drones

IMAGE Lab

Page 13: Project #4: Simulation and Experimental Testing of Allocation of UAVs Tim Arnett, Aerospace Engineering, Junior, University of Cincinnati Devon Riddle,

Experimental Setup

• AR Drone– Inexpensive, commercially available quadrotor– “Black box” with limited support– Can be controlled by a device using wireless

network adapter

13

Page 14: Project #4: Simulation and Experimental Testing of Allocation of UAVs Tim Arnett, Aerospace Engineering, Junior, University of Cincinnati Devon Riddle,

Experimental Setup14

AR Drones

OptiTrack Cameras

IMAGE Lab

Page 15: Project #4: Simulation and Experimental Testing of Allocation of UAVs Tim Arnett, Aerospace Engineering, Junior, University of Cincinnati Devon Riddle,

Experimental Setup

• Optitrack System– Cameras provide real time position data– Data can be imported into MatLab

15

Page 16: Project #4: Simulation and Experimental Testing of Allocation of UAVs Tim Arnett, Aerospace Engineering, Junior, University of Cincinnati Devon Riddle,

Experimental Setup16

AR Drones

OptiTrack Cameras

Wireless Router

PC with MatLab and OptiTrack Tracking Tools

IMAGE Lab

Page 17: Project #4: Simulation and Experimental Testing of Allocation of UAVs Tim Arnett, Aerospace Engineering, Junior, University of Cincinnati Devon Riddle,

Experimental Setup

• Software Interface– PC client with wireless capability, MatLab, and

camera software– Wireless router to connect to multiple drones

17

Page 18: Project #4: Simulation and Experimental Testing of Allocation of UAVs Tim Arnett, Aerospace Engineering, Junior, University of Cincinnati Devon Riddle,

Waypoint Algorithm

• Needed to dictate flight path of UAV• Control Methods

– Proportional-Derivative Control– Fuzzy Logic Control

18

Page 19: Project #4: Simulation and Experimental Testing of Allocation of UAVs Tim Arnett, Aerospace Engineering, Junior, University of Cincinnati Devon Riddle,

Control Diagram19

Page 20: Project #4: Simulation and Experimental Testing of Allocation of UAVs Tim Arnett, Aerospace Engineering, Junior, University of Cincinnati Devon Riddle,

Waypoint Navigation Controller

• Proportional-Derivative controller– Used for Yaw Rate, Ascent Rate

• Provides good response and settling time• Simple implementation

20

Page 21: Project #4: Simulation and Experimental Testing of Allocation of UAVs Tim Arnett, Aerospace Engineering, Junior, University of Cincinnati Devon Riddle,

Waypoint Navigation Controller

• Fuzzy Logic Controller– Used for Pitch, Roll

• Does not require system model• Robust to stability issues

21

Page 22: Project #4: Simulation and Experimental Testing of Allocation of UAVs Tim Arnett, Aerospace Engineering, Junior, University of Cincinnati Devon Riddle,

AMASE

Automatic Test System Modeling and System Environment

22

Page 23: Project #4: Simulation and Experimental Testing of Allocation of UAVs Tim Arnett, Aerospace Engineering, Junior, University of Cincinnati Devon Riddle,

History of AMASE

• AFRL– Air Force Research Laboratory (Wright Patterson)

• Desktop simulation environment developed for UAV cooperative control studies

• Used to develop and optimize multiple- UAV engagement approaches

• Self contained simulation environment that accelerates iterative development/analysis

23

Page 24: Project #4: Simulation and Experimental Testing of Allocation of UAVs Tim Arnett, Aerospace Engineering, Junior, University of Cincinnati Devon Riddle,

Why AMASE?• Control algorithms can be assessed and compared

effectively • Free for University research• An environment that provides a formal simulation of

the algorithm as a precursor to large scale flight tests.• Proven as a legitimate way to set up realistic flight

simulations.• Provides good visual description of what’s happening

24

Challenge: No technical support… Learned through trial and error.

Page 25: Project #4: Simulation and Experimental Testing of Allocation of UAVs Tim Arnett, Aerospace Engineering, Junior, University of Cincinnati Devon Riddle,

Important Features25

The Map XML Editing

Event EditorCreate

Scenario

Plan Request (CMASI) Validation

Run Scenario

Connect with Client

Record and Analyze data

Page 26: Project #4: Simulation and Experimental Testing of Allocation of UAVs Tim Arnett, Aerospace Engineering, Junior, University of Cincinnati Devon Riddle,

AMASE Set Up Tool: This is where all of the scenarios are created and the progress is saved.

26

The Map Event Editor

Toolbar

Error Box

Page 27: Project #4: Simulation and Experimental Testing of Allocation of UAVs Tim Arnett, Aerospace Engineering, Junior, University of Cincinnati Devon Riddle,

Simulation of test data on a world wide scale

27

What runs the simulation

Characteristics of the aircraftThe Map

Aircraft

Path line

Page 28: Project #4: Simulation and Experimental Testing of Allocation of UAVs Tim Arnett, Aerospace Engineering, Junior, University of Cincinnati Devon Riddle,

Experimental Results28

-0.5 0 0.5 10

0.5

1

1.5

2

1

2

3

4 56

7

8

1

-0.5 0 0.5 10

0.5

1

1.5

2

1

234

5

67

8

1

Minimum Delivery Latency Route Minimum Distance Route

Page 29: Project #4: Simulation and Experimental Testing of Allocation of UAVs Tim Arnett, Aerospace Engineering, Junior, University of Cincinnati Devon Riddle,

Analysis29

𝐽𝑇=∑𝑖=1

𝑁

𝐷 𝑖 𝑅𝑡𝑜𝑡𝑎𝑙=√ (𝑥1−𝑥𝑑𝑒𝑝𝑜𝑡 )2+( 𝑦1− 𝑦 𝑑𝑒𝑝𝑜𝑡 )

2+∑𝑖=2

𝑁

√ (𝑥𝑖−𝑥 𝑖−1 )2+ (𝑦 𝑖− 𝑦 𝑖− 1)2

Total Time Cost Total Distance Travelled

Minimum Delivery Latency

Minimum Distance Difference

Total Time Cost

751.96 1134.92 -33.74%

Total Distance Travelled

25.11 18.04 -28.14%

D = Delivery Time

Page 30: Project #4: Simulation and Experimental Testing of Allocation of UAVs Tim Arnett, Aerospace Engineering, Junior, University of Cincinnati Devon Riddle,

Simulation 1(a)

30

Simulation Results

Page 31: Project #4: Simulation and Experimental Testing of Allocation of UAVs Tim Arnett, Aerospace Engineering, Junior, University of Cincinnati Devon Riddle,

Simulation 1(b)

31

Simulation Results

Page 32: Project #4: Simulation and Experimental Testing of Allocation of UAVs Tim Arnett, Aerospace Engineering, Junior, University of Cincinnati Devon Riddle,

Analysis32

𝐽𝑇=∑𝑖=1

𝑁

𝐷 𝑖

Total Time Cost

D = Delivery Time

Minimum Delivery Latency

Minimum Distance Difference

Total Time Cost

31915 38807 -17.76%

Page 33: Project #4: Simulation and Experimental Testing of Allocation of UAVs Tim Arnett, Aerospace Engineering, Junior, University of Cincinnati Devon Riddle,

Comparison33

Ideal Experimental AMASE Test 1 -21.89% -44.35% -21.45% Test 2 -18.65% -16.81% -11.98% Test 3 -24.90% -32.92% -11.50%

% Improvement of Total Time Cost for the Minimum Delivery Latency route compared to the Minimum Distance route

% Improvement of Total Distance Travelled for the Minimum Distance route compared to the Minimum Delivery Latency route

Ideal Experimental Test 1 -21.91% -8.88% Test 2 -28.75% -39.93% Test 3 -29.94% -32.55%

Page 34: Project #4: Simulation and Experimental Testing of Allocation of UAVs Tim Arnett, Aerospace Engineering, Junior, University of Cincinnati Devon Riddle,

Acknowledgements34

• NSF Grant # DUE-0756921 for Type 1 Science, Technology, Engineering, and Mathematics Talent Expansion Program (STEP) Project

• Kelly Cohen, Ph.D, Faculty Mentor, University of Cincinnati, Cincinnati, OH• Chelsea Sabo, Ph.D, GRA, University of Cincinnati, Cincinnati, OH• Stephanie Lee, AFRL, Wright-Patterson Air Force Base, Dayton, OH• Manish Kumar, Ph.D, University of Toledo, Toledo, OH• Balaji Sharma, MS, University of Toledo, Toledo, OH• Ruoyu Tan, MS, University of Toledo, Toledo, OH

• Task Allocation Algorithm sourced from work done by Dr. Chelsea Sabo 

Page 35: Project #4: Simulation and Experimental Testing of Allocation of UAVs Tim Arnett, Aerospace Engineering, Junior, University of Cincinnati Devon Riddle,

Questions?35

Page 36: Project #4: Simulation and Experimental Testing of Allocation of UAVs Tim Arnett, Aerospace Engineering, Junior, University of Cincinnati Devon Riddle,

Command Value Conversion

AR Drone requires commands in text strings with values formatted as a 32-bit signed integer• Command string example

36

CMD = sprintf('AT*PCMD=%d,%d,%d,%d,%d,%d\r',i,1,0,1036831949,0,0);fprintf(ARc, CMD);

Sequence

admin
Less words - explain each piece.
Page 37: Project #4: Simulation and Experimental Testing of Allocation of UAVs Tim Arnett, Aerospace Engineering, Junior, University of Cincinnati Devon Riddle,

Command Value Conversion

AR Drone requires commands in text strings with values formatted as a 32-bit signed integer• Command string example

37

CMD = sprintf('AT*PCMD=%d,%d,%d,%d,%d,%d\r',i,1,0,1036831949,0,0);fprintf(ARc, CMD);

Flag

admin
Less words - explain each piece.
Page 38: Project #4: Simulation and Experimental Testing of Allocation of UAVs Tim Arnett, Aerospace Engineering, Junior, University of Cincinnati Devon Riddle,

Command Value Conversion

AR Drone requires commands in text strings with values formatted as a 32-bit signed integer• Command string example

38

CMD = sprintf('AT*PCMD=%d,%d,%d,%d,%d,%d\r',i,1,0,1036831949,0,0);fprintf(ARc, CMD);

Roll

admin
Less words - explain each piece.
Page 39: Project #4: Simulation and Experimental Testing of Allocation of UAVs Tim Arnett, Aerospace Engineering, Junior, University of Cincinnati Devon Riddle,

Command Value Conversion

AR Drone requires commands in text strings with values formatted as a 32-bit signed integer• Command string example

39

CMD = sprintf('AT*PCMD=%d,%d,%d,%d,%d,%d\r',i,1,0,1036831949,0,0);fprintf(ARc, CMD);

Pitch

• Value corresponds to a command value of 0.1• Values are a ratio to the full value allowable by the drone

admin
Less words - explain each piece.
Page 40: Project #4: Simulation and Experimental Testing of Allocation of UAVs Tim Arnett, Aerospace Engineering, Junior, University of Cincinnati Devon Riddle,

Command Value Conversion

AR Drone requires commands in text strings with values formatted as a 32-bit signed integer• Command string example

40

CMD = sprintf('AT*PCMD=%d,%d,%d,%d,%d,%d\r',i,1,0,1036831949,0,0);fprintf(ARc, CMD);

Ascent Rate

admin
Less words - explain each piece.
Page 41: Project #4: Simulation and Experimental Testing of Allocation of UAVs Tim Arnett, Aerospace Engineering, Junior, University of Cincinnati Devon Riddle,

Command Value Conversion

AR Drone requires commands in text strings with values formatted as a 32-bit signed integer• Command string example

41

CMD = sprintf('AT*PCMD=%d,%d,%d,%d,%d,%d\r',i,1,0,1036831949,0,0);fprintf(ARc, CMD);

Yaw Rate

admin
Less words - explain each piece.
Page 42: Project #4: Simulation and Experimental Testing of Allocation of UAVs Tim Arnett, Aerospace Engineering, Junior, University of Cincinnati Devon Riddle,

The Event Editor

• AirVehicleConfiguration– Characteristics of the UAV – Given

• AirVehicleEntity– Characteristics of where the UAV starts in a

scenario and where it will go first• MissionCommand

– Tells the UAV where to go from homebase

42

Page 43: Project #4: Simulation and Experimental Testing of Allocation of UAVs Tim Arnett, Aerospace Engineering, Junior, University of Cincinnati Devon Riddle,

CMASI

• Common Mission Automation Services Interface– A system of interactive objects that pertain to

the command and control of a UAV system.– Where the MissionCommand is used. – Example of two scenarios to show why

CMASI is important.

43