Experimental Cooperative Control ofFixed-Wing Unmanned Aerial Vehicles
Selcuk Bayraktar, Georgios E. Fainekos, and George J. Pappas
GRASP Laboratory
Departments of ESE and CIS
University of Pennsylvania
UAV Initiative at Penn
• OutlineDesign philosophyTest-bed
Piper J3 Cub ModelSystem architecture (Cloud cap technologies)
Hybrid Modeling (of the Piccolo autopilot)Experiments
Autonomous Flight FormationSurveillanceUAV-UGV cooperation
Future directionsTemporal logic motion planning
UAVs @ Penn
Piper Cub J3 (Scale ¼)
2.7m
Max speed: ~30m/secNominal speed: ~15m/sec
Max alt: 5000ftAutonomy: 15 - 20min
UAVs @ Penn
Servos controlling the payload
LaptopPCDell X200
3.5HP fuel engine
Deployable pod
AntennaPiccolo avionics
Pod with IMU 3DM-G and
Camera DragonFly
GPS receiver
UAVs @ Penn
Hi Res Camera & IMU POD Controllable
Deployable POD(Sensor , Beacon,
Landmark, micro-UAV)
Avionics Box EnclosureCloudCap Piccolo
Ground Station
Avionics
System architecture
OperatorPC
RS 232
Remote PCTCP / IP
Airframe
UAV 1
Avionics
Airframe
Onboard PC
RS 232 CAN
Onboard PC
RS 232 CAN
UAV nGround Control
Possible Configurations
Typical flight plan (mission)
TargetsWaypoints
Track line segment
Orbit
Take pictures
Danger
Turn rate controller
“Service”Target
Hybrid UAV Control Loop* Piccolo autopilot *
Sensing Models
GPS, IMUPressure Sensors
Continuous DynamicsAirframe Dynamics
+ Low Level PID
Controllers
Outputs Inputs
Mode Switching
Hybrid States+
Outer Control Loops
Hybrid model of thePiccolo autopilot
Turn Rate track (state 5)
Ÿturn_cmd ⁄ Ÿwpi_cmd⁄ Ÿvel_cmdG
51 R51
G12 R12 G13 R13
G11 R11
G52
R52
G14
R1
4
G31 R31
G 32
R 32
Circle track (state 3)
Ÿwpi_cmd ⁄ Ÿturn_cmd⁄ Ÿvel_cmd
Line track (state 1)
xtrack > 0 ⁄ Ÿwpi_cmd ⁄Ÿturn_cmd ⁄ Ÿvel_cmd
G 15
R 15
G54 R54
G 33
R 33
G 34R 34
G53
R53
( )cmdcmd u
UXFX
_
1111 ,
ωω ==&
G55 R55
G16 R16
G35 R35
( )( ) ( )( )
( )θω)r)rr
)&
,,,
,
1
1
1111
aptcmd xxxg
tXGtXUXFX
=
=
=
( )( ) ( )( )
( )( )θω
)r)rr
rr)r
)&
,,,
,
,
1
2
1
1111
aptcmd
wpat
xxxg
xxgx
tXGtXUXFX
=
=
=
=
G21 R21
Altitude from WP(state 2)
ŸG21 ⁄ Ÿalt_cmd
Altitude from Cmd.(state 4)
Ÿalt_cmd ⁄Ÿaltfromwp_cmd
G 22
R 22 G
42 R
42
G41 R41
( )2222 ,UXFX =&
( )2222 ,UXFX =&
Autonomous Formation Flight (AFF)Simple leader-follower formationProof of Concept
Continuous visual tracking of multiple objects On the ground or in the airCoverage transition from one UAV to another
Aerial, mobile sensor network Collective but complementary visual coverage of an areaProvide mobile, communication coverageAerial 2D or 3D mosaics
Air-ground integration (ARO MURI)Search in the air, rescue on the ground
UAV-UGV cooperation
Experiments
-400 -300 -200 -100 0 100 200 300 4000
100
200
300
400
x (meters)
y (m
eter
s)
LeaderFollower
Starting Point
Autonomous Formation Flight (AFF)
-400-300
-200-100
0100
200300
400
0
200
400
0
100
200
300
x (meters)y (meters)
z (m
eter
s)LeaderFollower
Starting Point
Fort
Ben
ning
Aug
. 200
3
S. Bayraktar, G. E. Fainekos, and G.J. Pappas, ”Hybrid Modeling and Experimental Cooperative Control of Multiple Unmanned Aerial Vehicles”, Technical Report, Department of CIS, University of Pennsylvania, 2004.
Autonomous Formation Flight (AFF)
Eye in the sky (150m)
Eye in the sky (65m)
Reconnaissance Mission
Aerial mosaic†
Fort
Ben
ning
Aug
. 200
3
†with Rahul Swaminathan
UAV-UGV cooperation
•B. Grocholsky, S. Bayraktar, V. Kumar, C. J. Taylor, and G. J. Pappas, ”Synergies in Feature Localization by Air-Ground Robot Teams”, Proceedings of the 9th International Symposium on Experimental Robotics 2004, Singapore, June 2004.•B. Grocholsky, S. Bayraktar, V. Kumar and G. Pappas, ”UAV and UGV Collaboration for Active Ground Feature Search and Localization”, AIAA 3rd ”Unmanned Unlimited” Technical Conference, Workshop and Exhibit, Chicago, Illinois, Sep. 20-23, 2004.
Fort Benning, GeorgiaNov. 2004
Synchronous formationsFixed relative distance or orientationRobust formations critical as UAVs get smaller
Cooperative control with sensor constraints Asynchronous cooperation in dynamic environments
Include sequencing and concurrency constraints Formal languages for mission/task descriptionsAbstractions for controller, sensor primitives
Decompose missions into controller primitivesDynamic verifiable composition of components
Some Technical Challenges* Future Work *[H. Tanner et. al. 42nd CDC 2003]
Building 1 Building 2
Start
UAV go to Building 1 and then to Building 2 and then return to Start and UGV cover Ground Area.
Ground area
∃ (◊UAV.Building1 ∧ (◊UAV.Building2 ∧(◊UAV.Starting Area))) ∧∃ (UGV.cover(GroundArea) while UGV.avoid(Building1 ∧ Building1))
Georgios E. Fainekos, Hadas Kress-Gazit and George J. Pappas, Temporal logic motion planning for mobile robots, Submitted to ICRA'05
1 2
34
5
6 7
8
9
10 11
1213
14 15
16 17
1819
20
1
2
3 4 5
6
7
8
9
10
11
12
13
14
15
161718
1920
2122
23
24
2526
2728
29
30
3132
33
34
Temporal logic motion planning
[C. Belta and L. Habets 43rd CDC 2004]
Temporal logic motion planning
Georgios E. Fainekos, Hadas Kress-Gazit and George J. Pappas, Temporal logic motion planning for mobile robots, Submitted to ICRA'05
1 2
34
5
6 7
8
9
10 11
1213
14 15
16 17
1819
20
1
2
3 4 5
6
7
8
9
10
11
12
13
14
15
161718
1920
2122
23
24
2526
2728
29
30
3132
33
34
Temporal logic motion planning
Georgios E. Fainekos, Hadas Kress-Gazit and George J. Pappas, Temporal logic motion planning for mobile robots, Submitted to ICRA'05
Conclusions
Development of a fleet of autonomous UAVs
Various experiments that demonstrate the operational capabilities of the test-bed
Current work in the development of algorithms for the composition of control primitives that satisfy user specifications
The ENDThank you!Questions?
