Application of DDDAS Principles to Command, Control and ...€¦ · Swarm Mission Planning and...
Transcript of Application of DDDAS Principles to Command, Control and ...€¦ · Swarm Mission Planning and...
Application of DDDAS Principles to
Command, Control and Mission Planning
for UAV Swarms
Greg Madey & R. Ryan McCune Department of Computer Science and Engineering
University of Notre Dame
DDDAS PI Meeting – Dec 1-3, 2014
IBM T. J. Watson Research Center
Project History • Goal: Apply DDDAS Concept to Command & Control of UAV Swarms
• Two year design, development, prototypes, demonstration, & evaluation
– Test bed
– Synthetic (simulated) UAV Swarms
– Demo with basic UAVs – Parrot AR.DRONE 2.0
– Proof-of-concept => DDDAS concept can be applied!
• Current: DURIP –> UAV hardware with improved payload, communication, performance, open development features
– Continuation of development, demonstration, & evaluation with more capable hardware (CrazyFli, Hummingbird, Iris, Dr. Robot, DragonFly)
– Supplemental support from GAANN and NSF
• Applying new computational and networking paradigms, e.g., emergent computing, swarm intelligence, vertex-oriented computing
1
Edwin OnattuUAV assembly,
configuration, flight
Andrew GnottHardware
Evaluation & Recommendations
Sean GleasonUAV assembly,
configuration, flight
Travis PattersonUAV assembly,
configuration, flight
R. Ryan McCuneGAANN Fellow
Research Course Instructor
Greg MadeyAgent-Based Simulation
DDDAS Applications
M. Brian BlakeSoftware Service
Discovery and Composition
Chris PoellabauerWireless Networks,Sensor Networks, Mobile Computing,Real-Time Systems
R. Ryan McCune & Rachael PurtaCSE PhD Students
Mikolaj Dobski & Artur JaworskiCSE REU Students
Alexander MadeyHigh School Research Intern
David WeiCSE PhD Student
Hongsheng LuCSE PhD Student
Unmanned Aerial Vehicles (UAVs)
• Trends – Numbers increasing
– Costs decreasing
– Small to full sized
– Capability increasing
– Swarms
Unmanned Aerial Vehicles (UAVs)
• Challenges – Operator
overload
– Training costs
– Flying the swarm
– Emergent behavior in swarm, how to control?
Operator Overload
• Unmanned Aerial Vehicles (UAVs) – No on-board pilot – Intelligence, Surveillance, and
Reconnaissance (ISR) missions – Becoming smaller and
cheaper with increased capabilities
• How to efficiently operate many UAVs?
5
DDDAS Concept
• Simulations of the Swarms for Ground-based Operators – Mission Planning
– Dynamic Mission Re-planning
– Command & Control
• Agent-based Simulations – Dynamically Updated by UAV Sensors
– What-if Predictive Modeling to Support Re-planning and Command & Control (Fly the Swarm)
Sequence
1 Research Test-bed • Virtual UAV Swarms • Simulation-to-Simulation
Modeling • Investigate DDDAS Research
Questions • Proof-of-concept
2 DURIP • Physical UAV Swarms • Evaluate, Calibrate,
Demonstrate and Validate
Ground Station
Operator TeamMission Planning & Re-Planning
Command & Control
Virtual UAV Swarms
Applica' on*of*DDDAS*Principles*to*Command,*Control*and*Mission*Planning*for*UAV*SwarmsResearch*Test?Bed
TestBed Virtual Swarms
Ground Station
Operator TeamMission Planning & Re-Planning
Command & Control
Physical UAV SwarmsDURIP Physical Swarms
Results
• A DDDAS test-bed was developed utilizing web-services middleware to communicate between a “real-world” UAV swarms and agent-based simulations. Six Parrot AR.Drones 2.0 quadrocopters were demo’ed as a “real world” UAV swarm communicating over the test-bed to a ground-based command & control application.
0
G.R. Madey, M.B. Blake, C. Poellabauer, H. Lu, R.R. McCune, Y. Wei, “Applying DDDAS Principles to Command, Control and Mission Planning for UAV Swarms”, Procedia Computer Science, Volume 9, 2012
Testbed System – Console & UI
Console DDDAS Application
System View
Virtual UAV
Swarm
UAVs in Flight
2
Results
• Another investigation examined dynamic mission scheduling for swarms by incorporating DDDAS principles into a global-local hybrid-planning scheme. A global agent utilized simulation to determine optimal task assignment, while UAVs locally determine the execution order for assigned tasks.
3
Wei, Y., Madey, G.R, and Blake, M.B. Blake. "An Operation-time Simulation Framework for UAV Swarm Configuration and Mission Planning, ICCS 2013, Barcelona, Spain. June 5-7, 2013
Mission Planning Example
T1
T2
T3
Mission Swarm
V1
V2
V3
V4
Two Missions, Multiple Tasks Each, Global Task
Assignment, and Scheduling by UAVs Locally
5
Results
• Another investigation applied the DDDAS paradigm to two swarm command and control scenarios. In both scenarios, UAV swarms were augmented with mission-specific sensors, providing real-time measurements to a ground operator. Based on real-time measurements, the operator could adjust a single, global swarm parameter to achieve mission objectives.
6
Madey, A.G., and Madey, G.R. 2013. "Design and Evaluation of UAV Swarm Command and Control Strategies, ADS'13/SpringSim2013. SCS, San Diego, CA. 2013.
DDDAS Inspired Control of UAV Swarm
7
DDDAS Control/Sensor Data Local Swarm Behavior
Two Types of UAVs
Swarm Performance Improved with One Control Input – the “Cohere Value”
8
Target UAV - Active Pursuer UAV - Inactive Pursuer UAV - Search
Results
• Another investigation quantified swarm performance with agent-based modeling. The modeling demonstrated the utility of an explanatory model in the DDDAS framework, and provided an interesting juxtaposition of a bottom-up model analyzed by a top-down clustering algorithm, where both calculated results based on the distance of agent neighbors. Example: UAV Swarm for disaster management
9
McCune, R. R. and G. R. Madey. “Decentralized K-Means Clustering with MANET Swarms”, ADS'14/SpringSim 2014, SCS, Tampa, FL. April 13-16, 2014 McCune, R. Ryan, and Greg R. Madey. "Control of Artificial Swarms with DDDAS." Procedia Computer Science 29 (2014): 1171-1181.
10
Tanker Moves
Swarm
Clustering
11
Simulation Snap-Shots
• Swarm clustering simulation
• Voronoi diagram overlay – Tankers as seed
points
• UAV Swarm for disaster support
More Results in Publications
12
McCune, R. R. and G. R. Madey. “Decentralized K-Means Clustering with MANET Swarms”, ADS'14/SpringSim 2014, SCS, Tampa, FL. April 13-16, 2014. Madey, A.G., and Madey, G.R. 2013. "Design and Evaluation of UAV Swarm Command and Control Strategies, ADS'13/SpringSim2013. SCS, San Diego, CA. 2013. McCune, R., Y. Wei, R. Purta, A. Madey, M. B. Blake, and G. Madey, “Investigations of DDAS for Command and Control of UAV Swarms with Agent-Based Modeling.” WSC 2013. Washington D.C. December 8-11, 2013. McCune, R. R., and G. Madey. "Swarm Control of UAVs for Cooperative Hunting with DDDAS”, ICCS 2013, Barcelona, Spain. June 5-7, 2013. McCune, R. R., and G. R. Madey. "Agent-Based Simulation of Cooperative Hunting with UAVs.”ADS'13/SpringSim2013, SCS, San Diego, CA. 2013. Purta, R., M. Dobski, A. Jaworski, and G. Madey. "A Testbed for Investigating the UAV Swarm Command and Control Problem Using DDDAS”, ICCS 2013, Spain. June 5-7, 2013. Purta, R., Saurabh N., and G. Madey. "Multi-hop Communications in a Swarm of UAVs.” ADS'13/SpringSim 2013, SCS, San Diego, CA. 2013. Wei, Y., Madey, G.R, and Blake, M.B. Blake. "An Operation-time Simulation Framework for UAV Swarm Configuration and Mission Planning, ICCS 2013, Barcelona, Spain. June 5-7, 2013.
Wei, Y., G. Madey, and M. B. Blake. "Agent-based Simulation for UAV Swarm Mission Planning and Execution.” ADS’13/SpringSim 2013, SCS, 2013. Y. Wei and M.B. Blake, “An Agent‐based Services Framework with Adaptive Monitoring in Cloud Environments”, WETICE 2012, IEEE Press, Toulousse, France, June 2012 -‐ Best Student Paper Award. Wei, Y., and M. B. Blake. "Adaptive Web Services Monitoring in Cloud Environments." International Journal on Web Portals (2013). G.R. Madey, M.B. Blake, C. Poellabauer, H. Lu, R.R. McCune, Y. Wei, “Applying DDDAS Principles to Command, Control and Mission Planning for UAV Swarms”, Procedia Computer Science, Volume 9, 2012 McCune, R. Ryan, and Greg R. Madey. "Control of Artificial Swarms with DDDAS." Procedia Computer Science 29 (2014): 1171-1181. Madey, A., “Unmanned Aerial Vehicle Swarms: The Design and Evaluation of Command and Control Strategies using Agent-based Modeling”, International Journal of Agent Technologies and Systems (IJATS), Vol. 5, Issue 3, 2013, 1-13. P. Mitra and C. Poellabauer, “Opportunistic Routing in Mobile Ad-Hoc Networks”, Routing in Opportunistic Networks, Isaac Woungang Ed., Springer, 2013. McCune, R. R., A. Madey and G. R. Madey. “UAV Swarm Command and Control With Agent-Based Modeling”, Book Chapter (under review)
UAVs (DURIP)
CrazyFlie Parrot AR.Drone 2.0
Hummingbird
Dragonfly Dr. Robot Iris
UAV Research: FAA Challenges
• FAA
• Campus in flight path for SBN airport
• Purchasing Office
• Office of Risk Management
• Campus Security
• University Legal
• OK if indoors!
Indoors Facilities
21,000 square feet 614 feet by 210 feet = 128,940 square feet
Movie Removed to Reduce File Size – can be viewed at: http://youtu.be/N-Y2u3loidQ
Acknowledgements
• This research was supported in part under grants from the Air Force Office of Scientific Research: – AFOSR FA9550-11-1-0351 (testbed, virtual SWARMS of UAVs)
– AFOSR FA2386-13-1-3027 (DURIP – physical SWARM of UAVs)
• The National Science Foundation – Award No. 1063084 – REU Site
– Award No. 1062743 – REU Supplement
• GAANN Fellowship (Graduate Assistance in Areas of National Need) from the Department of Education
• Center for Research Computing, University of Notre Dame
• Department of Computer Science & Engineering, Notre Dame
17
Vertex-Oriented Computing
Application to DDDAS
Swarm Application Architecture
19
Large-Scale Graph Processing
• Graphs capture relationships
– Networks of nodes and edges
• Graph analytics are valuable
– Node or graph quantities
• Big Data driving volume
– Millions of nodes, billions of edges
• Can’t centrally process
– Random access impractical
Vertex-Oriented Frameworks
• Distributed graph processing framework – Abstract away details like MPI
• User-defined vertex program
• Iterative execution – Vertices communicate through
messages
• Equivalent result to shared-memory algorithms
21
Machine 1 2
3 4
22
3 6 2 1
3 2 1 1
2 1 1 1
1 1 1 1
1 1 1 1
STEP 0
1
2
3
4
Active
Halt
Min Value
DDDAS and Vertex Computing • DDDAS principles for bottom-up approach
• Lots of Applications
• Power grids – Decentralized control
algorithms
– Disturbance monitoring
• Framework properties – Synchronicity
– Streaming partitioning
THANK YOU!
Questions?
24
Project URL: http://www3.nd.edu/~dddas/AFOSR/