Swarms for Airborne Forward Air Control

1
Swarms for Airborne Forward Air Control
2
Swarm Tasks

• Implemented mission tasks
• Parallel Sweep Search
• Synchronized, multi-point strike
• Layered, multiple-perimeter asset patrolling with
  threat elimination
• Mobile target tracking and surveillance

3
Inspiration Slow FAC and UAVs

• Marine Corps Dragon Eye
• Used for surveillance tactical reconnaissance
• Small UAVs

• Marine Corps OV-10 Bronco
• Used in Slow FAC
• Viet Nam through Desert Storm
• Now used for Fire Fighting in CA

4
Inspiration Marine Corps Doctrinal Publication 1

• Due to the fluid nature of war, gaps will rarely
  be permanent and will usually be fleeting. To
  exploit them demands flexibility and speed. We
  must actively seek out gaps by continuous and
  aggressive reconnaissance. Once we locate them,
  we must exploit them by funneling our forces
  through rapidly. (p. 93, MCDP 1 Warfighting.
  1997.)

5
Inspiration Marine Corps Doctrinal Publication 1

• Combined Arms is the full integration of arms in
  such a way that to counteract one, the enemy must
  become more vulnerable to another. We pose the
  enemy not just with a problem, but with a
  dilemmaa no-win situation. (p. 93, MCDP 1
  Warfighting. 1997.)

6
Goals

• Minimize individual and system complexity through
  task/role differentiation and specialization
• Address
• Fleeting targets of opportunity
• Close air support of troops-in-contact situations

7
Goals (2)

• Minimize need for communication
• Low operational altitude
• Prevent eavesdropping/detection
• Positive target identification and BDA
• Minimize collateral damage
• Ensure graceful degradation of performance
  through decentralized control and redundancy

8
Swarm FAC Applications

• Reconnaissance
• Close Air Support (CAS)
• Interdiction
• Battle Damage Assessment (BDA)
• Search and Rescue (SAR) Support
• Artillery and Naval Gunfire Spotting
• Convoy Support
• Leaflet Dropping
• Communications Relay

9
Hunter-Killer Overview

• Many UAVs with differing operational capabilities
  and physical features can be employed as
  hunter-killer teams
• Hunter finds target
• Killer dispatches target
• A controlled strike mission
• Application of Combined Arms
• Enemy doesnt move to avoid detection but has to
  move to avoid strike

10
Hunter-Killer Overview (2)

• Hunters
• Low and slow with long endurance
• Equipped for reconnaissance
• Deployed in tandem pairs
• One tracks the target
• One serves as a communications relay
• Find and mark targets (laser or GPS)
• Call in strike support.
• Conduct BDA

11
Hunter-Killer Overview (3)

• Killers
• Faster
• Armed with air-to-surface weapons
• On station outside area of interest until
  recruited
• Follow the guidance provided by hunters
• Perform strikes on the target

12
Killer (Striking)
Hunter (Relay)
Hunter (Tracking)
13
Mission Phases

• Search
• Discovery
• Recruitment
• Communications Handshaking
• Target Tracking and Strike Approach
• Target Marking and Strike Arrival
• Strike Clearance and Strike
• BDA
• Break off

14
Target confirmed / Recruit
Strike inbound / Paint target
Possible Target / Descend
Target intact / Recruit
Regrouped / Search
Bombs away / BDA
Go High / Climb
Target destroyed / Regroup
15
Increasing Intelligence

• There are many routes to increase system
  intelligence
• Intelligence efficient decision making
• Acting intelligently often requires a plan
• Plan requires a priori information
• Updated via observation or communication
• May be based on assumptions or implicit
  communication (MCDP 1-3 Tactics)

16
Bayesian Belief Networks

• Bayesian Belief Networks
• Decision making under conditions of uncertainty
• Based on conditional probabilities p(AB)
• System learning Initial probabilities updated
  through observations or experiments
• Have States, Actions, Probabilities, Experiments,
  and Rewards

17
Bayesian Belief Networks (2)

• States (real state of the world)
• S1 observed object is target Type 1
• S2 target Type 2
• S3 not a target Type 3
• Actions
• A1 call in strike
• A2 ignore object and continue searching
• Or conduct experiment

18
Bayesian Belief Networks (3)

• Prior probabilities
• Prob(S1) 0.2
• Prob(S2) 0.3
• Prob(S3) 0.5
• Payoff matrix

19
Bayesian Belief Networks (4)

• Experiment
• Flying low for a closer look
• Messages from this experiment are
• Z1 object is target Type 1
• Z2 object is target Type 2
• Z3 object is not a target Type 3

20
Bayesian Belief Networks (5)

• Conditional probabilities
• Prob(Z1S1) 0.8
• Prob(Z2S1) 0.05
• Prob(Z3S1) 0.15
• Prob(Z1S2) 0.03
• Prob(Z3S3) 0.9

21
Observe T1
0.80
Type1
0.05
Observe T2
0.2
Observe T3
0.15
Observe T1
0.25
Type2
0.25
Observe T2
Search
0.3
Observe T3
0.5
Observe T1
0.25
0.5
0.25
Type3
Observe T2
Observe T3
0.5
Z
S
E
22
(No Transcript)