16.886
Human Factors for
Autonomous Formation Flying
Mary (Missy) Cummings
Humans and Automation Division
Aeronautics & Astronautics
Formation Flight Basics
? Section v. division differences
? The tighter the formation, the
higher the workload
– Implications for system
management
? Wingman uses significantly
more fuel
? Formation flights account for
14 percent of all midairs
U.S. Navy photo
Autonomous Formation Flying in
Action
Keeping Position
? http://www.dfrc.nasa.gov/Gallery/Movie/AFF/HTML/EM-0081-01.html
Refueling
? http://www.dfrc.nasa.gov/Gallery/Movie/AAR/HTML/EM-0053-01.html
Position Issues
? Wing must stay within 10% of lead’s wingspan for 30%
fuel savings (Proud et al., 1999):
– Military: C-17: 171’, C-141: 160’, C-5: 223’
– Commercial: 747: 196’, A300: 147’
? Previous flights
– 2001: 55’ (two F/A-18s) (12% savings)
– 2003: 200’ (DC-8/F/A-18) (29% savings)
? Pilots most sensitive to changes in roll
– The most significant vortex disturbance when positioned for
maximum drag reduction is a strong rolling moment effect
(Hansen et al., 2002)
? Vortex turbulence generally avoided
– Step up and down
Spatial Disorientation
? A false perception of one’s position and motion
with respect to the earth
– Sensory illusions
? Primarily due to transition between inside/outside
scans
? Especially prominent in transition between
VMC/IMC in formation flying
– False horizons
– “The leans are most commonly felt when flying
formation on the wing in the weather or at night (Wright
Patt).”
Vigilance Issues
? Sustained attention
– Not a human strength
? Vigilance can deteriorate significantly after
30 mins
? What is the threshold for pilot intervention?
– False alarms
– Cost of premature pilot intervention
? Reaction times could be affected
? Alerting systems can help
– Advisories versus warnings
Previous Flight Test Results
? Air Force Flight Test Center, 2 & 3 T-38s
– October 2001
? Pilot workload assessments
? They found that maintaining the minimum drag
formation was a comparable workload to
maintaining other types of formations. (not a good
thing)
? The longest duration the pilots could maintain the
position operationally was approximately 20-30
minutes.
– Recall vigilance discussion
General Research Areas
? Alerting systems
–Prediction
– Probabilistic representations
– Signal detection theory
? 2D versus 3D displays
– Is one remarkably better or more confusing than the
other?
– Which one produces more false alarms?
? Situation awareness
– How do these design issues impact pilot’s SA for
both AFF alerting system as well as other systems?
Signal Detection Theory
Correct
Rejection
Miss
False AlarmHit
True in the World
Pilot
Response
Signal
Occurred
Signal Did
Not Occur
Signal
Occurred
Signal Did
Not Occur
d’ = sensitivity to signal
B = criterion
Graph generated at http://psych.hanover.edu/Krantz/STD/
Receiver Operating Curves
Graph generated at http://psych.hanover.edu/Krantz/STD/
2D Versus 3D Displays
? Human is supervising, not actively flying
? Need to know where the system is now and where it is
predicted to be at some point in the future.
? Is one type a better alerting system?
? Does one promote SA more than the other?
Situation Awareness
? Knowing what is going on around you both
now and in the near-term future
– Geospatial
– Temporal
– System
– Environmental
? Mental model
– Categorization mapping
? Not the same as workload
? Automation impact
Other Research Areas
? Relationship of distance/size of aircraft to pilot
workload/vigilance
– Ability to respond to problems/failures
? Trust issues
– Stress at close ranges
? Long range missions
– Both physical and cognitive fatigue
? Division issues
? Take the human out of the loop?
– Ground controller