16.400/453J
Human Factors Engineering
Manual Control I
1
Levels of Control
16.400/453
Human
Operator
Display Controller
Human
Operator
Human
Operator
Human
Operator
Human
Operator
Display Controller
Display Controller
Display Controller
Computer
Computer
Computer
Display
Computer
Sensor Actuator
Sensor Actuator
Sensor Actuator
Sensor Actuator
Sensor Actuator
Task
Task
Task
Task
Task
Manual Control
Supervisory Control
Fully Automatic
Control
Image by MIT OpenCourseWare.
Aircraft Basics
16.400/453
Image by MIT OpenCourseWare.
4
Control of Aircraft
16.400/453
– Propulsive
– Aerodynamic
– Gravitational
W
Distance
• Forces on an A/C
A
MT
Time
• Control responses
• Handling qualities
– Disturbance regulation
– Maneuverability
Velocity
– Stabilization
(a)
Time
(b)
Image by MIT OpenCourseWare.
• Reactive vs. proactive human response
– The innate human response to “home in” on a signal
Pilot Input
16.400/453
Actual pitch
attitude θ
Pitch attitude
error θe
Digital flight
control computer
Desired pitch
attitude θc
Cockpit inceptor
Fly-by-wire
Sensed aircraft
motion
Actuator
Control effector
Image by MIT OpenCourseWare.
6
The Basic Pilot/Plant Feedback Loop
16.400/453
+
What
you
want
-
Display
Pilot
Sensor
•
•
•
•
•
Airplane/
Plant
Where
you are
Disturbance
Input
Pilot sees error state (i.e., not on altitude, heading, etc.)
Pilot understands what must be done to correct for error
Pilot corrects
Airplane corrects
Sensed state is communicated to pilot…
7
Some important definitions
16.400/453
• Pilot/operator never needs to respond as long as there is
no error…
• Two input types
• Command
• Disturbance
• Transient vs. continuous
• Continuous
• Periodic vs. random
• Can model these in the frequency domain
• A good assumption?
• Pursuit vs. compensatory displays
• Linear vs. non-linear
• Transfer function
8
• Mathematical relationship between input and output
Transfer Function Interpretation
16.400/453
(a)
Pure gain
o(t) = ki(i)
O = kl
(b)
Pure time delay
o(t) = i(t - τ)
O = e-τSl
1
(c)
τ
Exponential lag
0.63
O=
0
l
T iS + 1
Ti
Image by MIT OpenCourseWare.
9
• These are for
step response
input – why is
this a good
assumption?
• Pure gain
and pure
time delay
have no
effect on
each other
Transfer Function Interpretation, II
16.400/453
• Constantly held position produces constant rate of movement
• Input withdrawn, velocity =0, but new position is achieved
(a)
(b)
o(t) = ∫ i(t) dt
O=
l
S
o(t) = ∫ ∫ i(t) dt
O=
l
S2
First-order
integral control;
velocity control
Second-order
acceleration
control
Image by MIT OpenCourseWare.
• Typical pure 2nd order systems are large mass/high inertia for
constant input force
• Sluggish, unstable when tracked, &requires a series of reverse corrections
Transfer Function Interpretation, III
16.400/453
1
(c)
Exponential lag
0.63
O=
0
l
TiS + 1
Ti
Image by MIT OpenCourseWare.
First order response to error (as opposed to commanded input)
will look like lag (response velocity is proportional to response
error)
o(t) = ∫ i(t) dt
(a)
O=
l
S
First-order
integral control;
velocity control
Image by MIT OpenCourseWare.
11
Human Performance & System Order
16.400/453
• Human information processing causes a delay
• Perceptually humans track position, velocity, and acceleration
changes the best (in that order)
• 0th and 1st order systems are tracked with time delays of
~150-300ms
• 2nd order systems: 400-500ms
• Systems can cause delay
• Computationally
• Communication
• Best human can do is 2 corrections per second
• Serial reaction time experiments show people can make 2.5
decisions per second
• Countering inherent human delays with display
preview/prediction
12
Demo
16.400/453
• http://studiolab.io.tudelft.nl/controltheory/demo.html
13
Order and System Response
16.400/453
Order
Input
Response
to correct
System
response
0
Time
1
2
3
Position
Image by MIT OpenCourseWare.
14
System Order Design Considerations
16.400/453
• Which system order is best for humans?
• 0th and 1st equivalent
• Cost-benefit analysis (time vs. magnitude of response)
• Economy of movement & space
• Level of expertise
• Not mutually exclusive, rate aided displays are a
combination
• 2nd order and higher should be avoided
• Operators must perceive higher-error derivatives
continuously
• Near-constant monitoring reduces cognitive capacity for
other tasks
• Error and subjective workload increase significantly
15
Stability Design Considerations
16.400/453
• High gain = minimal effort
but high likelihood of
overshoot
• Instability exacerbated by
system lags
• Finding the right gains
• PIO
• Caused by high gain & large
lags/time delays in loop
closure
• Inherent aspect of negative
feedback systems
Good
High
Unstable
Oscillatory
Acceptable
Sluggish
Long
Time delay ( τ )
Short
Gain (K)
• Systems with intermediate
gains have lowest error and
easiest to track
Low
Image by MIT OpenCourseWare.
Some PIO footage…
16.400/453
• F-22
• http://www.youtube.com/watch?v=03rz8vqjGdQ
• Gripen
• http://www.youtube.com/watch?v=k6yVU_yYtEc
• Shuttle
• http://www1.dfrc.nasa.gov/Gallery/Movie/STS/640x/EM0084-02.mov
17
Predictive Interface Design
16.400/453
• Should control by error rates vs. error values
• When humans respond to predictive cues, they act as
differentiator and cancel out an integrator (reducing
order)
• Predictive displays can mitigate consequences of
time delay
• Two types
• Preview
• Prediction
• Quickening – where system error is likely to be in the future
• Issues with assumptions
18
Measuring Pilot Response
16.400/453
300
Vertical Speed (fpm)
200
B
100
0
1
21
41
61
C
A
-100
C
RMSE 
 (Y
a
 Yd )
n
-200
-300
Time (sec)
Actual
Desired
+/- 10%
Image by MIT OpenCourseWare.
• Measuring signal strength as opposed to average
values
2
Multi-axis Control
16.400/453
• Cross-coupled
& hierarchical
tasks
• Lower order
variables must
be controlled
to regulate
higher order
variables
• Cognitive
workload &
design
interventions
φ
(a)
δ
∫
Steering wheel
angle
φ
(b)
Aileron position
(rate of roll)
∫
Heading
∫
Bank angle
Lateral position
δ
0
∫
Heading
∫
Lateral
deviation
(c)
Rubber plane
angle
∫
Pitch rate
∫
Pitch angle
∫
Depth
Image by MIT OpenCourseWare.
Inner Loop
Outer Loop
MIT OpenCourseWare
http://ocw.mit.edu
16.400 / 16.453 Human Factors Engineering
Fall 2011
For information about citing these materials or our Terms of Use, visit: http://ocw.mit.edu/terms.