Vous êtes sur la page 1sur 24

05WAC-62

Fundamentals of Fly-By-Wire Augmented Manual


Control
Anthony A. Lambregts

FAA Chief Scientist Flight Guidance and Control

ABSTRACT flight decks, airplane handling qualities and


common type ratings among different
This paper examines various Fly-By-Wire models of a family of airplanes;
(FBW) augmented manual control design enhancing flight safety, by providing
options and criteria for achieving desired protection against exceeding safety limits in
handling qualities. Design concepts with airspeed, load factor, bank angle, side slip,
different response types will be reviewed. The etc., thereby reducing the risk of loss of
author will re-interpret some of the classical control and providing the capability to safely
Handling Qualities criteria in light of desired achieve maximum airplane performance,
response characteristics in the time domain, when needed;
which can then be evaluated through simulation reducing development and production costs.
time histories. A design methodology is
discussed that can provide the same aircraft FBW design holds out the promise of
responses to pilot control inputs for different simplifying and standardizing the flight control
response types, e.g. Pitch Rate Command/ system design, while shortening the traditionally
Attitude Hold and Flight Path Angle (FPA) Rate complex, laborious and lengthy flight control
Command / FPA Hold designs. The role of system design development process and reducing
various feedback and feed forward signal paths associated costs and risks.
and ways to minimize the need for gain Change embodies risk in itself, but failure to
scheduling, is discussed. change, over time, ensures loosing out to the
competition. Airplane manufactures must
INTRODUCTION therefore make many difficult FBW design
decisions. They must decide
The motivations for incorporating Fly-by-Wire how to evolve the flight guidance and
flight control system design in transport and top control system, relative to the last
end General Aviation aircraft are manifold. At production airplane model (training
the top of the list is perhaps the aircraft implications);
manufactures need to offer a distinct product overall functional architecture and
that is very competitive to potential customers in integration of manual and automatic flight
terms economics, safety and passenger appeal. guidance and control functions;
At the design level this can be interpreted as a overall system hardware architecture and
need for redundancy management concepts
optimization of aerodynamic performance, o use of mechanical or electronic analog
by operating with the airplane center of back up control modes;
gravity moved aft, to reduce trim drag; specific FBW flight control system details
reduction in structural weight, by replacing o control algorithm response type;
mechanical system elements with lighter o control inceptor type (traditional or
electric/ electronic components; sidestick);
reduction of aircraft maintenance and spare o feel system type (active or passive);
parts inventory associated with mechanical o control inceptor back drive during
flight control systems (e.g. rigging, automatic flight control;
lubrication, etc.); o envelope protection functions and
reduction in flight crew training methods (hard or soft limiting);
requirements, by providing standardized

1
o control surface actuation type and position essence an automatic control system with a
loop closure method. specialized input device (the pilots control
inceptor). So it is reasonable to expect continuity
At the simplest FBW system level, the in design philosophy for augmented manual and
mechanical signaling of the flight control surface fully automatic control and sharing control
actuators is replaced by electric/electronic functions. Thus, the introduction of FBW design
signaling, with little or no control augmentation forces the manufactures to make a value/cost
that would require additional inertial and air judgment on what system elements can be
data sensors, computers, software and data bus shared, what new and unique elements are
interfaces. For high-end FBW designs everything needed, and which legacy elements from a
may be included. The Federal Airworthiness previous era can be dispensed with, to simplify
Requirements are silent on many of the specific the design and reduce costs. In general, simpler
design choices. is better. Simplicity often enhances safety,
based on the premise that what isnt there, isnt
The relative merits of various control algorithms needed, cant fail and cause a safety liability.
response types have been debated extensively. These principles can be used to formulate the
The handling qualities of classical un- following high-level FBW flight control system
augmented airplanes, sometimes referred to as design objectives. In this paper, transport
an alpha-command response type, should not be category or high-end general aviation airplanes
idealized, because modern FBW technologies are assumed, which already include powered
can greatly improve airplane handling qualities actuators on the control surfaces.
and safety, relative to the best classical un-
augmented airplane designs. In fact, handling DISCUSSION
qualities depend more on the detailed design
features than on the response type. For example, High-Level FBW Control System Design
airplane response characteristics at constant Objectives
speed can be made to look identical for a pitch
rate command, an alpha command, an nz- Augmented manual and automatic control
command and a flight path angle rate command modes should be functionally integrated,
control concept, as will be shown in this paper. It based on a common design philosophy.
requires sufficient knowledge to select the key
design features and sufficient experience to put FBW technology can and should be used to
the design together correctly. The designer must simplify the pilot's airplane control task and
also understand the importance of suitable visual improve flight safety.
display and tactile cueing information the pilot
uses to close the control loop while executing a The augmented manual mode must provide
specific task. Therefore, controls and displays suitable handling qualities for all required
must be suitable for the task and well integrated. control tasks throughout the flight envelope.
For example, satisfactory handling qualities for
direct flight path angle control concept cannot be Flight envelope protection features must be
achieved without suitably quickened flight path implemented in the simplest way, serving
angle display information and even then such a both the augmented manual and the
concept may not be suitable for the landing task. automatic modes. Envelope protection
modes must not unnecessarily compromise
The more sophisticated the FBW system airplane performance.
functional design is, the more it is necessary to
re-examine the entire Flight Guidance and The first objective implies the use of a
Control system architecture with all its generalized multi-input/multi-output (MIMO)
dependencies, as well as the entire design control concept that accommodates all automatic
integration process. To take full advantage of and augmented manual control modes by
very capable modern hardware and software selectively concatenating needed building
technologies, consideration must be given to re- blocks. Decoupled control responses should be
integrate all automatic and augmented manual provided by simultaneously coordinated
control functions and hosting them with the pitch/thrust control and roll/ yaw control
fewest possible hardware components. A commands for airplane control in the vertical and
sophisticated FBW control algorithm is in horizontal plane. Pitch and roll inner loops

2
should be shared between the augmented manual digital Mission Computer system with 384K
and all automatic control modes. Pilot manual bytes of memory. Condor used all electric
control augmentation should be provided by a actuation with Mil-Spec-1553 data bus
separate module that further shapes the communications. The Condor program
augmented airplane responses to achieve the convincingly demonstrated the advantages of
desired handling qualities. generic functionally integrated FG&C: it made a
fully automatic take off and landing on first
Earlier Control Function Integration flight, flew to 68,000 ft altitude and successfully
Research handled a considerable number of very
challenging failure conditions, without human
The industry has long felt a need for simpler, intervention or the need for significant design
more generalized and better integrated flight improvements. A further discussion on control
control systems. Recently, the NASA/FAA function integration as it relates to safety is
Aviation Safety Program has again identified the found in [3].
need for safer / simpler automatic Flight
Guidance and Control (FG&C) systems. In the The TECS FPARC/FPAH manual control
early eighties NASA sponsored research to this concept was again used in the 7J7 technology
end, resulting in a design called Total Energy development program and flown on a B757. It
Control System (TECS). This design integrates was considered for implementation on the B777
all vertical flight path and speed control modes, airplane, but in the end a C*U algorithm was
including an augmented manual control mode developed and implemented to provide control
[1]. TECS uses a generalized energy based flight response characteristics that are very similar to
path angle and longitudinal acceleration control traditional un-augmented airplane control
strategy to overcome the long-standing design, responses.
operations and safety deficiencies identified in
existing production FG&C systems. It eliminates The FPARC/FPAH control concept has still not
unnecessary mode overlap and associated found its way onto modern FBW airplanes
complexity and provides all desired envelope because it is seen as a complex higher order
protection safety features. For augmented design that embodies considerable handling
manual control TECS uses a Flight Path Angle qualities and design certification risks. This
Rate Command /Flight Path Angle Hold paper argues that this impression is false and that
(FPARC/FPAH) concept, which was first a FPARC/FPAH augmented manual control
developed and flight tested as a control wheel concept can provide superb handling qualities for
steering mode on the NASA TCV program in the a variety of tasks with relative ease and
late seventies [2]. The complete TECS system simplicity, compared to other concepts.
with all its modes was extensively evaluated in
pilot-in-the-loop simulations and validated in FBW Augmented Manual Control
flight test on the NASA B737 technology
demonstration airplane in the early eighties. Handling qualities requirements. The military
TECS was very successfully applied on the handling qualities requirements and discussions
Condor UAV project. Its counterpart generalized may be found in [4, 5]. Additional Handling
lateral-directional control system, called Total qualities requirements are found in [6]. Most of
Heading Control System (THCS), was also these requirements were developed for various
developed on the Condor project. Condor was a categories of un-augmented military airplanes.
very large (200 ft wingspan) all-composite High With the advent of augmented control system the
Altitude Long Endurance autonomous UAV. need arose to reinterpret the basic handling
TECS and THCS helped to substantially reduce qualities requirement so it is possible to apply
hardware /software complexity, by eliminating them on highly augmented or High Order
the traditionally separate Autothrottle, Yaw manual control systems. These efforts have met
Damper, Turn Coordination and Trust with mixed success. As a result various designers
Asymmetry Compensation functions. These and companies have developed their own set of
functions are generically and integrally provided design criteria. Regardless of the manual control
as part of the multi-axes control algorithms. All augmentation system complexity considered, the
Condor Flight Guidance and Control (FG&C), important handling qualities requirements can be
FDIR and mission management functions were expressed by the following airplane response
implemented in a dual channel fail operative

3
characteristics for step pilot control input, see satisfactory response sensitivity;
figure 1. : satisfactory response damping, minimal
a smooth buildup of the rate of change of the target overshoot and short settling time at
task related control variable (pitch attitude, maneuver termination.
flight path angle or vertical speed), in direct
proportion to the pilot's input, converging In addition:
quickly on a pure integrator-like (K/S) the pilot must be given display information
response for a sustained control input, appropriate to the task, to initiate the
without undue initial response delay, or maneuver, gauge his progress toward the
undo pitch rate overshoots at maneuver target and allow him to properly anticipate
initiation; the termination of the maneuver;
a minimal steady state response lag of the the tactile and motion cues must be in
task related control variable, relative to K/S harmony with the pilot's control input and
(adequate bandwidth, gain and phase expectation and not cause the pilot to react
margins), consistent with the overall inappropriately.
airplane physical limitations and satisfactory
ride qualities at all passenger locations;

No overshoot,
K Short settling time
Response cmd
, S

Low SS response lag


Cmd
Correct stick sensitivity

stick Smooth start-up

Time
Figure 1. Desired Response Characteristics

If the augmented manual control system is well understanding of airplane dynamics, a lot of
designed, an active side stick cannot provide the design experience and attention to every design
pilot with much additional airplane information detail.
to further improve the handling qualities. In that
case it may increase the design risk due to extra Control algorithm options. There are many
complexity, failure modes and the creation of FBW control algorithm options. Each can be
possible disharmonious control force cues. The described by a set of three basic design elements,
use of an active side stick would certainly which may not all be employed :
increase design cost. Stability Augmentation function;
Pilot Command Augmentation function;
Good "handling qualities" imply that the pilot Command Reference Hold function for
can do everything that the task demands, without unattended, pilot out of the loop operation.
having to compensate mentally or physically for Distinctive concepts, in increasing order of
bad or annoying design characteristics and other complexity, include:
shortcomings. In that case the pilot perceives simple electrical signaling of the control
only airplane responses in precise harmony with surfaces with little or no stability and
his control inputs and expectations. Achieving command augmentation (Embaer 170);
all this is not an easy task. It requires a good

4
pitch rate stability plus command response at which nz = & * Vco /g. This C* handling
augmentation (Gripen and Eurofighter); qualities criterion has long been shown to be
Pitch Rate Command /Pitch Attitude Hold seriously flawed [8]. A popular explanation of
concept (Space Shuttle); this criterion is that during maneuvering the pilot
Flight Path Angle Rate Command/ Flight responds predominately to pitch rate cues at low
Path Angle Hold concept (Dassault Falcon speed and predominately to incremental normal
7X) load factor cues at high speed. Thus, the load
C* concept (Dassault Mirage fighters, factor cue and the pitch rate cues are said to be
Airbus A320 - A380 models); equally compelling at the crossover speed.
C*U concept (Boeing 777); This explanation is also flawed, because these
cues are perceived by different human sensory
Control concepts are often characterized by their channels and there is no objective way to
response type, but this is not very meaningful, establish equivalence between these cues.
because the response characteristics within one Furthermore, the incremental load factor the pilot
algorithm type can vary dramatically, depending generates during normal transport airplane
on the specific system elements employed and maneuvering is not speed dependant, so there is
the tuning of the algorithm, as will be no reason to believe pilots use load factor at one
demonstrated later on in this paper. Some experts end of the speed range, but not at the other end.
like to include an angle of attack command For passenger comfort pilots simply maneuver
system as a distinct algorithm possibility. the airplane with approximately the same
However, the use of an angle of attack sensor in relatively low g-levels for all flight conditions.
a basic control algorithm is a difficult The natural airplane dynamics dictate the
proposition. Furthermore, assuming control inherent relationship between the steady state
responses at constant airspeed, incremental angle incremental load factor (nz)ss and the steady
of attack is equivalent to incremental normal
state pitch rate ( & )ss : (nz )ss = ( & )ss *Vtrue/g.
load factor and equivalent to flight path angle
The pilot has no way to influence this
rate.
relationship.
C* and C*U will be discussed first because these
The C* algorithm, presented in figure 2, uses
algorithms are used in the first generation of
FBW transport airplanes. Next, the more feedback loops of & * Vco /g and nz signals fed
fundamental Pitch Rate Command / Pitch through an integral signal path, in addition to a
Attitude Hold (PRC/PAH) algorithm will be proportional pitch rate signal path. nz is derived
discussed, because some fundamental stability from body mounted accelerometer sensitive
and command augmentation issues are easier to along the airplane Z-axis, compensated for the
discuss and understand with respect a PRC/PAH signal bias due to pitch and roll attitude. The
system. This will help introduce the discussion integral control signal path of & * Vco /g and
related to the FPARC/FPAH algorithm nz signals is a very peculiar arrangement,
development. because integration makes the design sensitive to
sensor bias errors, so high precision signals are
C* Control Algorithm. The first FBW needed, otherwise the control algorithm will
commercial transport airplane, the Airbus A320, diverge, unless the pilot provides corrective
uses a C* algorithm. The C* algorithm derives inputs. The pilots control inceptor signal is fed
its name from a handling qualities criterion forward through the same integral signal path
called C*, proposed in the mid sixties [7]. This with a certain gain, as well as through a
criterion postulates that an airplane has good proportional signal path, which may include a
handling qualities, if the time history of the response-shaping filter. The pilots input signal
quantity dubbed C*, obtained from the airplane path to the feedback control integrator is said to
responses to a pitch control input, falls within a command a value of the C* quantity and thus is
certain envelope. The C*quantity is defined as called the C* command. The algorithm is
the sum of the incremental normal load factor complex, higher order, and requires extensive
(nz) and normalized pitch rate ( & *Vco /g). gain scheduling with flight condition for
The normalization factor is the crossover achieving the desired tuned response for each
airspeed (Vco ) divided by the gravity constant flight condition.
(g). The crossover speed is defined as the speed

5
Figure 2. C* /C*U algorithm

The higher order aspects relate to the use of an result, the C* algorithm does not meet the
accelerometer signal, the integral control signal Federal Airworthiness requirement for stick
path and the response shaping filter(s), making force/speed stability. Stick force / speed stability
the airplane response transfer functions of higher requires that a pull force is needed to achieve and
order than those of an un-augmented airplane, or maintain a steady state speed decrease and vice
those of an airplane with only classical stability versa, and that the speed returns to trim speed
augmentation. when the control is released. An Equivalent
Safety solution to meet the intent of the speed
In effect, integration of the vertical stability certification requirement is to provide
accelerometer and pitch rate gyro signals results speed envelope protection. In practice, the
in pseudo attitude and vertical speed feedbacks. tendency for the speed to diverge if thrust is not
The dominant effects of pitch attitude feedback controlled carefully, causes the pilot to resort to
are an increase in Short Period frequency and the use of the autothrottle to hold speed.
Phugoid damping. Vertical speed feedback also
adds to the Phugoid damping. The net result is a Speed stability without simultaneous thrust
non-classical, heavily damped nz - response to control tends to undue the pilots flight path
pilot control inputs, with pseudo vertical speed control input. In a comparison study between
hold characteristic when the pilot leaves the control augmentation algorithms with and
control loop. It is not a true vertical speed hold, without speed stability, Mooij [9] found a small
because the integrator continues to integrate its advantage in flight path control handling
sensor inputs which are affected by external qualities during final approach for a PRC/PAH
airplane disturbances, airplane configuration control algorithm that has no speed stability
changes and thrust changes. As a result, over (albeit at the expense of speed deviation). This
time the actual vertical speed tends to drift away may be explained by the fact that the nz response
and the pilot must periodically reenter the control is sustained longer without being washed out and
loop to keep the airplane on a desired earth reversed due to the speed deviation feedback, as
referenced flight path. A change in vertical is the case for a speed stable airplane. This
speed also causes a corresponding longitudinal clearly indicates that handling qualities cannot be
acceleration, unless compensated by a thrust fully assessed based on constant speed Short
change. Therefore the pseudo vertical speed Period criteria alone. The final assessment needs
hold characteristic destroys airplane speed to be based on complete high fidelity airplane
stability and the concept of a trim speed. As a dynamics, taking into account the effects of

6
thrust/speed control dynamics and the effect of The resulting responses look more or less
non-minimum phase flight path response due to classical depending on the Phugoid damping
tail lift. provided, allowing the C*U algorithm to be
certified under the basic Federal Airworthiness
Another characteristic of the C* algorithm is its Requirements, without the Special Condition
response to normal load factor change when the issued for the C* algorithm. The pilot can
airplane enters a banked turn. The feedback of change the trim speed through the use of
integrated nz appears like a spurious increase in conventional trim switches on the column,
vertical speed to which the controller responds activation of which commands a rate of change
with a nose down control command, in effect of trim speed. On the B777, the so developed
increasing the maneuver stability. This trim speed command is not synchronized with
characteristic may require additional the autothrottle speed command selected on the
compensation in order to achieve satisfactory FG&C mode panel. This can lead to thrust and
handling qualities. Classical airplane maneuver flight path divergence if the two speed
stability provides the pilot with a stick force cue commands differ when the pilot leaves the
due to nz when the airplane enters into banked control loop, because the autothrottle and the
turn, which is a useful characteristic for hand C*U control algorithm each pursue their own
flying, especially in Instrument Meteorological speed command. In that case the C*U algorithm
Conditions. When a passive feel system is used wins and the throttles end up either at the
there is no stick force change due to nz to alert forward or the aft limit, because the C*U has far
the pilot. Excessive maneuver stability at low greater authority controlling the elevator than the
bank angles or neutral maneuver and spiral autothrottle controlling thrust. If this happens the
stability at high bank angles result in pilot must recognize the situation and take
deterioration of handling qualities ratings. With corrective action. In future designs there should
enough attention to design details and correct only be one speed command and the elevator and
tuning of the algorithm for the flight condition throttle should not both control speed.
and task, the C*control algorithm can provide
excellent handling qualities with low pilot Handling Qualities characteristics similar to the
workload. C* and C*U algorithms can be achieved in a
simpler, more classical way using fewer sensors,
C*U control algorithm. The C*U algorithm is by replacing the integral signal path of nz and
the same as the C* algorithm, except it features & with proportional feedback paths of vertical
an additional airspeed feedback loop and a speed speed and pitch attitude, using equivalent gains,
command (trim) input operated by the pilot, see and putting the pilots input signal integrator
figure 2. Since airspeed (error) is the highest outside the feedback loop. This in effect changes
order state variable fed back to the integrator, the the algorithm into an nz-command /vertical speed
C*U algorithm drives the speed back to the hold concept, depicted in figure 3. This
commanded trim speed when the pilot leaves the approach eliminates the need for pitch and roll
control loop, thereby re-establishing in principle attitude compensation for nz and eliminates the
conventional lowly damped Phugoid behavior. vertical speed drift tendency because the vertical
But since low Phugoid damping is not speed command, developed purely by
particularly desirable and return to trim speed is integration of the pilots input, is no longer
the objective, an additional longitudinal affected by feedback signals.
acceleration feedback may be added to achieve a
better-damped speed and flight path responses. If the designer wants to preserve classical speed
However, too high a Phugoid frequency and stability and use FBW design mainly for other
damping interferes with flight path control and reasons than to improve handling qualities, then
adversely affects the flight path control handling a much simpler system may be used. For
characteristics. Therefore the speed error example, on the EMB 170-190 airplanes the only
feedback gain must be kept low, otherwise the control augmentation provided is pilot inceptor
flight path change is undone by the control sensitivity scheduling.
algorithm, when the pilot leaves the control loop.

7
Stick
PFD
Throttle

T
Command engine
sensor
Augmentation
Airplane
Kstick + e
S +_ actuator

Kq q
+
+
+ K
K h& h&

Figure 3. C*, evolved into an nz- command/Vertical Speed Hold algorithm

Pitch Rate Command / Pitch Attitude Hold (response shaping filter(s)) in the pilots forward
Algorithm. In this type of algorithm, pitch loop signal path, as presented in figure 4. Many
attitude feedback is an essential element of the variations are possible. To create the pitch
system, but the pitch attitude feedback loop may attitude hold function requires a pitch attitude
not need to be closed during pilot maneuvering. command generation function and a pitch
The pilots control input may simply command attitude feedback. The pitch attitude command
pitch rate or elevator position, with or without generation can be a simple pitch attitude
explicit pitch rate feedback to augment the short track/hold function that memorizes the achieved
period stability. Command response shaping may pitch attitude while the pilot is in the control
be accomplished by control inceptor sensitivity loop, and holds its last value when the pilot
scheduling and possible command augmentation leaves the control loop.

Stick PFD
Throttle

T
sensor Command engine
Augmentation
Airplane
e
Gain
+
++ +_ actuator

error Stability q
Track augmentation
&

Hold
Figure 4. Pitch Rate Command /Pitch Attitude Hold Concept.

8
The pitch attitude error is only fed back to the Stability augmentation. First the Short Period
elevator control when the pilot leaves the control frequency and damping are augmented to some
loop (column or stick centered). Issues with such ideal middle of the road values, taken from
a design include the difficulty in achieving the Mil Specs Level I boundary box, e.g. sp = 2
smooth maneuver initiation and termination rad/sec and sp = 1.0. A damping coeffient sp
without excessive pitch attitude drop back and = 1.0 is selected because it results in a lower
long term droop. An alternate approach is to phase rate around the critical bandwidth
close the pitch rate and pitch attitude feedback frequency. It may be necessary to choose sp on
loops full time and develop the pitch attitude the low side to safely accommodate the
command by integration of the pilots control automatic control modes with the same inner
input signal, outside the feedback loop. In that loop design, without having to fear problems
case, the integrator in the pilots forward loop associated with too high inner loop gains, which
signal path forms an ideal K/S response model, can cause adverse coupling with airplane
against which to shape the pitch attitude structural modes and interaction with control
response. With all these design approaches the surface actuator dynamics. As will be shown, the
problem is how to shape the response in a augmented Short Period frequency, resulting
simple reliable way to a desired ideal response from closing only the pitch rate and pitch attitude
(e.g. Figure 1), which may be different for loops, will not appreciably affect the final
different flight conditions and for different tasks. attitude response, after the second and third
design steps have been completed. So, selection
When reading many papers on FBW control of a relatively low augmented short period
system development, one cannot escape the frequency will not compromise the final designs
impression that there is no well-established low handling qualities. Selection of sp = 2 rad/sec
risk systematic design process. This may perhaps and sp = 1.0, requires normalized loop gains
explain the many difficulties encountered in for both pitch rate and angle of attack of 4,
development of satisfactory designs. It is often neglecting the effects of actuator dynamics. The
difficult to discern in a specific algorithm what effect of additional actuator dynamics will be
the function of a specific design feature is. The discussed later. In this context the normalized
end product often appears to be the result of angle of attack (AOA) loop gain is the sum of
empirical trial and error methods. Some design
methods have resulted in higher order systems, the internal AOA feedback due to Cm and the

which cannot be analyzed directly using classical external pitch attitude feedback. Likewise, the
Handling Qualities evaluation criteria. Lower normalized loop gain of pitch rate is the sum
Order Equivalent System representations must of the airplanes internal pitch rate feedback due
be developed for such systems. Even with to Cm and the external pitch rate feedback. In
reasonably good Lower Order Equivalent q
System representations, success with analytical a simplified analysis based on constant speed
assessment and prediction of handling qualities Short Period airplane dynamics, pitch attitude
has been limited. As a result, final system feedback fulfills the same role as AOA feedback.
development still relies heavily on costly high In this process of determining normalized loop
fidelity pilot in the loop simulations and flight gains moments of inertia, control surface
testing. Thus simpler, more transparent design effectiveness and the effect of dynamic pressure
processes are needed, that produce reliable and and normalization constants involved (e.g. Wing
analyzable algorithms. area and mean aerodynamic cord) must be
properly accounted for.
In this paper a systematic analytic approach,
based on simple physics principles is proposed. Static Airplane Model Inversion. It is easier to
The design problem is broken down into three work with normalized loop gains, because
functional elements: normalized feedback gains can be directly
stability augmentation, related to the airplanes dynamic mode
pilot command augmentation characteristics. One simple approach which
development of command reference and makes this possible is to use a book keeping
control tracking or Hold function, for trick, by first canceling the airplanes internal
operation with the control inceptor at feedbacks of state variables that have important
neutral. stability derivatives associated with them, and
then building up the desired airplane dynamics

9
New SP SP model inversion Unaugmented Aircraft SP model
Dynamics
design q
cm q
K q q
ec cm q

+- +-
1 1 1
Cmd
- - cm e
Act cm e +++ S S
Iyy 1
K _ _ cm cm Sc
S +1
qSc I YY
_ 2


_
q -+

Figure 5. Static Model Inversion based Augmented SP Design

by new external feedbacks, as shown in figure 5. effects can be neglected. There is no reason why
Cancellation of the internal airplane feedbacks is the normalized control algorithm, once checked
achieved by externally feeding back these state out, would not work properly when coupled to
variables to the control surface with the same the actual airplane, except in case the airplane
gain and opposite sign of the internal feedbacks. model used in the design does not properly
This reduces the airplane dynamics to a series of represent the actual airplane dynamics. In that
concatenated integrators. This author prefers to case the right thing to do is to fix the model and
call this cancellation process, static airplane its static inversion, and leave normalized control
model inversion. Again, the inversion process algorithm alone! In fact, this approach holds out
must properly account for all elements in the hope to reduce overall risk and design
control loops (moments of inertia, control development and flight testing time, because the
surface effectiveness, dynamic pressure, etc.) process is very straight forward and suitable for
Use of angle of attack (AOA) feedback in this automation.
static inversion process may not be a satisfactory
approach, but this issue will be dealt with later. The desired new airplane dynamics are built up
The control algorithm design process can then by closing pitch rate and pitch attitude feedback
proceed without consideration of the actual loops around the successive integrators that
airplane, control surface and actuator dynamics produce the airplane states, as shown in figure 6.
details, producing a normalized control The first one integrates pitch acceleration and
algorithm design that is airplane independent. produces pitch rate, the second one integrates
In the end, the airplane inversion can be undone pitch rate and produces pitch attitude. Now, for
by reversed bookkeeping, if desired, to create such a system of concatenated loop closures
the traditional single feedback path control around successive integrators, it can be shown
augmentation design, but this results in complex that all the resulting system poles will line up on
flight condition dependant feedback gain the negative real axis, if the normalized gains
schedules. Keeping the inversion feedbacks and associated with the control loop for each state
using the separate normalized stability variable is at least a factor 4 smaller than the
augmentation feedback system has the advantage gain for the adjacent inner loop. Following this
of eliminating gain schedules in the core control rule produces an extremely conservative/robust
algorithm. Accurate modeling and inversion of control system. So, for the above selected sp = 2
the actual airplane dynamics then become the rad/sec and sp = 1.0 , it follows that the new
responsibility of the airplane dynamic modeling normalized pitch rate gain Kq = 4 and that the
department! This process does not introduce new new normalized pitch attitude feedback gain K
risks, because as before, the same simplifying = 1 (concatenated form), resulting in both short
assumptions can be made as to which small period poles located on the real axis at S = -2.

10
Figure 6. Normalized Pitch Rate Command / Pitch Attitude Hold Algorithm

Here, pitch attitude feedback fulfills the role of due to airplane trim change when the algorithm
AOA for SP dynamics augmentation and is used on a real airplane. Following the above
provides the basis for pitch attitude control. Now principles a gain KI =.25 may be selected for the
integral control feedback is added to the new feedback integrator signal. The resulting
algorithm, to prevent large pitch attitude droop transfer function is

KIKKq 1 1
(1) = = =
cmd S + K qS + K K qS + K I K K q
3 2
S + 4S + 4S + 1 (S + 2.618)(S + 1)(S + 0.382)
3 2

Pilot Command Augmentation. Now, in order a free integrator in the denominator and have a
to make it possible to turn the above /cmd second order polynomial in S in the numerator.
transfer function into classical looking /stick (One numerator zero will be placed to cancel one
transfer function into a 3rd order denominator denominator pole). Therefore the following form
that includes one free integrator and a first order is selected for this relationship:
numerator, the cmd /stick relationship must have

cmd K stick 1
(2) = (K FFPS2 + K FFIS + 1) = K stick (K FFPS + K FFI + )
stick S S
The resulting normalized /stick transfer function becomes

K stick (K FFPS2 + K FFIS + 1)


(3) =
stick S {(1/K I K K q )S3 + (1/K I K )S2 + (1/K I )S + 1}

This transfer function represents the entire produced at the output of the stick command
normalized pitch attitude control algorithm integrator;
shown in figure 6. The feed forward gains KFFI KFFP is a direct feed through gain from the
and KFFP provide the command response shaping stick to the final control effector command
tools. By selecting the feed forward gains KFFI (pitch acceleration in this algorithm model).
and KFFP such that one of the poles is cancelled Thus, KFFP controls the initial pitch
by one of the zeros, the transfer function is acceleration to a stick input. With complete
reduced to the precise classical third order form airplane dynamics, the gain KFFP determines
of the /stick transfer function of an un- the value of Control Anticipation Parameter
augmented airplane. Expressed in control &&
(CAP), defined as t =0 .
handling qualities terms:
The gain Kstick represents the stick
n zSteady State
sensitivity - steady state pitch rate per stick : When the feed forward command response
qsteady state /stick ; shaping gains KFFI and KFFP are set to zero, the
The gains KI and KFFI control the steady steady state pitch attitude response lag to a step
state pitch attitude response lag relative to stick command is =1/KI = 4 seconds for KI
the command response model stickKS/S, =.25 , relative to the output of the stick command
integrator, which represents the pitch attitude

11
command. For the general case with non-zero words: minimize pole zero mismatch. For
KFFI and KFFP , is determined by the difference a selected greater than the response lag of
between the coefficient of the S1 of the the transfer function (3) with KFFI and KFFP
denominator and the coefficient of S1 of the set to zero, there should be no pitch rate
numerator of the normalized transfer function: overshoot. For a selected smaller than the
= 1/KI -KFFI. Thus, to achieve = 0 with KI response lag with KFFI and KFFP set to zero,
=.25 requires a gain KFFI = 4. There are several there should be no more than one significant
ways to determine KFFP : overshoot, to minimize residual wobble due
1. Dictate that the denominator pole with the to pole-zero residues. At least one overshoot
slowest dynamics will be cancelled by one is unavoidable in this case.
of the numerator zeros. Then the remaining 3. Selection of equal to a time constant
numerator zero must be selected to provide corresponding to one of the denominator
the desired steady state response lag of poles results in a special case for which the
pitch attitude relative to KIstick /S. For numerator zeros will cancel the other two
example, using the above gain selection denominator poles. The response then
KI =.25, K = 1, Kq = 4, canceling the root at reduces to pure first order. For example, a
S = -.382 and demanding a final = 0, pure first order response with = .382,
requires KFFI = 4.0 and KFFP = 4 - .382 = relative to the cmd developed at the output of
3.618. Cancellation of the pole with the the forward path integrator, is achieved by
slowest dynamics is a good strategy, selection of KFFI = 3.618 and KFFP = 2.618.
because the pole with the slowest dynamics This possibility may be especially attractive
tends to produce the largest pole zero if one of the denominator poles can be
mismatch residue, which manifests itself in placed to correspond to the final desired .
a residual response wobble at the initiation Either one of these procedures assures that the
and termination of a maneuver. pitch attitude response is smooth and meets the
2. Directly compute KFFI for the selected , response objectives. In all practical cases the
then experimentally determine KFFP such feed forward response augmentation gains will
that the pitch attitude response converges as be used to reduce the pitch attitude response lag
fast as possible to KIstick /S , with minimal of the transfer function (3) with KFFI and KFFP
pitch rate/pitch attitude wobble, in other set to zero.

Pitch Rate Cmd /Pich Att Hold Agorithm responses


Kq = 4, Ktheta = 1, Ki = .25, Kffi = 3.5, Kffp = 2.309
Pitch Rate, Pich Attitude, Theta Cmd ~deg.sec, deg

Tautheta = .5 sec, no actuator dynamics


6
Pitch Rate
Pitch Attitude
5
Pich Att Cmd

0
0 5 10 15 20 25
Time ~sec

Figure 7. Responses for Step Stick Command of PRC/PAH Algorithm, = .5 second

12
In fact, for the landing approach flight path might be more suitable for the landing approach
control task, it is advantageous to select a task. In a real application, selection of should
negative final , as is generally the case for an
un-augmented airplane. Figure 7 and 8 show the take value of 2 into consideration. 2
responses for =.5 which may be close to ideal represents the time constant of the first order lag
for a pitch pointing task, and for = -1.0 which response of flight path angle relative to pitch
attitude.

Pitch Rate Cmd / Pitch Att Hold Algorithm responses


Kq = 4, Ktheta = 1, Ki = .25, Kffi = 5, Kffp = 6.236
Pitch Rate, Pitch Attitude, Pich Att Cmd ~deg/sec, deg

Tautheta = -1, no actuator dynamics


6
Pitch Rate
5 Pitch Attitude
Pitch Att Cmd
4

-1
0 5 10 15 20 25
Time ~sec

Figure 8. Responses for Step Stick Command of PRC/PAH Algorithm, = -1 second

Pitch Attitude control system responses to step cmd = 1


system 1: Kq = 4, Ktheta = 1, Ki = .25, Kffi = 0, Kffp = 0
system 2: Kq = 3, Ktheta = 1, Ki = .33, Kffi = 0, Kffp = 0
1
0.9
Pich Att Cmd, Pitch Attitude ~deg

0.8
0.7

0.6

0.5
0.4

0.3

0.2 Pitch Att Cmd


0.1 Pitch Attitude system 1
Pich Attitude System 2
0
-5 0 5 10 15 20
Time ~sec

Figure 9. Pitch Attitude Responses to step Pitch Attitude Command for two sets of system feedback gains.

13
Another consideration for the selection of is be seen by comparing the responses of the
that too much pitch attitude lead (negative ) complete system with gains Kq = 3, K = 1, KI =
relative to KIstick /S causes an unacceptably ratio .333, KFFI = 2.5 and KFFP = 1.73, presented in
of peak pitch rate to steady state pitch rate and figure 10, to the responses shown in Figure 7. In
too much of a pitch attitude drop back after the both cases the design was tuned to achieve =.5
pilot centers the controller (Gibson Pitch Rate with fast response convergence and negligible
and Drop Back HQ criteria [10]. wobble.

The selected feedback gains for the responses in High feedback integrator gain, consistent with
figure 7 and 8 are very conservative and result in maintaining the desired system damping (here
a rather slow response of the basic pitch attitude 1), is desirable for good disturbance rejection
response with feed forward gains augmentation and preventing response droop due to trim
KFFI and KFFP set to zero. It is possible to select effects on a real airplane. It is clear that the basic
considerably faster basic pitch attitude dynamics, stability augmentation feedback gains can be
while maintaining all real negative poles and not selected within a wide range, while still being
exceed a pitch rate loop gain of 4. For example, able to achieve nearly the same overall pitch
the response for a system with triple poles at S = attitude response by proper selection of the feed
-1, requires gains Kq =3, K = 1, KI = .333. forward gains. This capability has great merit
Figure 9 shows that this system responds when integrating the augmented manual control
considerably faster to a step command than the and fully automatic control modes. In effect the
system with gains Kq = 4, K = 1, KI = .25. One feedback gains can be designed to serve first and
would expect that for the full stability and foremost the automatic modes and then the feed
control augmentation system the pitch attitude forward gains can be designed to provide the
dynamics could be made to converge faster to desired pitch responses for the augmented
the final K/S for a system with a faster basic manual mode. These principles were used in the
feedback system response. However, this effect TECS design to simplify automatic and manual
is hardly noticeable, except perhaps when the mode integration.
system is tuned for large negative .This may

Pitch Rate Cmd / Pitch Att Hold Control Algorithm


Kq = 3, Ktheta = 1, Ki = .333, Kffi = 2.5, Kffp = 1.73
Pitch Rate, Pich Attitude, Pitch Att cmd ~ded/sec, deg

Tautheta = .5, No actuator dynamics


6
Pitch Rate
5 Pich Attitude
Pich Att cmd
4

-1
0 5 10 15 20 25
Time ~sec

Figure 10. Response for PRC/PAH system with =.5 second

14
In the process discussed above, actuator damped poles in the system). In that case the
dynamics have been neglected. This is justified overall system robustness is maintained and the
on the premise that the actuator loop gains will effect of the added actuator dynamics will be
again be at least a factor 4 higher than the pitch minimal. Compare responses of Figure 11 for
rate loop gain, in order to assure that the actuator system including actuator dynamics to responses
dynamics will not couple with the pitch rate of figure 10 for system without actuator
control loop (and therefore not produce under- dynamics.

Pitch Rate Cmd / Pitch Att Hold control Algorithm responses


Kq = 3, Ktheta = 1, Ki = .333, Kffi = 2.5, Kffp = 1.73
Pitch Rate, Pitch Attitude, Pitch Att Cmd ~deg/sec, deg

Tautheta = .5 sec,
6
Pitch rate
5 Pich Attitude
Pitch Att Cmd
4

3
Actuator dynamics:
2
576/(S +60S+576)
2

-1
0 5 10 15 20 25
Time ~sec

Figure 11. PRC/PAH responses for system of figure 10, including actuator dynamics

Finally, the described wide range of pitch trained to use a pre-cognitive open loop control
attitude response tuning capability may be used strategy instead. For the same reasons this
to tune the algorithm to achieve the desired approach was also used for landing the
handling qualities for the flight path angle Concorde.
control task, e.g. for the landing approach. For
that case the may be selected to achieve a The frequency responses /stick for unity (full)
desired flight path angle response lag . For stick input, for selections of = 1, 0 and 1 are
example, if the basic flight path angle to pitch shown in figure 12. The bandwidth increases
2 with decreasing . For the configuration with
attitude response lag = 2 seconds, and the
= 0, 1800 phase lag is reached at ~ 7 rad/sec,
desired =1 second, then this requires = -1 yielding a gain margin of ~6 db (factor ~2).
second, which is the response shown in Figure 8. range. Here Kstick = 5 deg/sec/full stick. Thus,
linear system stability will be assured as long as
It is interesting to note in the application of an the pilot uses a gain KP < 2 units of stick
earlier PRC/PAH concept on the Space Shuttle, command / degree of pitch attitude error. This
2 large linear gain margin provides enough system
which has an inherently large 3 seconds,
robustness to be able to incur a considerable loss
the need to reduce to achieve good flight path of gain margin due to non-linearities in a real
control handling qualities was not fully airplane application. The stick sensitivity Kstick
understood, or perhaps the algorithm did not used here is typical for an approach
have the inherent capability to achieve lower . configuration. For higher speeds Kstick must be
As a consequence, the Space Shuttle is reduced, to assure suitable nz /stick. As should
notoriously difficult to land with a closed loop be expected, the configuration with = 0 has an
pilot control strategy. Therefore astronauts are

15
Bode Diagram
Pitch Rate Cms / Attitude Hold Algorithm
Kq = 4, Ktheta = 1, Ki = .25, Kstick = 5 deg/s ec/full stic k
Frequenc y Response Theta/stick~deg/full stic k
100

50
Tautheta =-1

Magnit ude ( dB)


0
Tautheta = 1
-50

-100 Actuator:
2 Tautheta = 0
-150 1024/(S +64S+1024)

-200
0

-90 Tautheta = -1
Phase (deg )

-180 Tautheta = 1

-270 Tautheta = 0

-360
-2 -1 0 1 2 3
10 10 10 10 10 10
Frequency (rad/sec)

Figure 12. Frequency responses for Pitch Rate Command/ Pitch Attitude hold Algorithm

amplitude response that is K/S-like over the considerably to the reduction in pilot workload,
widest frequency range. because in that case the system will
automatically compensate for the effects of
Flight Path Angle Rate Command / Flight winds, turbulence and airplane configuration
Path Angle Hold algorithm. For most flight changes. The concept requires the addition of
conditions the pilots manual control objective is flight path angle information on the Primary
to control the airplanes flight path angle, or path Flight Display. Furthermore, a FPARC/FPAH
in space, e.g. capture and maintain an altitude or algorithm greatly simplifies continuous descent
the glide slope. If the pilot has only pitch attitude maneuvers involving altitude crossings at
to control and observe on his display, then he specified navigation waypoints, because, when
must change pitch attitude iteratively to arrive at the airplane reliably maintains the pilot
a desired flight path angle. To control a path in established flight path angle, it is easy to reliably
space also involves iterating the flight path angle predict the altitude crossing point and display it
(or vertical speed) to capture and track the on the navigation display. The FPARC / FPAH
desired path in space. Thus the pilot uses pitch algorithm is an ideal control mode for use with
attitude control as an inner loop control to Head Up Display, which also uses flight path
achieve his outer loop control objective. angle.
Attitude inner loop control is essentially a
process of executing a series of educated Flight path angle relates directly to normal load
guesses, because the final pitch attitude is not factor (nz) and Angle of attack, so it is easy to
known and the control sensitivity gain - how implement normal load factor and stall
much pitch attitude change to use in each protection. Thus the FPARC/FPAH algorithm
iteration - varies with the flight condition. can functionally integrate many desirable design
Therefore the pilot could be helped a lot by and operations features, in a relatively simple
providing him with the capability to directly way, compared to many other design approaches.
control the flight path angle, without having to
use the pitch attitude as an inner loop control, The industry has been reluctant to move to flight
and by providing a design feature to maintain the path angle based control, because of feared
pilot established flight path angle when the pilot difficulties in achieving good handling qualities,
gets out of the control loop. A true flight path gaining pilot acceptance and being able to
angle hold feature will also contribute certify an airplane that uses FPA as the basic

16
control mode. The step toward introduction of based augmented manual control and pitch inner
flight path angle based automatic and augmented loop design.
manual control algorithms on commercial
transport is in fact a relatively small one, with The relationship between flight path angle and
few risks and very substantial rewards. pitch attitude is described by the transfer
function
The initial research to develop and evaluate the 1
merits of a FPARC/FPAH algorithm dates back
(4) = wherein
to the NASA TCV program in late seventies. S + 1
This author was involved in sorting out many of 2
the design implementation choices and ironing C .V
out most bugs of the early system. This work
(5) = L True
resulted in a stand-alone design [2]. Then, in the 2 g.CL
early eighties a functionally integrated flight
control systems was developed that uses a
generalized energy based control strategy to The basic structure of the FPARC/FPAH
provide all automatic and augmented manual algorithm shown in Figure 13 and described
modes for airplane control in the vertical plane. here, is very similar to the PRC/PAH algorithm
This concept called Total Energy Control (Figure 6.), described above. Instead of the
System (TECS) features a core controller, integral control feedback loop being closed on
based on energy related variables of flight path pitch attitude, it is closed on flight path angle,
angle and longitudinal, that serves all its and an additional proportional flight path angle
automatic and augmented manual control modes feedback path is introduced (switches in the
[1]. The FPARC/FPAH naturally fits in the System 1 position). The same pitch rate, pitch
TECS architecture and requires only a front-end attitude and feedback integrator gain KI are
pilot control input processing module to realize maintained as in the PRC/PAH algorithm. Now
this augmented manual control mode. This the output of the pilots inceptor command
research laid the groundwork for a more integrator in the forward signal path may be
complete understanding of the relationship considered a FPA-command (cmd).
between pitch attitude and flight path angle

Figure 13. FPARC/FPAH control Algorithm

The transfer function /stick becomes


K stick (K FFPS2 + K FFIS + 1)K q K K I
(6) =
stick S S(S2 + K qS + K q K )( 2 S + 1) + K q K K K IS + K q K K I
The denominator of this transfer function is a isolated by reworking the transfer function as
follows:
function of 2 , but this dependency can be

17
K stick (K FFPS2 + K FFIS + 1)K q K K I
(7) =
stick S S(S2 + K qS + K q K )( 2 S + 1) + K q K K I (K S + 1)
Now by making K = 2 and normalizing the transfer function we obtain

K stick (K FFPS2 + K FFIS + 1)


(8) =
stick S ( 2 S + 1){(1/K q K K I )S3 + (1/K K I )S2 + (1/K I )S + 1}

Notice that this /stick transfer function (8) for 2


For a condition with = 2, selecting = 1
the FPARC/FPAH algorithm differs from the
/stick transfer function (3) by the exact airplane and canceling the same pole at S = -.382 as in the

transfer function = 1 /( S + 1) . This
2
PRC/PAH algorithm (figure 8) requires KFFI = 5
and KFFP = 6.236 and results in, = -1.
means that the for the same 2 , feed
It is possible to select the feed forward gains
forward gains KFFP and KFFI and feedback such that one of the numerator zeros cancels the
gains, both algorithms produce identical pitch
attitude and FPA responses, keeping in mind that pole associated with 2 . Then the remaining
for the FPARC/FPAH algorithm relates to the zero can be selected to yield the desired steady
of the PRC/PAH algorithm via the state . As an example, for a condition with
relationship = - 2 . KFFP and KFFI can 2 = 2 and keeping the other feedback gains
be selected to satisfy two conditions. KFFI is the same as in the first example of the ARC/AH
determined by selecting a desired or . Then algorithm (Kq = 4, K =1, KI = .25), the result is
KFFP can be selected to cancel one of the poles.

K stick (K FFPS2 + K FFIS + 1)


(9) =
stick S (2S + 1)(S + 1)(2.618S + 1)(.382S + 1)

In the simple model of figure 13 with the overshoot, which is attributable to the KFFP gain
switches in the down position (system 1), determination method used for this case. This
calculations for KFFI and KFFP are shown based small FPA overshoot at maneuver termination
can be minimized, if desired, by a increasing
on selected values of the parameters 2 , KFFP to ~ 6.7, but a barely observable FPA
and KI, resulting in cancellation of the pole response wobble at the start and finish of the
maneuver remains. This is the result of the
associated with 2 . For demonstration
imperfect poles/zeros cancellation.
purposes, 2 has been made independently
It is possible to entirely eliminate the
selectable. To achieve a = 1 (and therefore a
= -1) requires KFFI =2+1+2.618+.382-1= 5 and
2 effect in the /stick transfer function.
KFFP = 6. The resulting /cmd and /cmd time This is accomplished by using a proportional
histories are shown in figure 14. Notice that the feedback of the flight path angle error, instead of
pitch attitude response in figure 14 is very a proportional feedback of the flight path angle
similar, but not identical , to the attitude response 2
signal and again selecting K = . This
in figure 8, obtained for the PRC/PAH system
with identical = -1. The slight difference is algorithm configuration is shown in figure 13
attributable to the different order of the /stick with the switches in the up position (System 2).
transfer function for the two cases. The flight In that case the /stick transfer function reduces
path angle response in figure 14 exhibits a small to

18
FPA RAte Cmd /FPA Hold algorithm responses
Kq = 4, Ktheta = 1, Ki = .25, Kffi = 5, Kffp = 6, Kstick = 5 deg/sec/full stick
6

FPA Cmd, FPA, Pich Attitude ~deg


5

1 Pich Attitude
FPA
FPA cmd
0
0 5 10 15 20 25
Time ~sec

Figure 14. FPARC/FPAH algorithm (system 1) responses to step 10 % of full stick for 5 seconds

K stick (K FFPS2 + K FFIS + 1)


(10) =
stick S (1/K q K K I )S3 + (1/K K I )S2 + (1/K I )S + 1

Note that this transfer function is the same as the =4, K = 1, KI = .25, and KFFP = 1. This case
/stick transfer function, equation (3). To achieve results in a first order FPA response, shown in
= 1 for this FPARC/FPAH algorithm, with Kq figure15.

FPA Rate Cmd / FPA Hold algorithm responses (System 2)


FPA Cms, FPA, Pitch Attitude ~deg, Pitch Rate ~deg/sec

Kq = 4, Ktheta = 1, Ki = .25, Kffi = 3, Kffp = 1, Kstick = 5 deg/sec/full stick


6

3 Pitch Attitude
FPA
FPA cmd
2 Pitch Rate

-1
0 5 10 15 20 25
Timw ~sec

Figure 15. FPARC/FPAH algorithm (System 2) responses using proportional FPA error feedback instead
of proportional FPA feedback

19
For the general case of an arbitrary the rules responses. Furthermore, to achieve harmony
presented above may be used to determine KFFP between the initial pitch acceleration and the
for an optimized response. The advantage of the final nz response (figure 1), KFFI (here
algorithm configuration of System 2 is that the controlling the steady state nz lag relative to K/S)
feed forward gains KFFI and KFFP are now only a must be selected within a narrow range, to
function of the desired and KI . KI is selected achieve the desired value of . For high speed
at the initial design and will be fixed thereafter. 2
conditions will be smaller and with the
Thus for the System 2 configuration, the gains
KFFI and KFFP can remain constant throughout the same this results in lower KFFP and lower CAP
flight envelope, if a constant is desired. Only values. So in its present form the CAP criterion
does not provide meaningful design guidance,
K needs to be adjusted in proportion to 2 . because the allowed range is too large and
It should be noted that with the system 2 because this range should be a function of
configuration (switches in figure 13 in the up 2 . This may be explained as follows. In [2,
position), there is a new sneak proportional
signal path from the stick to the control surface 10] it was found that a constant 1 provides
(here pitch acceleration command), through KFFI satisfactory FPA control for a transport aircraft at
and the proportional FPA-error signal path with all flight conditions. So, to achieve = 1 at high
gain K. When the gains of both proportional 2
speed where is relatively small, does not
stick signal paths are combined a total gain KFFP
2 require a large (lead of pitch attitude relative
= 7 is found (for =2), close to the
to stickKS/S). Therefore a lower initial pitch
optimum KFFP = 6.7 found for the FPARC/FPAH acceleration will suffice for the build up this lead
system 1. This sneak proportional signal path (if any is needed) within a certain time limit, to
provides the inherent adjustment of the gain of establish a desired steady state .nz . What is
the overall proportional feed forward stick missing is a criterion for the allowed time it
command signal for different 2 values.
takes to establish a steady state nz for a step
stick input. This then drives the requirement for
KFFP (and possibly , if = 1 was found to be
The CAP Handling Quality Parameter re- unsatisfactory.)This in turn would drive new
examined. As already noted above the criteria for KFFI and the normalized stability
FPARC/FPAH (system 1) and the PRC/PAH augmentation feedback gains Kq , K and KI.
algorithms can be tuned to produce identical
FPA and Pitch Attitude responses. In that case Flight Path Angle display. A very important
the transfer function /stick will be the same for aspect of the application of a FPARC/FPAH
both systems and therefore CAP will be the algorithm is the need to provide proper FPA
same for both systems. From equation (3) for the display information on the PFD or HUD and
PRC/PAH system, it follows that proper pilot training. The flight path angle lags
&&
t =0 = stick KS K FFP and for a constant stick the pitch attitude response by 2 seconds. On
input the final incremental load factor is
V V V approach, 2 may be as large as 2.5 seconds
n z = & = & = stick KS , therefore
t = g g g for a transport airplane. This is too large a lag to
produce acceptable handling qualities when the
gK FFP pilot attempts direct flight path control by
CAP = . Thus, for an approach
V closing his control loop using a pure FPA
display. To achieve satisfactory handling
condition with 2 = 2 and a selected = 1, it qualities for direct flight path angle control, it is
follows that the required value KFFP = 6.236, necessary to use a quickened FPA-symbol, as
resulting in CAP = .836. This is well within the was found during the original FPA based
CAP range between .28 to 3.6 allowed in MIL- augmented control algorithm development in the
STD-1757A for level 1 handling qualities. late seventies [2, 11], and also in NASA HUD
However, KFFP is a critical design parameter that studies [12]. The FPA command, developed at
cannot be varied much from the calculated value the output of the stick command integrator, can
in the above analysis and still achieve acceptable serve as

20
external disturbances, thereby providing
the quickened FPA signal, because it has perfect
considerably better flight path tracking
K/S characteristics and because, for the
performance. However, the price for this
algorithms presented above, will be constant improved path tracking is pitch bobbing, and
throughout the flight envelope. This results in high elevator and pitch activity in turbulence.
two display quickening possibilities: display Also, for such an inner loop it is difficult to
FPAcmd and FPA full time, or blend FPAcmd with achieve satisfactory damping and robustness,
the actual FPA using a first order washout filter because of sensitivity to non-minimum phase
on FPAcmd. For the latter approach only one normal acceleration response caused by tail lift,
symbol, defined as sensitivity to accelerometer location and
S structural mode pickup. An extensive research
(10) FPA Quickened = FPA Cmd + FPA program was conducted in the late seventies
S + 1
under the NASA TCV program to develop high
is used on the PFD or HUD. FPAquickened performance automatic landing flare algorithms
responds like FPAcmd when the pilot maneuvers, [13]. In this research frequency responses of
and returns to true FPA when the pilot centers path deviation due to horizontal and vertical
the stick and leaves the control loop. With this gusts were studied for two equivalent high gain
quickened FPA display the pilot can directly flare algorithms, one based on a pitch
control the airplane to his desired FPA target, rate/normal acceleration inner loop and the other
without the need to iterate through pitch attitude. derived from the same flare algorithm by
Another advantage of using FPAcmd display, replacing the vertical acceleration feedback by
either as a full time or blended display, is that its state variable constituents. It was found that
when the pilot closes his control loop on the algorithm using a pitch rate/normal
FPAQuickened, he cannot induce a PIO condition, acceleration inner loop has vastly superior
because the phase lag relative to his input does altitude tracking performance for frequencies up
not exceed 90 0. to ~ the short period frequency, relative to the
algorithm using a conventional pitch rate/pitch
During the original research, it was found that attitude inner loop. Above this frequency the
display of a pure K/S response is not ideal, pitch rate/pitch attitude based algorithm has
because of the pilots neuro-muscular and superior performance, both in flight path tracking
cognitive delays. Experiments showed that pilots and in elevator activity. Therefore, in order to
were able to hit the desired FPA target better if achieve the best overall flight path tracking and
the displayed symbol responded with a small touchdown dispersion, a new flare algorithm was
additional first order lag (e.g. display = ~.2 synthesized by using the pitch rate/normal
seconds), rather then as a pure K/S. However, acceleration based algorithm for control in the
this re-introduces an additional time lag in the low frequency range and the pitch rate/pitch
control loop that can affect PIO susceptibility. attitude based algorithm for control in the high
frequency range. The coupling was achieved by
Inner loop Feedback Signal Synthesis using either a first or second order
The above discussions considered only basic complementary filter, with the crossover
feedback control augmentation and pilot frequency equal to the above turbulence response
command response shaping to achieve good crossover frequency.
handling qualities. In a final design also control
tracking and control activity under external Here, a similar approach is proposed to achieve
disturbances must be considered, especially if the good overall performance, considering all
basic stability augmentation is used to serve as important design requirements. In the static
the inner loop for the automatic control modes. It inversion process sketched above, a pure angle
is a well known fact that a pitch rate/pitch of attack feedback signal, corrected for position
attitude inner loop makes the airplane stiff in error, could be used. Un-modeled sensor effects,
attitude, which is desirable for a pitch attitude turbulence noise and inadequate sensor reliability
hold function, but not so good for an automatic can make this approach unacceptable for a real
mode requiring tight flight path tracking, such as system design. A simple derived inertial angle of
in the case of glide slope, or landing flare control attack substitute: der = - inertial works better.
in turbulence and windshear. An inner loop using Another approach is to use the following
pitch rate and normal acceleration can provide complementary filtered angle of attack signal,
more pitch agility by sensing and counteracting using a position corrected sensor signal s for the

21
low frequency range and inertial angle of attack stick to elevator transfer function approaches a
rate signal & I for the high frequency range: straight gain and the elevator time histories
resemble the pilot controller time history. At low
K1S& I + (K2S + 1)S
(11) F = 2 , where speed and large 2 , the augmentation effect
K1S + K2S + 1 is very noticeable and the elevator must work in
a z + g.cos.cos a +g overdrive, to maintain the same FPA response
(12) & I = +q z +q as at high speed. When the FPARC/FPAH and
VI VI PRCPAH algorithms are tuned to yield the same
pitch attitude and FPA responses, the elevator
In (11) again der may be substituted for s. In responses are identical, as expected. The
(12) az is the output of an z-axes body mounted overdrive effect is less pronounced for the
accelerometer. The filter time constants must PRC/PAH algorithm, if the algorithm is tuned
then be optimized for best performance in for K/S-like pitch attitude response.
turbulence /windshear and for accelerometer bias
rejection. Control Responses at Constant Thrust.
Control responses at constant thrust differ
Algorithm Performance with Full Airplane considerably between the different algorithms.
Dynamics. The above algorithm analyses and For such conditions speed change has a
time responses used constant speed short period dominant effect on the control response after the
airplane dynamics approximations. The pilot leaves the control loop and to a lesser
performance of the above algorithms (C* , C*U, extend during short duration with the pilot in the
nzC/HdotH, PRC/PAH, FPARC/FPAH), was loop. Short term responses to a constant stick
evaluated by Matlab Simulink simulations using input appear similar to conditions with automatic
full airplane and actuator dynamics including speed control, but show an apparent washout
rate and position limits, with and without an over time of the maneuver rate (droop). When
automatic thrust control to maintain airspeed. the pilot leaves the control loop after establishing
The performance of these algorithms, with the a new flight path angle (vertical speed ), the C*U
full airplane dynamics and speed automatically algorithm returns to trim speed and this causes a
maintained to within a knot or two, differed little relatively fast washout of the flight path angle
from the analysis using simplified constant speed change achieved during the time with the stick
airplane dynamics without an actuator. The deflected, as well as a temporary nz reversal. The
FPARC/FPAH algorithm was also evaluated as C* has a slower flight path angle washout and nz
an integral part of the TECS design, with a static reversal , while energy is being transferred from
inversion based inner loop. The difference in airspeed to flight path to partially make up for
performance was small, except in the case where the thrust deficiency. Speed continues to bleed
a substantial thrust pitching moment was down or build up and this is the reason why a
simulated, but not compensated for in the speed envelope protection function is needed.
airplane static inversion. For this case the trim For the above PRC/AH algorithm the pitch
change due to thrust requires an adjustment of attitude starts to droop after stick release, but
KFFP. Alternatively, after including thrust there is a tendency to establish a new pitch
pitching moment compensation in the static attitude and airspeed equilibrium. The
inversion, the performance using full airplane FPARC/FPAH algorithm exhibits a flight path
dynamic model was again nearly identical to the angle washout after stick release similar to C*.
performance with the simplified constant speed
airplane model and no adjustment to KFFP was Augmented Manual Roll Control
required. With the inclusion of a realistic The generalized design approach used above for
actuator model, including rate and position the PRC/PAH algorithm development can serve
as a carbon copy for the development of a
limits, the effects of variation of 2 and generalized Roll Rate Command/Roll Attitude
dynamic pressure for different flight conditions Hold (RRC/RAH) control algorithm. Only the
becomes very noticeable in the control surface stick (or wheel) sensitivity changes to ~25-30
response. For high dynamic pressure and small degrees/ sec roll rate command for a full lateral
stick input. Here the roll mode time constant
2 conditions, the augmentation effects on is typically selected to be in the range of .5 to 1
the responses are small. In that case the overall second. The same rules apply for determining the

22
feed forward gains as used in the pitch axis providing the correct sensitivity around
control algorithms described above. This neutral. At high speed a full control inceptor
RRC/RAH algorithm was implemented and deflection should command the nz-limit,
evaluated as an integral part of THCS, using the while at low speed a full pull-up deflection
full lateral-directional airplane dynamics and the should command the available safe nz
static inversion process to design the roll and command authority, consistent with a pre-
yaw innerloops. The roll responses and turn selected angle of attack limit.
coordination performance were very good. The response shaping design should not use
differentiation of the control inceptor
PIO Assessment command signal, which will cause a gain
A complete PIO assessment of the above increase and phase advance at high control
augmented manual control algorithms is beyond inceptor input frequencies and increase the
the scope of this paper. It requires a high fidelity risk of actuator rate limiting and PIO.
model of all the control algorithm details, It is not possible to maintain constant stick
including actuator dynamics with rate and force/g sensitivity over the flight envelope,
position limits, a complete airplane dynamics using a passive feel system and a fixed
model and representation of the low frequency control inceptor deflection range. A ~
flexible airplane modes. A limited assessment constant stick force/g can be achieved by
was conducted, using the rigid airplane model. using an active/programmable feel system or
Some general comments can be made to help by a using a constant force gradient control
avoid unnecessary PIO susceptibility: inceptor with an actively varying deflection
Feedback integrators need to be protected hard stop. A constant stick force /g may not
from open loop wind-up due to actuator rate be needed, as long as the control inceptor
and/or position limiting, which causes maneuver command sensitivity around
additional phase lag in the feedback loop. In neutral and the full deflection authority are
a final design this can be remedied by appropriate. Increasing maneuver command
placing the feedback integrator at the end of sensitivity with control inceptor deflection
the control algorithm processing chain. This can cause pilot over control and increases
requires differentiation of the proportional PIO susceptibility.
signal paths, so these proportional signals The control algorithm should provide
can be fed through the integrator. This continuously and smoothly varying control
differentiation and integration processing surface commands as a function of the
should be designed to avoid noise control inceptor deflection. There should not
amplification. The output of this integrator be any abrupt gain, or control response
then represents the final control surface behavior changes. Any needed control
command which needs to be limited to the algorithm adaptation to the flight condition
actuator authority limit. The input to this should be programmed as a function of
integrator, should be amplitude limited to slowly changing variables, e.g. weight,
the actuator rate limit, so that the actuator dynamic pressure, speed and altitude.
can keep up with its command. Unnecessary processing delays due to digital
The control inceptor authority needs to be data bus signal input/output transmission,
scheduled in such a way that a full sensor and control algorithm processing
deflection control inceptor never commands should avoid. This requires an efficient IO
more than the available control surface handling and an efficient Real Time
authority. This is necessary to assure that the Executive scheduler.
pilot will never experience a dead zone at
the end of the control inceptor defection Conclusions
range for any flight condition. Since the Methodologies used in developing augmented
basic airplane control surface authority is manual control algorithms vary widely and the
sized to handle all foreseeable maneuver resulting designs are often very complex and
conditions, proper scheduling of the control difficult to analyze. Important characteristics of
inceptor authority should not in any way different existing algorithm types were
restrict the maneuver capability. This reviewed. This paper describes a simple,
objective can be achieved by scheduling transparent design process for pitch /roll attitude
control inceptor sensitivity as a function of and flight path angle type of control algorithms,
deflection and flight condition, while also consisting of three distinct elements:

23
augmentation of the basic airplane stability, pilot Using Total Energy Principles, AIAA Paper
command augmentation and design of special 83-2239CP.
provisions (e.g. to track the established control 2. Lambregts, AA., (1979), Development of a
reference). It was demonstrated that Control Wheel Steering Mode and Suitable
Application of static airplane model Displays that Reduce Pilot Workload and
inversion is an effective and practical Improve Efficiency and Safety of Operation
approach to synthesizing the basic airplane in the Terminal Area and in Windshear,
stability augmentation part of the design: AIAA Paper 89-1887.
o It simplifies the implementation of basic 3. Lambregts, A.A.,(1997), Automatic Flight
design objectives, such Short Period Controls Concepts And Methods,
natural frequency and damping and Netherlands Society of Aeronautical
provides transparency to the meaning Engineers Annual Report 1996.
of parameters used; 4. Anon, 1980, Flying Qualities of Piloted
o It provides inherent compensation for Airplanes, Military Specification, Mil-F-
variation in basic airplane dynamics, 8785C.
separated from the control algorithm; 5. Anon, 1995, Flying Qualities of Piloted
o It provides the basis for generalizing the Airplanes, Military Specification, Mil-STD-
control algorithm design, making it 1797A.
usable in a variety of airplanes, with 6. Mooij, H.A., (1984), Criteria for Low Speed
little or no adaptation for the specific Longitudinal Handling Qualities of
airplane and flight condition, while Airplanes with Closed Loop Flight Control
providing uniform performance Systems, NLR Report TN 83037 U.
throughout the flight envelope; 7. Tobie, H.N., Elliott, E.M., Malcolm, L.M.,
o The process should not decrease design (1966). A New Longitudinal Handling
robustness, as long as the same Qualities Criterion.
conservative design considerations are 8. Neal, T.P., Smith, Rogers E., (1970), An In-
applied as used in more classical design Flight Investigation to Develop Control
approaches; System Design Criteria for Fighter
The final airplane responses to pilot control Airplanes, AFFDL-TR-70-74.
can be shaped largely independently from 9. Mooij, H.A., van Gool, M.F.C., (1977). The
the basic stability augmentation part of the Need of Stick force Stability for Attitude-
design, by using appropriate feed forward Stabilized Aircraft. NLR Report TR 77027
signal paths; U.
Different control algorithms, specifically 10. Gibson, J.C., (1999), Development of a
PRC/PAH and FPARC/FPAH types, can be Methodology for Excellence in Handling
designed to produce identical maneuver Qualities Design for Fly By Wire Airplanes
control responses; Delft University Press, ISBN 90-407-1841-
The use of feedback integral control need 5.
not lead to lower airplane control 11. Lambregts, A.A., (1991). US Patent
bandwidth. Algorithms can be designed such 5,016,177 Aircraft Flight Path Angle
that the final order of the response transfer Display System.
function is the same as, or lower than for the 12. Brae, R.S., Scott, B.C., (1981), Head-Up
basic un-augmented airplane; Display Format for Application to Aircraft
Important algorithm distinctive features Aorcraft Approach and Landing, NASA
include: airplane response characteristics TM-81199; NASA HUD Report 11.
when the pilot leaves the control loop and; 13. Lambregts, A.A., Creedon, J.F.,(1980).
the effect of thrust/speed control on the Development and Flight Evaluation of
target control response. Automatic Flare Laws with Improved
Touchdown Dispersion, AIAA paper 80-
References: 1757.
1. Lambregts, A.A., (1983), Vertical Flight
Path and Speed Control Autopilot Design

24

Vous aimerez peut-être aussi