testing tunnel-in-the-sky displays and flight control systems with and without flight simulator...

Upload: joseph-mangan

Post on 08-Apr-2018

220 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/7/2019 Testing Tunnel-in-the-Sky Displays and Flight Control Systems With and Without Flight Simulator Motion ISAP_035

    1/6

    TESTING TUNNEL-IN-THE-SKY DISPLAYS AND FLIGHT CONTROL SYSTEMS

    WITH AND WITHOUT FLIGHT SIMULATOR MOTION

    Max Mulder1, Jerome Chiecchio2, Amy R. Pritchett2, M. M. (Ren) van Paassen1

    1 Delft University of Technology, Faculty of Aerospace Engineering, Control and Simulation division

    Delft, the Netherlands2 Georgia Tech, Schools of Industrial and Systems Engineering and Aerospace Engineering,

    Atlanta, USA

    Tunnel-in-the-sky displays have shown great potential for reduced pilot workload and for reduced navigation error.

    However, further improvements are s till worthy of investigation. One challenging consideration is the relationshipbetween modern flight control systems and the tunnel display. For example, flight control systems may now be

    developed that allow the pilot to directly control the direction of the aircraft motion relative to the ground,

    facilitating the tracking of a ground-based tunnel considerably. Another major c onsideration is the means by which

    tunnel displays may be accurately tested in a cost-effective manner. Nearly-identical studies of tunnel displays in

    flight tests and fixed-based simulator tests have achieved dramatically different results, suggesting that the task of

    flying a tunnel may be very sensitive to pilot-perceived motion. This study tests several combinations of the tunnel

    display with advanced flight control system concepts, and examines in particular the impact of flight simulatormotion on pilot workload, control behavior and ability to track the tunnel.

    Introduction

    Tunnel displays may enable aircraft to closely follow

    intricate trajectories as a means of improving air

    traffic management efficiency, meeting noise

    abatement conc erns, etc. (Grunwald, 1984, Mulder,1999). While studies to date have demonstrated the

    potential benefits of tunnel displays, further

    improvements are still worthy of investigation. One

    important consideration is the relationship between

    modern flight control systems and the tunnel display.

    That is, advanced fly-by-wire control algorithms may

    now be implemented that allow the pilot to directly

    command the direction of the aircraft motion relativeto the ground, facilitating pilot tracking of a ground-based tunnel considerably (Veldhuijzen, Mulder, Van

    Paassen, & Mulder, 2003).

    Another major concern is the means by

    which tunnel displays may be accurately tested in a

    cost-effective manner. Nearly-identical studies of

    tunnel displays in flight tests and simulator tests have

    led to remarkably different results. E.g., pilot path-

    following performance with a tunnel display in

    conducting a straight-in approach was a factor 2 to 3times worse when conducted in real flight as

    compared to performance in a fixed-base simulator

    (Mulder, Kraeger, & Soijer, 2002). This suggests thatthe task of flying a tunnel may be very sensitive to

    pilot-perceived motion. While extensive tests in the

    past have outlined the type of testing facility suitable

    for many piloting tasks, tunnel displays are

    sufficiently novel that appropriate methods of testinghave not been fully explored; these insights may also

    help relate flying the tunnel to other piloting tasks

    and further understand the relationship between pilot

    perception of motion and flight control behavior.

    The study presented in this paper tests thecombination of a basic tunnel-in-the-sky display with

    several advanced flight control system concepts, and

    focuses in particular on the impact of flight simulatormotion on pilot control behavior and ability to track

    the tunnel. The paper is structured as follows. First,

    the rationale behind the experiment is discussed in

    detail. It will be shown that the results presented in

    this paper are a sub-set of a much larger multi-

    objective experiment, the results of which are

    reported in other papers at this conference

    (Chiecchio, Pritchett, Kalaver, Van Paassen, &

    Mulder, 2003; Veldhuijzen et al. 2003). Then, the

    experiment design and setup are described, followedby a discussion of the experimental results.

    A Multi-Objective Experiment:

    Testing Tunnel Displays, Flight Control Systems,

    and Pilot Fault Detection Behavior

    In the summer of 2002, a group of researchers from

    the Georgia Institute of Technology (USA) visitedDelft University of Technology (the Netherlands).

    Their mission was to study the pilot detection of

    faults in the flight control system, using Delft

    Universitys high-fidelity flight s imulator, SIMONA.

    This investigation fitted nicely into an existing line ofresearch at Delft University on the development of

    fly-by-wire control algorithms suitable for being used

    in conjunction with the tunnel-in-the-sky display. Thevarious research perspectives from both parties led to

    the definition of what we called the multi-objectiveexperiment. Before elaborating further on this

    experiment, first the main initial research interests of

    the Atlanta and Delft parties will be outlined.

  • 8/7/2019 Testing Tunnel-in-the-Sky Displays and Flight Control Systems With and Without Flight Simulator Motion ISAP_035

    2/6

    Aiding pilot in detecting faults in the flight control

    loop

    Georgia Tech is involved in a NASA-funded study

    on the pilot detection of faults in the flight control

    loop. Problems with loss of control are a major

    safety issue. Any degradations in performance in the

    flight control loop impact the aircraft stability and the

    pilot ability for control. These faults can be caused

    by many systems, such as the control mechanisms by

    which the pilot enters commands, the actuation of theaircraft control surfaces, faults in the sensors and

    displays providing feedback to the pilot, etcetera.

    Compounding pilot difficulty in responding to these

    faults, they may be masked by flight control systems

    that attempt to compensate for the fault until they

    reach the limit of their control authority.

    Few systems currently exist for aiding pilotsin detecting these types of faults, in particular those

    not caused by simple mechanical failures. Several

    technologies may enable such pilot aids, includingon-board simulation and adaptive estimation of the

    expected aircraft dynamics in nominal conditions for

    comparison to actual conditions. However, the role

    and function of these aids remains to be determined.

    The design of a simple aural and visual alert facesconcerns with setting its threshold and commensurate

    potential for false alarms and mis-use (Pritchett,

    2001). Likewise, while a unified display of flight

    control loop health can provide the pilot beneficial

    information and predictive power, it may also be

    difficult to comprehend, especially when a flight

    control system is masking the impact of the fault, and

    may not by itself enable the pilot to understand howto control the aircraft in its damaged state.

    The aim of the Georgia Tech research team

    was to investigate the possibilities for providing pilot

    support in detecting faults in the flight control loop

    through alerts and a display of flight control loop

    health. To provide pilots with an environment that is

    as realistic as possible, the SIMONA research

    simulator was used. The pilot task targeted for this

    study was a relatively high workload manualtrajectory-following control task with flight control

    systems ranging from conventional direct-link to the

    more advanced maneuver-demand control concepts.

    During this task, pilot monitoring for faults is to beobserved, as well as pilot reactions to faults.

    Task-oriented control-display systems

    The ability of pilots to manually control their aircraftalong complex curved approaches using the tunnel

    display is widely recognized to be one of the main

    qualities of the display. By presenting the pilot the

    trajectory to follow in an intuitive three-dimensional

    fashion, while at the same time including the

    constraints of the trajectory-following guidance, the

    tunnel display is related very closely to the pilots

    primary aircraft guidance and control task.

    Since the early investigations of the tunnel

    display, however, it has been known that a basic

    tunnel does not always lead to acceptable levels ofpilot performance and workload (Grunwald, 1984).

    The mainstream of tunnel research has focused on

    augmenting the display with symbols that can help

    pilots in controlling the aircraft through the tunnel.

    Symbols like the flight-path vector, showing the

    instantaneous direction of the aircraft motion, and,

    most importantly, the flight-path predictor, showing

    the aircraft future position a couple of seconds ahead

    (usually in conjunction with a frame moving ahead inthe tunnel that acts as the reference for the predictor

    symbol), showed to greatly facilitate the pilots task(Grunwald, 1984). The presentation of especially the

    flight-path predictor, however, considerably changes

    the task and also a pilot control behavior, namelyfrom the integrated feedback of aircraft attitude,

    flight-path and position relative to the tunnel to atwo-dimensional pursuit tracking task with preview

    (Mulder, 1999). That is, the pilot is constantly trying

    to keep the flight-path predictor symbol in the middle

    of the predictor r eference frame moving ahead of the

    aircraft, rendering the tunnel geometry itself to be

    merely a situation awareness aid. Also, with these

    display-augmentation features, the pilot remains

    operating in the innermost control loop, constantlycompensating for the path-following errors caused by

    weather effects such as turbulence or wind. This can

    result in considerable levels of pilot workload and asub-optimal path-following performance.

    In (Veldhuijzen et al., 2003), a radically

    different perspective on improving the pilot-display

    interaction is presented. Here, the aim is not to

    augment the display through the addition of symbols,

    but rather to investigate the possibilities ofmanipulating the basic aircraft control. That is, with

    the introduction of fly-by-wire (FBW) flight control

    systems in modern aircraft, the relation between the

    pilots control inputs and the corresponding response

    of the aircraft can be designed. Veldhuijzen et al.

    (2003) compared the basic, direct-link pilot manual

    control with a tunnel display with two different FBWcontrol concepts: attitude-orientedcontrol and flight-

    path-oriented control. The attitude-oriented FBW iscomparable to the current way Airbus aircraft are

    being controlled. The pilot stick inputs act as attitude-

    rate (i.e. pitch-rate and roll-rate) commands to theFBW system. When the stick input is zero, the FBW

    controllers keep the selected attitude constant, even

    in the presence of disturbances acting on the vehicle.In the flight-path-oriented FBW solution the pilot

  • 8/7/2019 Testing Tunnel-in-the-Sky Displays and Flight Control Systems With and Without Flight Simulator Motion ISAP_035

    3/6

    stick inputs act as flight-path vector rate (i.e. track-

    angle rate and climb-angle rate) commands, or, inother words, the pilot commands the aircraft direction

    of motion relative to the ground. This control solution

    comes very close to the primary task of pilots when

    flying with the tunnel display, i.e. the accurate

    tracking of a trajectory fixed with respect to the

    ground. Optimizing the display as well as the controls

    from the perspective of the pilots primary guidance

    task (tracking ground-based tunnels in the sky) yields

    a so-called task-oriented control-display interface.

    Veldhuizen et al. (2003) show that the

    combination of the tunnel display with an advanced

    flight-path-oriented fly-by-wire control system

    significantly improves the pilot path-following

    performance and reduces the pilot workload

    considerably. This study was conducted in a fixed-

    base flight simulator. In the same period of this study,results of a flight test campaign showed remarkable

    differences in pilot path-following performance in

    tracking straight tunnel trajectories as compared toearlier experiments conducted in the same fixed-base

    flight simulator (Mulder et al., 2002). This raised the

    concern whether the pilot behavior and the

    corresponding metrics like performance, as

    obtained in the fixed-base simulator are indeedrepresentative for (or can be extrapolated to) pilot

    behavior in real flight. Hence, a follow-up study was

    planned in the SIMONA flight simulator, to examine

    the possible effects of having simulator motion on the

    results of the Veldhuijzen et al. (2003) study.

    Rationale of the multi-objective experiment

    An experiment has been designed in which theresearch interests of the Delft and Atlanta parties

    were served as good as possible. It was decided to

    invite the pilots for the Delft experiment in which

    the combination of tunnel displays and three flight

    control systems was investigated with and without

    simulator motion. When pilots arrived in the

    morning, they were quickly briefed and

    measurements started. Pilots were left completely

    unaware of the fact that, first of all, the objective was

    to study the effects of flight simulator motion, and

    also, even more important, that at the end of the

    morning trials, just before the lunch break, anintended and easily detectable fault occurred in the

    flight control loop. This allowed us to test the pilot

    agility and response to such a fault in a high

    workload situation, in the case they really did not

    expect a fault to happen at all.And surprised they were. During lunch, after

    the pilots were calmed down a little, they were

    briefed about the complete intentions of the

    experiment, that is, the Atlanta experiment. It was

    made explicit to them that during some of the

    afternoon sessions a variety of faults in the flight

    control loop could occur, and they were asked to

    detect when such a fault had happened and comment

    on the type of f ault (e.g. sensors, flight controls, etc.)

    they thought it was. They were also introduced in the

    operation of an alerting system, the so-called fault-o-meter, that was designed to help them in detecting

    and coping with the fault as it occurred. This alerting

    system always provided pilots with the right

    information at the right time. But again, pilots were

    tricked in the last run, as in that run the alerting

    system indicated that a fault had occurred when it did

    not. This last surprise was included to test the

    potential for over-reliance on an indication (and alert)

    of FCS health, as indicated by accepting a false alarmfrom the system as truth. After this second surprise,

    some pilots made it very clear that they would neveragain participate in these kinds of experiments.

    Design of the multi-objective experiment

    From the above it is clear that the whole experiment,which lasted one day for each pilot, can be regarded

    as the concatenation of four smaller experiments. In

    the firstexperiment, the emphasis was only on pilot

    behavior in flying complex trajectories with a tunnel

    display, with a var iety of fl ight control systems, with

    and without simulator motion. The second

    experiment is the first surprise trial, where the pilot

    is confronted with an unexpected fault. In the third

    experiment, the pilot fault detection capability is

    investigated. Finally, the fourth experiment consisted

    of the second surprise trial, where the alertingsystem presents a false alarm.

    This paper presents the results of the first

    experiment only. The results of the other experiments

    are all discussed in detail in (Chiecchio et al., 2003).

    Experiment

    The goal of the experiment was to investigate the

    impact of simulator motion on pilot tunnel tracking

    performance with various flight control systems.

    METHOD

    Subjects and instructions. Twelve professional pilots

    participated in the experiment. They were instructedto fly the aircraft through the tunnel as accurate as

    possible. They were briefed correctly about all

    objectives of the experiment, except one: that ourintention was to examine the effects of flight

    simulator motion. Rather, pilots were briefed with the

    statement that In this experiment we will be testingdifferent motion models (italics not included in the

  • 8/7/2019 Testing Tunnel-in-the-Sky Displays and Flight Control Systems With and Without Flight Simulator Motion ISAP_035

    4/6

    briefing). In the experimental trials, all cues with the

    exception of motion were present, such as theinitialization motion procedure of the simulator, even

    when motion was turned off during the run. This was

    done such that pilots could not predict if motion was

    going to be on or off in a particular run.

    Apparatus and setup. Subjects were seated in

    the cabin of the SIMONA flight simulator. The

    tunnel display was shown on a 15 inch LCD display

    in front of the subject. A green flight-path vectorsymbol was presented. To the side of the tunnel

    display a standard navigation display and engine

    indication display were shown. The aircraft was

    flown through the control column; the pilots had no

    rudder panel at their disposal.

    Independent measures. The experiment had two

    independent variables. First of all, three flight control

    systems (FCS) were defined: (1) the direct-link

    manual control, i.e. no control-augmentation at all;(2) the attitude-oriented fly-by-wire system; and (3)

    the flight-path-oriented fly-by-wire system. These

    flight control systems were identical to those tested in

    (Veldhuijzen et al, 2003). For the flight-path FCS, a

    second (yellow) flight-path vector symbol was

    presented on the tunnel display, representing thecommanded flight-path.

    The second experimental variable was the

    flight simulator motion: motion was either on or off.

    Experiment design. A factorial within-subjects

    design was employed, consisting of six conditions (3

    FCS systems x 2 motion states). All conditions wererun three times, yielding 18 measurement runs.

    Aircraft model. The aircraft flown was a small

    business jet, the Cessna Citation 500. A realistic non-

    linear dynamic model was used. The aircraft was

    equipped with a yaw-damper to improve lateral

    stability, and also with an auto-throttle that kept the

    velocity constant at 150 kts (TAS). This allowed

    subjects to fly the a ircraft using the control column

    (giving commands to the flight control system) only.

    Weather model. The weather model included a

    constant headwind, with magnitude 15 kts andheading 25 degrees off the runway centerline. A

    patchy turbulence was simulated with an intensity

    that resulted in a relatively strong disturbance.

    Procedure. The pilots had to fly curvedapproaches to a fictitious airport. Two tunnel

    trajectories were used, which were mirrored relative

    to the runway centerline (note that the wind vector

    was mirrored as well). The tunnel width was 45 [m].

    The tunnel trajectories were quite complex and wereonly possible to fly with the tunnel display, at

    reasonably high levels of workload.

    Each pilot flew a total of 24 runs, of which 6

    were training runs. The runs were blocked by flight

    control system, i.e. they were run as 3 sets of (2 + 6).

    During the training runs the motion was always on.

    Each run lasted approximately 5 minutes. After each

    run the pilot workload was assessed using the NASATask Load Index (TLX). After these 24 runs the first

    surprise run was flown, with the FCS fault, see

    (Chiecchio et al., 2003) for more details.

    Dependent measures. The dependent measures

    in this first experiment were: (1) the path-following

    performance, expressed in the aircraft position and

    flight-path angle errors (measured relative to the

    tunnel trajec tory); (2) pilot control activity, i.e. thecontrol column deflections and their derivatives; (3)

    aircraft-related variables such as the attitude-rates

    and the normal acceleration; and (4) the pilotsubjective workload (the NASA TLX ratings).

    Data processing. Of each run, the first 10 seconds

    were disregarded to allow the pilots some time to getinto the task. The data belonging to the last 500

    meters before the runway were also discarded, as

    here pilots were preparing for landing.

    Hypotheses. The main experimental hypothesis

    was that, independent of the three flight control

    systems used, the presence of motion stimuli helped

    pilots to better control their aircraft and therefore

    improve their performance. It was expected that thispositive effect of motion increased when the level of

    flight control automation decreased. That is, the

    advantage of having motion stimuli was hypothesized

    to be larger for the direct link FCS than for the

    advanced flight-path-oriented FCS. The reason for

    this hypothesis is that the motion stimuli are

    generally believed to be useful in particular for

    controlling the aircraft inner loops, such as attitudeand flight-path. With the attitude-oriented and

    flight-path-oriented flight control systems the

    attitude control loop is automated, and the relation

    between the pilot control command and the aircraft

    attitude response is less direct than in the situation offull manual control (the direct link).

    Experimental Results

    The experimental data were analyzed using anAnalysis of Variance (ANOVA), with pilot (12

    levels) as a random effect and fcs (3 levels) and

    motion (2 levels) as fixed effects. The main results

    of this analysis are summarized in Table 1.

  • 8/7/2019 Testing Tunnel-in-the-Sky Displays and Flight Control Systems With and Without Flight Simulator Motion ISAP_035

    5/6

    Table 1. Result s of a full-factorial ANOVA on the variables involving pilot control activity, path-following

    performance, aircraft-related variables, and workload. In this table **, *, and o represent significance levels of

    p 0.01, 0.01 < p 0.05, and 0.05 < p 0.10, respectively. The other symbols are explained in the text.

    pilot control activity path-following performance aircraft-related variables workload

    e a de/dt da/dt Xe Ve Te e e p Q Nz zTLX

    main effects

    FCS ** ** ** ** ** ** ** ** ** ** ** ** **Motion ** o ** ** o O o

    two-way effects

    FCS x motion ** o

    (a) (b) (c)Figure 1. The means and 95% confidence limits of the STD of the pilot elevator control derivative de/dt (a), the

    RMS of the total position errorTe (b) and the normalized pilot workload ratings zTLX (c) (data of all pilots). The

    three flight control systems are shown along the horizontal axis (1=direct link, 2=attitude-oriented, 3= flight-

    path oriented), with the results grouped in clusters of simulator motion (off = 0, on = 1).

    Control activity. The main effects fcs and motion

    are both significant for the pilot vertical stick

    deflections e (fcs: F2,22=71.204, p

  • 8/7/2019 Testing Tunnel-in-the-Sky Displays and Flight Control Systems With and Without Flight Simulator Motion ISAP_035

    6/6

    link FCS, and the other FCS in between. Simulator

    motion resulted in a significant increase (p