This action might not be possible to undo. Are you sure you want to continue?
4, July– August 2001
Assessment of Flight Simulator Fidelity in Multiaxis Tasks Including Visual Cue Quality
R. A. Hess¤ and W. Siwakosit† University of California, Davis, California 95616-5294
A technique for analytical assessment of ight simulator delity is presented as an extension of a methodology previously introduced in the literature. The assessment is based on a computer simulation of the pilot and vehicle and is inherently task dependent. A simple model of visual cue quality is introduced that is based on the classical concept of human operator visual remnant. The complete assessment procedure now includes proprioceptive, vestibular, and visual cue modeling. Inverse dynamic analysis is employed that allows the use of compensatory models of the human pilot in multiaxis tasks. The methodology is exercised by considering a simple rotorcraft lateral and vertical repositioning task in which visual and motion cue quality is varied.
[A; B; = state-space quadruple de ning closed-loop C; D] pilot/vehicle system [AG ; B G ; = state-space quadruple de ning inverse C G ; DG ] dynamics lter G di = number of times expression for i th pilot/vehicle response variable must be differentiated for tracking command to appear explicitly d var = parameter controlling multiplicative noise in visual cue model E; F = matrices involved in de nition of inverse dynamics system e ¡¿0 s = time delay in pilot model G = transfer function matrix de ning inverse dynamics lter GD = transfer function matrix describing pilot/vehicle response to tracking command inputs with inverse dynamic system h = rotorcraft height in repositioning task, ft h com = rotorcraft height command, ft he = rotorcraft heigth error, h c ¡ h, ft Km = gain on structural model vestibular signal P P = state-space quadruple [A, B, C, D] qi = polynomial whose reciprocal describes inverse dynamics between i th pilot/vehicle response variable and i th tracking command ts = sampling period in visual cue model V = vector of input tracking commands in inverse dynamics system Ye = gain and low-frequency integration on structural model visual error signal YFS = force-feel system dynamics in structural model of pilot YNM = neuromuscular system dynamics in structural model of pilot
YPF y ycom ye ±A ±C Á
= proprioceptive feedback dynamics in structural model of pilot = rotorcraft lateral displacement in repositioning task, ft = rotorcraft lateral displacement command, ft = rotorcraft lateral displacement error, yc ¡ y, ft = lateral cyclic input measured at pilot’s hand, in. = collective input measured at pilot’s hand, in. = vehicle roll attitude, rad
Received 5 January 2001; presented as Paper 2001-4250 at the AIAA Modeling, Simulation, and Technology Conference, Montreal, Canada, 6 – 9 August 2001; revision received 10 April 2001; accepted for publication c 11 April 2001. Copyright ° 2001 by R. A. Hess and W. Siwakosit. Published by the American Institute of Aeronautics and Astronautics, Inc., with permission. ¤ Professor and Vice-Chairman, Department of Mechanical and Aeronautical Engineering, One Shields Avenue; firstname.lastname@example.org. Associate Fellow AIAA. † Graduate Student, Department of Mechanical and Aeronautical Engineering; currently Lecturer, Department of Mechanical Engineering, Kasetsart University, Bankok 10900, Thailand. 607
LIGHT simulation facilities constitute an important tool for both pilot training and ight control research. The delity of the simulator is of prime importance in the successful use of these facilities. Here, simulation delity can be de ned as “the degree to which characteristicsof perceivablestates induceadequatepilotpsychomotorand cognitivebehaviorfor a given task and environment.”1 Assessment of simulator delity is usually undertakenthrough subjective pilot delity ratings and comments, for example, Refs. 2 and 3. However, analytical approaches to delity assessment are also needed. The research summarized here builds on the analytical framework introduced and developed in Refs. 4 – 6. In particular, a mechanism for including visual cue quality is introduced with origins in the classical treatment of human operator remnant, for example, Ref. 7. Furthermore, well-developed compensatorytracking models of the human pilot behavior are utilized with the aid of an inverse dynamic analysis technique similar to that employed in Ref. 6, but now rendered computationally ef cient. The use of compensatory tracking models is not a minor issue. These models and their parameterization have been the subject of considerable research over the past four decades. With such models the analyst can predict the basic form of pilot equalization given the dynamics of the vehicle and a hypothesized manual feedback topology.7;8 The primary drawback to the use of compensatorymodels arises in computer simulation of discrete tracking tasks such as those characterizing many low-altitude rotorcraft maneuvers. In such computer simulations, unrealistic control activity and vehicle angular velocities are typically predicted at maneuver initiation. This has led some researchers to propose dual-mode pilot models in which pilot dynamics are different at the beginning of a maneuver than at its conclusion.9 Although such models possess a certain amount of face validity,their parameterizationis dif cult, especially if they are to be used in predictive fashion. The procedure employed in Ref. 6 allowed the use of compensatory models by creating outer-loop position commands that produced desired pilot/vehicle trajectories in the tasks being studied. A technique for formalizing and automating this procedure is
as such. the coef cients of qi . the delay increase may not be associated with the visual process. 3.18 whereas noise corruption is capable of increasing tracking error. In the model of Fig. per se. The state-spacequadruple for the closed-loop system can be given as P D [A. 6. (2). high-frequency poles. The process just described has Fig. the (decoupled) transfer function between pilot/vehicle response yi .di ¡ 1/ B. The visual cue model is shown separately to distinguish it from the transfer function representationof the pilot.di ¡ j/ C ¢ ¢ ¢ C aid i (5) (4) The model for visual cue quality is based on the classical concept of human operator remnant and was developed with an eye toward ef cient computer simulation of the resulting pilot/vehicle system. 10.s/ is given by yi . B.s/ is selected by the analyst. . low-order transfer functions.s/ where qi . The lter matrix denoted G produces a decoupled system between the tracking commands and the pilot/vehicle system responses. the i th tracking command must be multiplied by qi . BG . the displayederror.s/ D diag[1=qi . 11. it was felt that a baseline delay increase associated with the visual process when using degraded displays would be a reasonable modeling assumption. With this in mind. R C G D ¡E¡1 F.608 HESS AND SIWAKOSIT been automated in a MATLAB ® M. the two operations considered to be essential to a visual cue model were 1) sampling and 2) noise corruption. is fed through a sample and hold element. The unstable poles appearing in G are replaced by their mirror images with respect to the imaginary axis. C. Thus. it was desired to minimize model parameterizationwhile still allowing the resulting structure to produce effective remnant spectral characteristics that approximated those resulting from classical human-in-the-looptracking experiments. a state-space quadruple [AG . 1 Inverse dynamics analysis. To obtain the nal diagonal transfer function matrix with elements approximating unity. Reference 12 surveyedpertinentliteratureregardingthe effects of variation in display quality on measured human operator describing functions in laboratory tracking tasks.s/]vi . summarized in the following section. the dif culties associated with applying dynamic inversion to systems with nonminimum phase dynamics pose no problems here.s/ and tracking command vi . whose realization as a transfer function matrix can be obtained from the state-space quadruple of Eq. R BG D BE¡1 R DG D E¡1 R (2) where E¡1 is the right inverse of the matrix E.s/ with the nal elements having been made proper by the addition of an appropriate number of stable. By the use of a technique introduced in Ref.s/ D [1=qi . with a particular emphasis on multiaxis tasks. 11. Figure 2 compares a single-axistracking task employing the injected-errorremnant with the tracking structure using a serial visual cue model. Fi D jD0 ai j Ci A. Ref. CG .pilot/vehicle system has been obtained. 3. Citing six studies (Refs.s/ (6) The mechanism that provides the tracking/response relations such as that in Eq.di ¡ 1/ C ¢ ¢ ¢ C ai j s . As a nal note. An example of doing this for a simple single-axistask is given in Ref. Thus. Periodic sampling (with a zero-order hold) is capable of producing an increase in effective time delay approximately equal to half the sampling interval. The transfer function matrix between commands and responses can be represented by a diagonal matrix whose elements are simple.s/ to the vector expression for tracking commands V.s/ can be given by Y.s/ D ai0 s di C ai1 s .di ¡ j/ (3) where di is the number of times one has to differentiate the expression for the i th vehicle response before a tracking command explicitly appears. with the sampling period and the polynomial qi . The resulting phase errors in the elements of G are minimized in the frequency range of interest for manual control (0. Matrices E and F are R de ned as di Ei D ai0 Ci A. The vector expression for the vehicle responses Y. that is. represents the interface between the display medium and the resulting stimuli to the human pilot. for example. The relatively simple visual cue model that includedthese operationsand that was capable of closely approximating measured remnant spectral characteristics is shown in Fig. However. D G ] of the lter matrix G can be obtained as A G D A ¡ BE¡1 . The model itself is intended to capture the effects of visual information display in a control theoretic context and.s/]V. Inverse Dynamic Analysis Assume that a state-space model of the closed-loop. 2 Comparison of injected-remnant tracking structure with visual cue model structure. that is. 13– 17). D] (1) Figure 1 shows the structure of the inverse dynamic system to be discussed. for example. denoted as model input. once the value of di is obtained for each tracking command/response pair. and that produced effects in human operator dynamic characteristicssimilar to those obtained in experiment. it was shown that reduction in display quality of various types resulted in 1) an increase in the effective time delay in the human operator describing function and 2) an increase in tracking error in the task(s) at hand.le.s/ can be selected.1– 10 rad/s) by time delays on the desired inputs. (6) is the lter transfer function matrix G. with appropriate pilot compensation in outer loops. This would be created using compensatory tracking models of the pilot in inner loops. Modeling Visual Cue Quality Model Fig. Reference 12 demonstrated that the increase in time delay could be attributed to a gain reduction in the propriopceptive feedback used by the human operator.
vehicle dynamics = 1/s2 . Figure 4 shows a portion of the visual model input and output time histories in the computer simulation of the pilot/vehicle system with 1/s vehicle dynamics. Acceptable matches are seen to result. 11. Note that the magnitudes of the PSDs are calculated as 10log10 j ¢ j. Fig. The parameter dvar is thus seen to play a role similar to the noise– signal ratio of injected observationnoise in algorithmic models of the human operator such as the optimal control model. vehicle dynamics = 1/s: Fig. 7 PSD of effective remnant from visual cue model for different values of dvar. Fig. Comparison with Classical Remnant Models A computer simulation of a compensatory tracking task was created with the human operator represented by the structural model of the human pilot. The effective remnant was created as the random signal that reproducedthe output of the visual cue model when added to the displayed error. 6 Comparison of the PSD of effective remnant from visual cue model with that from experiment. As Fig. t s D 2¼ =2:50 D 0:0628 s ¼ 0:06 s (7) The variance dvar was then selected as that which scaled the resulting effective remnant power spectral density (PSD) so that acceptable matches to measured remnant PSDs reported in the literature could be obtained. Figure 7 shows the effect of varying dvar on the PSD of the effective remnant. Note speci cally that the PSD of the effective remnant scales with the mean square value of the displayed error just as the case with the experimental results of Ref. 11 for the two-vehicle dynamics. the resulting discrete signal is multiplied by a magnitude-limitedrandom number. The resulting ltered signal is added to the output of the sample and hold element to produce the model output.6 Two sets of vehicle dynamics were chosen.HESS AND SIWAKOSIT 609 ts. 3 Visual cue model. The visual model sampling period ts was chosen to provide a sampling frequency approximately 50 times the closed-loop pilot/vehicle bandwidth. 2 indicates. which is approximately 2 rad/s. The tracking command was a random signal. 1/s and 1/s 2 .8 Fig. and is then ltered. with variance dvar . Figures 5 and 6 compare the PSDs of these effective remnant signals with those from Ref. 4 Visual model input– output time histories. . Fig. 5 Comparison of PSD of effective remnant from visual cue model with that from experiment. Thus.
i refers to the i th axis of control.707. 20.707. Example Nominal Vehicle and Pilot Models At this juncture.25(0)/(0. 6. In both area calculations just dened. They have been categorizedas large-motionand smallmotion systems as described in Ref. delity metric D 1 n n iD1 1area normalizing area (9) i where n is the total number of inner-loop axes being controlled by the pilot. 21. 10 in more detail. simpli ed version of the hover dynamics of an unaugmented BO-105 rotorcraft as taken from Ref. Fig. Hence.610 HESS AND SIWAKOSIT Table 1 Axis Vertical Lateral Roll a Motion system de nitions Small motion 0:13. Table 1 gives the dynamics of the two motion systems to be compared in the analysis.0:707/.0. and 1 area refers to the area between the HQSF curve for the nominal or ight vehicle and that for the vehicle as being simulated. j !/j (8) Fig.0:3/2 ].81) Large motion 0:8.The vehicle model is a linear. a simple example of the assessment technique describedin the precedingsectionsis in order. a delity metric can be de ned using the differences in areas beneath the HQSF curves for a nominal vehicle and the vehicle as simulated across all inner-loop axes being controlled by the pilot. The larger the value of the metric de ned by Eq.5] Equal to 0:8s 2 =[s 2 C 2. 22. and normalizing area refers to the area beneath the HQSF for the nominal or ight vehicle. The vestibular feedback implicit in the structural model of Fig.19 As discussedin Ref.0. The ight task is shown in Fig. 10 Pilot/vehicle system for the task of Fig.0. 10 deserves some comment.0/2 /[0.707.5] 0:4.707. Fidelity Assessment with the Structural Pilot Model Analytical assessment or prediction of simulator delity is undertaken in exactly the same manner as discussed in Ref. the poorer is the predicted simulator delity. whereas the small-motion system is representative of hexapod motion devices. (9).0/2 /[0. the pilot model parameters are based on the nominal vehicle and left unchanged when the simulated vehicle is analyzed via computer simulation.0/2 /[0.0. 8. 11 Pilot models of Fig. Fig. Figure 11 shows the pilot model blocks of Fig. A MATLAB-based computer-aided design package referred to as PVDNL is utilized in creating the pilot model and in calculating the HQSF that is plotted vs ! on linear scales. The vehicle dynamics are P w P v p P P Á 0 0 0 0 0 0 ¡0:036 ¡1:76 ¡0:063 ¡9:91 0 1 0 32:2 0 0 w v p Á D Fig. 6.0/2 /[0. The handling qualities sensitivity function (HQSF) is the magnitude of a transfer function derived from the structural model of the human pilot that is shown in single-axisform in Fig. For the sake of clarity. the visual cue model and motion system dynamics have been omitted. 8 derives from the study in Ref.707. C ¡11:1 0:0086 ¡0:059 0:807 ¡0:13 2:8 0 0 ±C ±A The amount of coupling between vertical and lateral axes is quite small in this model and is created by off-axis control inputs.9] 0:45.0. There it was .0:3/s C . The inclusion of Ásensed in the inner-feedback roll-attitude loop in Fig. The large-motion system is comparable to the NASA Ames Research Center vertical motion simulator (VMS). The HQSF is de ned as HQSF D jU M = C. 9 Hover lateral and vertical repositioning task. Figure 10 is a block diagram of the pilot/vehicle system to be modeled. 9 and represents a 100-ft lateral and vertical repositioningtask initiated and completed in a condition of stabilized hover. 10.0/2 /[0.9] 0. 9. 8 Structural model of the human pilot.3]a 0:5.
707. In the outer visual loops. Figures 12 and 13 show the lateralcyclicandcollectivecommands from the pilot models when the commands of Eq. the same value of dvar was used in all of the visual models.t /.06 0.829 10/(0)(10) 0.11 0. be directly explained by this model. The conditions represented in Table 3 involve variation in motion dynamics (large vs small-motion systems and motion time delays). whereas only a single variable was involved in the computer simulation of the single-axis task that yielded Figs.1=32:2/ y (10) where Ásensed is an output from the visual model.47 0. (10). visual cue models were employed in the error channels for each variable created by the outer-loop compensation. Áe and h e . demonstrated that inclusion of a feedback loop involving the rst derivative of the inner-loop vehicle response variable being controlled by the pilot could explain measured human pilot dynamic characteristicswhen motion cues were present.035 0.that is.25 on rate variables. The four-toone variation in gain values in Eq.045 0.1 0.29 Unstable 0. The heave or vertical response was nearly identical to that shown in Fig. 10. The command trajectoriesfor the lateraland verticalrepositioning task were identical and were given by ydes . The jagged appearance of the HQSFs for the cases when the visual model is in place is a result of the noise injectionand limiting that occurs within the visual model.06 0.05 0.95(0.1/0.24 0.25 0.1/0. and temporal mismatches between motion and visual cues (unequal del m and del v values).25 Unstable 0. Note that the parameter dvar is a factor of 10 smaller than the value used in obtaining the PSD plots of Figs.10] 22.1 0. s 0. respectively.05 0. the impact of this assumption on delity assessment is mitigated by the gain values remaining invariant in the analysis to follow. 14.25a 0.25 0. Ác = ye D Y p y 0.06 0.47 0.1 0. 17 shows the normalizing area for the roll-axis structural model HQSF . P h c = h e D Y ph 1. Also note that.06 0.06 0.05 0. 0. This implies that in creating Ác .2)/(10) Heave axis. The fth and sixth columns represent time delays that may accrue in the motion del m and visual del v systems. P separate visual cue models were employed for ye . The inner.05 0. s 0 0. However.4)/(0) 625/[0. Selection of dvar D 0:1 was somewhat arbitrary and based on obtaining stable simulations with realistic pilot control inputs.0 Here 0. ¢ ¼ ¢ t =10/ ¡ 9 cos.30 0. for simplicity and Table 3 shows 13 con gurations representing conditions that could lead to reduction of simulator delity.035 0.05 0.05 0. the pilot employs an internally generated signal proP portional to ye .05 0.05 0.075 0.05 0. The utility of outerloop motion quantities such as lateral acceleration could not. 3.t / D h des . thus reducing the visual workload associated with the roll-axis task.06 0.06 del m.13 Outer-loop Lateral axis. In simulating the pilot/vehicle system. 10.125 N/A del v. Varying Simulation Parameters The visual model parameters for the nominal or ight vehicle are given in the third and fourth columns of the rst row of Table 3.025 0. visual cue quality (dvar).25] 3. that is.HESS AND SIWAKOSIT 611 Table 3 Con guration de nitions dvar 0.¼ ¢ t =10/ C 8] ft (11) P P with the exception of translational rate cues h e and ye in con gurations 5 and 6. The pilot outer-loop compensation elements Y p y and Y pÁ are also given in Table 2 along with the dynamics of the force-feel systems in the cyclic and collective cockpit inceptors. This surrogatecue was blended with the visually sensed roll attitude as follows: R Ásensed D 0:75Ávisual C 0:25.21 0. is of course. The reason for using this smaller value is that ve visually sensed variables are involved in P P the rotorcraft simulation.707.1 0.06 0.075 0. ye . (11) were applied to the system of Fig.1 0. (9).t / and ye .025 0.1 on position variables.05 0.28 0. For example.05 Fidelity metric 0 0.05 0.045 0.1 0. ye . Here it is hypothesized that lateral accelerationcan serve as a surrogate cue for roll attitude. that is. Fig.1 0.and outer-loop crossover frequencies for each axis were selected as 2 and 1 rad/s.06 0.06 0.53 0.06 0. as indicated in Fig.28 Unstable Table 2 Elements Pilot model parameters Values Inner-loop Con guration Nominal 1 2 3 4 5 6 7 8 9 10 11 12 13 a Motion Perfect Large Small Large Small Large Small Large Large Large Large Large Large Static Common ¿0 YNM Roll axis Ye YFS Km P Heave axis Ye TFS Km P 0. Áe . Figures 15 and 16 show the HQSFs from the roll and heave innerloop structural pilot models. however. conjecture. the fundamental waveforms appear as one or two period sinusoids corrupted by noise.1 0. The HQSFs for cases with and without the visual model are shown in each Figs.125 0. 5 and 6. Figure 14 shows the vehicle lateral response. 15 and 16 for the sake of comparison. Visual cue models were employed in the input channel for each P structural model inner-visual loop of Fig.05 0.1 ts.06 0. h e and h e .t / D 100 [cos.t /.1 0. These time historiesare qualitativelysimilar to those obtainedin ight test or pilot-in-the-loopsimulation.06 0.1 0.25 0.238(0.06 0. 5 and 6. consider the compensation for the outer-loop element of the lateral axis given in Table 2 where an (s C 0:2) factor appears in numerator of the Ác = ye transfer function.3 16 Fig. The structural model parameters resulting from application of PVDNL for the inner loops controlling vehicle roll and heave are given in Table 2. As an exampleof the area calculationscalled out in Eq.1 0.t / C 0:2 ye .2 100/[0. 12 Lateral cyclic control input from computer simulation with nominal pilot/vehicle system. s 0 0.
Indeed. The nal column of Table 3 tabulates the value of the delity metric de ned by Eq. the computer simulation of the pilot/vehicle system was unstable in three cases. As Table 3 indicates. Fig. a comparison of the delity metrics for con gurations 1 and 2 and 3 and 4 would suggest that motion system characteristicsdominate visual system characteristics. large-motion systems will exhibit higher delity than small-motion systems in spite of differences in visual cue quality (Fig. Fig. This does not imply that the actual piloted simulation would be unstable. 20 is quite small. 14 Lateral response time history from computer simulation with nominal pilot/vehicle system. 17 Normalizing area for roll-axis structural model HQSF.612 HESS AND SIWAKOSIT Fig. 18 shows the 1 area for the heave-axis HQSFs between the nominal heave con guration and that of con guration 1 in Table 3. indicating poor delity. 19) and that in cases of temporal mismatches between motion and visual cues. These comparisons predict that. Fig. 15 HQSFs for roll-axis structural model. Also note in Table 3 that the static or no-motion condition was unstable. Recall that the delity assessment procedure freezes the pilot model parameters at values appropriate to the nominal or ight vehicle (Table 2) and then applies this model in a computer simulation of the system with simulator limitations. The results just summarized offer few surprises. Finally. the visual should lead the motion (Fig. Fig. the con gurations of Table 3 were deliberately chosen to allow the . 13 Collective control input from computer simulation with nominal pilot/vehicle system. 16 HQSFs for heave-axis structural model. for this task. Figures 19 and 20 graphically compare the delity metrics for 12 of the 13 con gurations of Table 3. 20). (9). and the differencesnoted are probably within the model’s margin of error. and Fig. The contraindicationof this last conclusionfor con gurations 7 and 8 in Fig.
. B. 5 Hess. compensatory models of human pilot behavior. and Dynamics. Control. Vol. pp. 2000. a means of correlating the value of the delity metric of Eq. “Evaluation of Simulation Motion Fidelity Criteria in the Vertical and Directional Axes.. Committee Tasks. Vol. Nov. 24.. 41. Paper A-84-08-4000.. 4. 1979. 8 Hess. and visual cues. 19 Comparison of delity metrics for con gurations 1– 6 and nominal con guration of Table 3. and Atencio. for example. The Grant Technical Manager was Duc Tran. Finally. W.” Journal of Guidance. 2. 6 Zeyada.” Ad Hoc Advisory Subcommittee on Avionics. and Human Factors. “Mathematical Models of Human Pilot Behavior. T. 18 D area for heave-axis HQSFs between nominal heave con guration and that of con guration 1 in Table 3. Conclusions A methodology for the analytical assessment of ight simulator delity previously introduced in the literature has been extended to include a measure of visual cue quality. vestibular. Clement. W.. A. 37. 38. The methodology thus includes the effects of proprioceptive.” American Helicopter Society. .. 191– 197. and Krendel. Ref. D. 1997. A. Reference 23 indicated that visual cues should be synchronous with motion cues. D. No. 7 McRuer. S. and two are immediately apparent. A. Vol. First. Fidelity assessment is based on a metric derived from pilot-model-based HQSF. and Malsbury. No. 1. for example Refs. 1993. Malsbury.. 1996. A. Wiley. although some progress has been made in this area in terms of the optimal control model of the human pilot. In addition. Chap.” Journal of Aircraft. Fig. A. Fig. 16. 1991.. Ref. now extended to multiaxis tasks.. W. S.” Journal of Aircraft.. “Assessment of Simulation Fidelity Using Measurements of Piloting Technique in Flight. R. 20 Comparison of delity metrics for con gurations 7 – 12 and nominal con guration of Table 3. Model De ciencies The model for assessing ight simulator delity discussed herein is not without de ciencies. “Feedback Control Models– Manual Control and Tracking. Acknowledgements This research was supported by Cooperative Research Agreement NCC2-5369 with NASA Ames Research Center. and Dynamics. NASA. (9) with subjective measures of simulator delity is needed. Salvendy. a means of correlating a given visual scene representation in a simulator to a speci c set of dvar or ts values is yet to be established. 2nd ed. “Evaluation of Motion Fidelity Criterion with Visual Scene Changes. 37. Controls.HESS AND SIWAKOSIT 613 analytical approach to quantify results that have been experimentally veri ed. List of De nitions.. 1 References “Proposed Study of Simulation Validation/Fidelity for NASA Simulators. No. edited by G. 44 – 57. pp. 2000. “Human Pilot Cue Utilization with Applications to Simulator Fidelity Assessment. A. and Hess. control theoretic representations of human pilot behavior. if not all. R. May 1984. Second. A. pp. The latter de ciency is less formidable and can be attacked by comparing values of the delity metric with subjective measures of simulation delity in analyzing the results of pilot-in-the-loopsimulations reported in the literature. 4. New York. pp. pp. E. and Hess. albeit with vehicles and tasks differing in detail from those studied here. 2 Schroeder. Y. Vol. R. R. A. R. “Closed-Loop Assessment of Flight Simulator Fidelity. F. Fig. and Study Scope. 9 Ferguson. the visual cues should lead the motion cues.” Journal of the American Helicopter Society. Vol. 580– 587. 1.. Chung. No.. The former de ciency is shared with most.. 3 Schroeder. J. W. W.. Y. “Flight Simulator Fidelity Assessment in a Rotorcraft Lateral Translation Maneuver.. No. Cleveland. 79 – 85. J. 14..” AGARDograph 188. inverse dynamics analysis allows the use of well-established.. 1974. Exercise of the methodology on a simple example of a rotorcraftlateral and vertical repositioningmaneuver indicated that delity assessments could be derived that were consonant with experimental results obtained in pilot-in-the-loopsimulation studies.” Journal of Guidance. T. Reference 21 demonstrated the superiority of large-motion systems as compared to small or no-motion (static) systems. and Key. T.. Control. 3 demonstrated that motion system dynamics had a considerable stronger effect upon delity ratings than did variations in visual scene characteristics. 21 and 23. Jr. L. 4 Hess.. 588– 597. or at worst.” Handbook of Human Factors and Ergonomics.
W. Digital Control of Dynamic Systems. and Teichgraber. 21 Schroeder. “PVD .614 HESS AND SIWAKOSIT 10 Snell. A. H. K. Control. 3. Powell. AMRL-TR-71-63. Vol. No. 476– 482. 21. Lehman. E. and Johnson. J. 535– 556.” IEEE Transactions on Systems. CA. Aug.” Proceedings of the Fifteenth Annual Conference on Manual Control. Addison Wesley Longman. MI. T. “Model for Human Use of Motion Cues in Vehicular Control. “A Compilation and Analysis of Helicopter Handling Qualities Data. May 1971. and Levison. “Display Format Effects on Precision Tracking Performance. Describing Functions. Elkind. and Workman. Univ.” NASA CR-1746. R. Tran. R. pp. Y.. Wright– Patterson AFB. E. R. 1990. Sampling.. Aug. D.” NASA CR-1569.. W.” Journal of Guidance. 1970. 1998.. Jewell. Aug. 1998. A. 18 Franklin. 2nd ed. I. Laforce. OH.” IEEE Transactions on Systems. Man.. W. of Mechanical and Aeronautical Engineering. 23. A. “Effects of Vehicle Bandwidth and Visual Spatial-Frequency on Simulation Cueing Synchronization Requirements. 338– 344. and Dynamics. 11 Jex. S.. L. H... of Michigan. Vol. No. 1979. and Ward. R. Air Force Flight Dynamics Lab. 1974.. A.. 343– 349. W.” IEEE Transactions on Systems. 1978. and Magdaleno.” AIAA Paper 97-3655. D. 647– 655. No. and Cybernetics. 25. 1995. “Pilot-Induced Oscillation Prediction with Three Levels of Simulation Motion Displacement. No. Chung. 15 Hess.. pp. Y. J. Univ. Man. W. “Decoupling Control Designs with Applications to Flight. “Research on Display Scanning. R. Vol. Control. 22 Hess.. M. and Remnant.. 13 Allen. AFFDL-TR-79-3134. and Van Winkle..” Proceedings of the Fourteenth Annual Conference on Manual Control. W. R. 17 Kenyon. . Los Angeles. 23 Chung. J. and Cybernetics.. Univ. and Bengford. and Kneller.. F. 14 Levison.” Journal of Guidance. MA... J.” Dept. of California. Schroeder. 187– 211. J. A. “A Theoretical and Experimental Analysis of the Outside World Perception Process. 1990.. p. 12 Hess.. R... Vol. and Cybernetics.. Davis..” AIAA Paper 98-4333. “A Performance Analyzer for Identifying Changes in Human Operator Tracking Strategies. An Interactive Computer Program for Modeling the Human Pilot In SingleAxis Linear and Nonlinear Tracking Tasks... G. 4. 24 Wewerinke. F.. P. W. 16 Zacharias. F. 183– 193. No.. S. 1993. 1. H. CP-2000. “Error Quantization Effects in Compensatory Tracking Tasks. N. Man. R. pp.” NASA CR-3144. Vol. and Reconstruction Using Separate Main and Secondary Tracking Tasks. “The Effects of Field-of-View Size on the Control of Roll Motion.. Clement. Vol. R.. R. J. V. 119. of Southern California. J. pp. H. NASA. and Jex. W. Ann Arbor.” Aerospace Medical Research Lab. A. 19 Zeyada. 2.... W. W. 1971. “Modeling the Effects of Display Quality upon Human Pilot Dynamics and Perceived Handling Qualities. L. 4. A. and Dynamics. H. R.. A. pp. W. and Hess. Allen. March 1979. G. Y.. W. pp. W. Reading. pp.. W. “Studies of Multivariable Manual Control Systems: A Model for Task Interference. 20 Hef ey. Pilot/Vehicle Dynamics NL Nonlinear .. SMC-4.. M. 1998. R. One: Data Compilation. 13.. 1997. M L.
This action might not be possible to undo. Are you sure you want to continue?
We've moved you to where you read on your other device.
Get the full title to continue listening from where you left off, or restart the preview.