Objective Motion Fidelity Qualification in Flight Training Simulators

AIAA 2007-6802

AIAA Modeling and Simulation Technologies Conference and Exhibit 20 - 23 August 2007, Hilton Head, South Carolina

Objective Motion Fidelity Qualification in Flight Training Simulators Sunjoo K. Advani 1 International Development of Technology Ruud J.A.W. Hosman 2 AMS Consult and Mario Potter 3 International Development of Technology

This paper proposes a criterion that defines the motion cueing performance of a flight simulator. The long-term aim of such a criterion is to use it to qualify and compare training and engineering simulators. The approach is based on measuring the total gain and phase of the motion cueing system components, which include the motion cueing algorithm, motion system mechanism and transport delay, over a range of frequencies relevant to the type of flight vehicle being simulated. The frequency response of the total motion cueing system is then plotted against the criterion in a modulus – phase distortion plot (Nichols diagram) indicating the end-to-end performance of the system in gain and phase. The performance of four research simulator motion systems was measured and plotted into this format in order to illustrate this approach. The physical properties of these devices are well known and documented. The paper describes the initial experimental results using the proposed approach and the advantages and limitations. This work is aimed to broaden the understanding of motion cueing, to develop objective guidelines for simulator qualification and, in the long run, to simplify the current technical tests.

Nomenclature φs θs φa θa axs ayx axa aya

= = = = = = = =

Simulator roll rotation Simulator pitch rotation Aircraft roll rotation Aircraft pitch rotation Simulator specific force in x-direction Simulator specific force in y-direction Aircraft specific force in x-direction Aircraft specific force in y-direction

I. Introduction

I

n order to maintain high quality training in aviation, flight simulation is a highly regulated industry. When a qualified simulator is applied to pilot training, it may be used instead of the real aircraft, thereby making training effective and efficient. This requires that the behaviour acquired and reinforced by the pilot in the simulator translates as closely as possible to actual flight. While almost all aircraft characteristics are accurately modeled in simulation and matched within the simulator, motion cueing is the most difficult to match and compare. This is due to the inability of the ground-based simulator to reproduce flight motions one-to-one. So far, civil flight simulator regulations, in the area of motion systems, are restricted to the quantification of only the mechanical properties of

1

President of IDT, Breda, The Netherlands, [email protected], Senior Member AIAA Director of AMS Consult, Delfgauw, The Netherlands, [email protected], Member AIAA 3 Project Engineer, IDT, [email protected] 2

1 American Institute of Aeronautics and Astronautics Copyright © 2007 by S.K. Advani and R.J.A.W. Hosman. Published by the American Institute of Aeronautics and Astronautics, Inc., with permission.

the motion platform, and to matching aircraft motion vibrations. Measuring the mechanical properties only defines what the device is capable of doing, and not what it actually does during the training process. The International Civil Aviation Organization (ICAO) publishes a master document called “The Manual for the Qualification and Testing of Flight Simulators” (document ICAO 96251), considered the master standards document. At the time of writing this paper, ICAO 9625 Amendment 2 was undergoing a major revision. This revision is performed under the umbrella of the Royal Aeronautical Society, and the collaboration is known as the International Working Group (IWG). The main authors are active contributors to the IWG. The resulting Amendment of ICAO 9625 (No. 3) will take on a new format. Instead of simply stating the technical requirements of full flight simulators (as in prior such documents), the international team of pilots, airline operators, simulator and airframe manufacturers, regulators and external experts have mapped training requirements to Flight Simulator Training Device technical features. Starting with the various types of training (for example, from MultiCrew-Pilot Licensing, to Take-off Recency training), the tasks are then defined. Tasks, in this context, refer to the entire set of activities occurring during normal, abnormal and emergency operations. For example, the start engines, or taxi, or perform initial climb and flap retraction. With the training levels and tasks in place, the training device system’s “features” and associated fidelity requirements are produced. For example, the performing of a circling approach in a multi-crew cockpit environment requires a very wide field-of-view of the visual display system, while the same task in another type of training may not be required. The next step is to define how accurately the simulation must be for these features, within four categories (Specific, Representative, Generic or “Not Required”), where each level, respectively, would require a lower fidelity of the validation of the aircraft data. With all the above combined, seven general types of “flight simulator training devices or FSTD’s” have evolved so far. The highest two of these are “full flight simulators” as we have known them to date. The five remaining FSTD’s are lower level devices, similar to fixed-base “flight training devices”. Not surprisingly, the subject of motion cueing has led to the most challenging of debates within the ICAO International Working Group, owing to the fact that there exists much subjectivity in this area. While motion has been shown to directly influence pilot skill-based control behaviour, which consequently is translated to the flight environment 2, 3, 4 , studies on quasi transfer of training have been inconclusive and suggest that motion cueing, when treated independently, does not transfer 5, 6, 7. With a lack of solid criteria, an objective means of measuring the data, and a clear data presentation format, the discussion on whether or not motion is required continues. Furthermore, there exists a significant difference in the motion cueing algorithms and tuning coefficients between otherwise identical simulators of identical aircraft types, since tuning is performed subjectively. Given that one of the major shortcomings in all regulatory requirements is the absence of any motion cueing criterion, the authors initiated an effort to propose such a standard. The initial proposal has described by AdvaniHosman 8, 9. While the process requires full validation in training simulators, this paper describes the execution and assessment of this criterion on four research simulators at the following institutes: • • • •

Japan Aerospace Exploration Agency (JAXA), Tokyo, Japan University of Toronto Institute for Aerospace Studies (UTIAS), Toronto, Canada Central Aerohydrodynamic Institute (TsAGI), Zhukovsky, Russia National Aerospace Laboratory (NLR), Amsterdam, The Netherlands

II. Background Motion feedback influences the way in which flight vehicles are controlled. The presence of motion cues in simulation influences the perception and control of the simulated flight vehicle, and also the overall acceptance of the simulator by its operator. In simulation, ideally, aircraft specific forces and angular accelerations should yield the similar responses in the simulator. Motion cueing has, however, remained a “creative science” due to the substantial subjective input in their tuning. The motion tuning process attempts to trade off providing of the linear and angular 2 American Institute of Aeronautics and Astronautics

aircraft motion representations while remaining within the constraints of the motion system’s operating envelope (position, velocity and acceleration). Whereas most of the process of designing and implementing motion platforms is based on engineering, the motion cueing algorithm is one which transforms aircraft motions to simulator motions in an attempt to stimulate the pilot to perceive he is flying. There are also several types of motion drive algorithms, including linear classical washout, adaptive washout, optimal control and hybrids of these. The design of these algorithms is based on knowledge at the time on what is considered of importance to the simulation task, and the available workspace of the motion base. In the commercial world of flight simulation, manufacturers tend to develop their own adaptations to one or more of the above, consider this as a part of their motion platform, and it is not common to share this knowledge. However, it determines for a great deal the end result – the impact that the entire motion cueing system has on the pilot – that matters to the effectiveness of flight simulation in training.

III. Criterion Requirements In order to qualify and regulate motion cueing performance of flight simulators through an objective criterion, the technical approach must also consider several practical aspects in the design of the procedure and its application as a legitimate standard. 1.

2. 3. 4. 5. 6.

If an objective criterion were to be established, it must enhance current practices and not attempt to completely replace them. It should be seen as a measuring standard to help the industry in those areas that need improvement, or where objectivity is lacking. The criterion must be well designed and substantially supported by analytical and experimental knowledge. The criterion must be easy to interpret, and the interpretation of the results should not require in-depth knowledge of the technical details on which it is based. The criterion should be independent of the technology within the simulator and allow maximum flexibility to the manufacturers in finding appropriate solutions to motion cueing. The measurement and analysis techniques and tools must be standardized and readily obtainable. The criterion must be acceptable to all major stakeholders, including regulators, operators and simulator manufacturers.

The entire process of flight simulation and its application is tightly qualified against technical standards, with the exception of the motion cueing algorithm. Flight simulator response to motion inputs is dependent on the entire process, from control input, to host computer computations, and the motion cueing system. In this context, the motion cueing system is comprised of both the following systems: a. Motion cueing algorithm. b. Motion platform hardware and controllers, and time delays. In simulators, pilots experience the combined effect over a limited frequency range of motion cueing algorithm and motion platform dynamics. The range of frequencies depends on motion platform degree-of-freedom and aircraft dynamics. The inevitable presence of transport delays in the process also impacts the perception of the motion cues by the pilot. Hence, it is the total sum of the motion cueing algorithm and the motion platform dynamics that need to be determined. Currently, the transport delay and the motion platform dynamics (frequency response, smoothness, leg balance) and vibrations are qualified. The total sum of these is however never specified or objectively checked. Advani and Hosman 8, 9 proposed a motion cueing criterion whereby the modulus and phase of the cueing system (motion cueing algorithm, motion platform transfer function and fixed transport delay) are bounded by tolerance limits over a certain frequency range, and the gain and phase are plotted against one another. When this is done for the characteristic and for the parasitic (cross-coupled) motions, it is possible to generally compare motion cueing systems with each other (without revealing the dynamic characteristics of the components of the cueing system, which are often proprietary to the manufacturer).

3 American Institute of Aeronautics and Astronautics

The background behind this approach is to characterize the entire transformation between the pilot input and simulator output. Hence, from the point-of-view of motion stimulation and pilot perception, the following transfer relations have been defined as being of direct importance: 1. 2. 3. 4.

Simulator rotational response to aircraft pure rotations. Simulator specific force response to aircraft pure translations. Simulator rotational response to aircraft pure translations. Simulator specific force response to aircraft pure rotations.

The first two relations are of direct importance for the reproduction of aircraft motions. They require a relatively high gain and correlation with the aircraft motions, and a small phase distortion. The second two relations, despite being undesirable, are inevitable in ground-based flight simulation. They provide information about the parasitic (undesirable) simulator motion response and, preferably, manifest with low gains. The fact that the proposed criterion is based on the final transformation and transfer of the motion cues provides freedom to the designer of the motion cueing system how to comply with the criterion.

IV. Criterion The criterion proposed by Advani and Hosman can be stated as follows: 1. In the frequencies range relevant to manual control of the aircraft being simulated, the minimum gain (or modulus) and the maximum phase of the total motion cueing system transfer function should occur within clear and accepted bounds. 2. Similarly, the parasitic cross-coupling (simulator rotations due to aircraft translations and vice versa) should be within clear and accepted bounds. These can also be defined in terms of their gain and phase.

V. Procedure The proposed criterion can easily be applied to simulators whose motion transfer functions are known, or where the motion transfer functions can be accurately measured. Additionally, the transport delay of the motion cueing system should be known. This is defined as the induced digital delay due to the cueing algorithm, the actuator extension transformation and the digital controller of the motion system, see Fig. 1. The motion cueing algorithm transfer function is currently not measured for regulatory purposes. Motion bandwidth tests that record the mechanical system transfer function are also customarily limited to the heave alone. Conversely, this criterion requires measuring all six degrees-of-freedom and capturing therein the total transfer function of the cueing algorithm, motion system and time delays. The characteristics of the motion platform may be a fitted transfer function to the measured frequency response of the motion system, or it may be the measured frequency response. The essence is to capture both the drive algorithm in its tuned, operation form, and the mechanical system dynamics. Aircraft translations and rotation

Flight Model (in Host Computer)

Simulator platform commands

Motion Cueing Algorithm Motion Drive Algorithm Transfer Function

Simulator motion response

Actuator Extension Transformation

Simulator Motion Platform

Transformation delay and motion platform characteristics Transport delay of motion cueing

Figure 1 - Components of the motion cueing process


A. Quantification of the Motion Cueing Algorithm The motion cueing algorithm is a combination of high and low-pass filters, transformation matrices, and non-linear blocks such as tilt rate and acceleration limiting. It may also contain special effects that generate motions not recorded in flight data. It aims to provide a good balance between cueing at low frequencies representing the gentle response of the aircraft to manual control by the pilot, and limitation of the simulator cab to within the motion envelope when it is subject to abrupt accelerations. For the qualification of motion cueing, the primary interest lies in the representation of the aircraft motions including pilot’s control, external disturbances and aircraft aero-elastic response to control and external disturbance. Pilot’s control inputs during normal maneuvering often occur up to the maximum aircraft rotational rates of approximately 3 to 5 degrees per second. In this region, the non-linear aspects of the algorithm should not even come into effect. For example, for a twin-engine turboprop commuter, the short period natural frequency ranges up to around 3 rad/s, depending upon the configuration. Control inputs in this area are of greatest interest in providing motion feedback from a control point-of-view. As proposed in Advani-Hosman [Ref 8 & 9], the boundaries of the criterion should cover the frequency range up to 2.5 times the highest natural frequency of the simulated aircraft dynamics or its class (ICAO code A through F). The procedure for analysis of the motion platform characteristics is laid out as follows: 1.

Determine the transfer functions of the motion cueing algorithm. This should be for a linearized form of the algorithm if performed analytically. If measured, the transfer function should be determined based on control inputs over the frequency range of the aircraft and motion platform dominant natural frequencies. Preferably, this algorithm’s parameters will have been tuned to a specific aircraft. Tuning can be done in “whatever” way (subjectively or objectively). The following transfer functions should be calculated: Main Products a. Simulator roll rotation φs as a function of aircraft roll rotation φa. b. Simulator pitch rotation θs as a function of aircraft pitch rotation θa. c. Simulator yaw rotation as a function of aircraft yaw rotation d. Simulator specific force in x-direction fxs as a function of aircraft specific force in x-direction fxa. e. Simulator specific force in y-direction fys as a function of aircraft specific force in y-direction fya. f. Simulator specific force in z-direction fzs as a function of aircraft specific force in z-direction fza Cross Products a. Aircraft fxa to simulator pitch rotation θs. b. Aircraft fya to simulator roll rotation φs. c. Aircraft pitch rotation θa to simulator specific force fxs. d. Aircraft roll rotation φa to simulator specific force fys. The total effects of the low-pass and high-pass filters on the final simulator motion should be considered. For example, fxs is generated by the high-pass translation filter and by the low-pass tilt coordination and, hence, the combined effect of these filters should be quantified by gain and phase. See Refs 1 and 2.

2.

Define the transport delay of the motion loop (Fig. 1), meaning from the aircraft model output to the actual response of the motion platform. This should be a nominal, constant value. In order to do this, the digital delay from the instant of input to the motion cueing algorithm to the response of the motion platform needs to be measured.

3.

Define the mechanical motion platform transfer function for all six degrees-of-freedom or the measured frequency response at enough discrete frequencies over the required frequency range. This should be valid for the system as it is actually used (hence, the correct mass and moment of inertia). The transfer function should include the motion controller and the motion platform.


4.

Determine the total transfer function of the motion cueing system by multiplication of the above components (motion cueing algorithm, motion platform transfer function and motion platform transport delay) to determine the total motion platform transfer function.

5.

Calculate this for the linear and angular motions, as well as for the linear and angular cross-products. Note that measurements of the complete system can also be made, allowing bypassing of intermediate steps.

6.

Plot the results as Bode plots and gain versus phase distortion as a function of frequency. These should be plotted for the Main Products and the Cross Products listed above.

7.

Convert these Bode plots to the Modulus – Phase distortion (Nichols) plots together with the bounds of the criterion.

B. Presentation Format The clearest means of presenting these results is through Nichols plots, which map modulus in the horizontal axis against phase distortion (absolute value of the phase angle) in the vertical axis. In addition to plotting over a frequency range, it is important to show the characteristic frequencies of the aircraft. This is dependent on the degree-of-freedom, and the characteristic natural frequency of the airplane in that degree-of-freedom. It is recommended that the modulus and phase be plotted for frequencies ranging from 0.1 rad/s to 2.5 times the highest aircraft natural frequency (often due to the short-period mode).

VI. Evaluation on Research Simulators This criterion has been assessed on four research simulators. These simulators are used for research where accuracy of motion cueing is critical. They are operated by staff that is familiar with the details of manual pilot control and motion cueing. A. UTIAS Flight Research Simulator The flight research simulator of the University of Toronto Institute for Aerospace Studies (UTIAS), shown in Figure 2, has a generic transport aircraft cockpit configuration. The simulator is driven by a CAE six degrees-of-freedom hydraulic motion platform with 30-inch stroke actuators. It uses a wide-angle collimating window-type display system. Research applications of this simulator include designing and tuning motion cueing algorithms. It is capable of accelerations of approximately 10 m/s2 in fore-aft, lateral, and vertical directions and approximately 300 deg/s2 in pitch, roll and yaw. The system employs hydrostatic bearings for smooth operation and it has a bandwidth of approximately 10 Hz. The complete motion characteristics of the FRS are described by Grant10. B. JAXA Flight Research Simulator The FSCAT-A research simulator of the Japan Aerospace Exploration Agency (JAXA) is used for human factors research in critical flying tasks. It is driven by a six degrees-of-freedom hydraulic motion platform. The motion cueing algorithm uses linear classical washout filter with hybrid filter combinations. Tuning has been based on empirical iterations. Improvement of motion performance will take place in the near future as there are some known issues due to latency and tuning. C. TsAGI Flight Research Simulator The FS-102 flight research simulator of the Central Aerohydrodynamic Institute (TsAGI) in Zhukovsky, Russia, is used for flight handling qualities research. It has a large 60 inch 6 degree-of-freedom hydraulic motion platform which is considered to be well tuned. D. NLR GRACE Flight Research Simulator The GRACE flight research simulator of the National Aerospace Laboratory (NLR) in Amsterdam, The Netherlands, is a generic, re-configurable transport aircraft simulator. Applications include research on pilot performance during aircraft handling. It has a six degree-of-freedom electro-mechanical motion platform using an asymmetric layout, and window-type displays.


Figure 2 - UTIAS Flight Research Simulator

Figure 3 - JAXA FSCAT-A Flight Research Simulator

Figure 5 – NLR GRACE Flight Research Simulator

Figure 4 – TaAGI FS-102 Flight Research Simulator

VII. Results The criterion proposed by the authors as presented in this document has been tested against four research simulators, each. The results, shown in Figures 6 through 14, do not present an opinion on the quality or performance of these systems, other than their objective data as brought forward by this criterion. In these figures, the boundaries on the lower right of each plot identify the proposed target in which the curve, or a significant portion thereof for the range of relevant and aircraft-dependent frequencies, should fall. Figure 6 shows the Nichols diagram of the Motion Drive Algorithm (MDA) for two of these simulators, NLR and UTIAS. Both these simulators have relatively small motion systems, however the motion-base of the NLR simulator has been optimized to provide enhanced pitch 11. Two cases are shown for the UTIAS simulator, prior to tuning, and 7 American Institute of Aeronautics and Astronautics

following tuning (which indeed shows lower phase and higher gain. Its Classical Washout algorithm was evaluated by flying the simulation of a 747 in a low-altitude clean cruise configuration. It was adjusted subjectively. The NLR algorithm was tuned for a generic large transport application. It demonstrates similar performance characteristics. Note that, at time of publication, only the motion cueing algorithm data of the NLR simulator was available. In Figures 7 through 12, the total response of the motion drive algorithm motion platform hardware and time delay for each degree-of-freedom, are presented for three simulators – UTIAS, TsAGI and JAXA. In general, the bandwidth of all of these systems is known to be very acceptable and at least as good as simulators used in training. It was specifically recorded for this experiment in the case of JAXA. However, when the MDA and transport delay components were added, its gain and phase shifted considerably. This is again prior to tuning, which is now under way. The TsAGI simulator has a motion envelope similar to JAXA, and considerably larger than UTIAS. It has been well tuned and thereby demonstrates an overall higher gain. As seen in Figures 7 (pitch) and 8 (roll), application of the criterion gives a corresponding indication. In Figure 9, the total transfer is shown for yaw. The motion system of the JAXA simulator shows an overall higher gain and lower phase than in other degrees-of-freedom. It now also becomes clear that bounding the frequency range (particularly the upper limit) to a level that is related to the aircraft natural frequencies allows reasonable fitting of the data. Figures 10, 11 and 12 show the total transfer in the translation degrees-of freedom. In surge and sway, the TsAGI and UTIAS simulators show reasonable performance in the mid-to high frequency range, yet the smaller size of the UTIAS motion base appears to limit the maximum allowable gain. Heave, however, is quite different for all cases and it is worthwhile exploring allowable limits here. The cross-couplings of translation inputs leading to simulator rotation outputs due to tilt coordination (pitch due to surge acceleration inputs), or tilt compensation (roll due to sway), are somewhat unavoidable. These are shown in Figure 13. The reverse cross-coupling (translations caused by rotations) occur at very low amplitudes, albeit with significant phase error. See Figure 14.

Pitch - Motion Drive Algorithms 120

UTIAS Pitch MDA (well tuned) UTIAS Pitch MDA (poorly tuned) NLR Pitch MDA 0.63 rad/s 1 rad/s 3 rad/s

Phase [deg]

100 80 60 40 20 0 0

0.1

0.2

0.3

0.4

0.5 Gain

0.6

0.7

0.8

0.9

1

Figure 4 – Motion Drive Algorithm gain and phase in pitch for UTIAS (two configurations) and NLR simulator


Pitch 120

UTIAS JAXA TsAGI 0.63 rad/s 1 rad/s 3 rad/s

Phase [deg]

100 80 60 40 20 0 0

0.1

0.2

0.3

0.4

0.5 Gain

0.6

0.7

0.8

0.9

1

Figure 5 – Total response of UTIAS, JAXA and TsAGI simulators in pitch

Roll 120


Phase [deg]

100 80 60 40 20 0 0

0.1

0.2

0.3

0.4

0.5 Gain

0.6

0.7

0.8

0.9

1

Figure 6 – Total response of UTIAS, JAXA and TsAGI simulators in roll

Yaw 120


Phase [deg]

100 80 60 40 20 0 0

0.1

0.2

0.3

0.4

0.5 Gain

0.6

0.7

0.8

Figure 7 – Total response of UTIAS, JAXA and TsAGI simulators in yaw


0.9

1

Surge 200


Phase [deg]

150

100

50

0 0

0.1

0.2

0.3

0.4

0.5 Gain

0.6

0.7

0.8

0.9

1

Figure 8 – Total response of UTIAS, JAXA and TsAGI simulators in yaw

Sway 200


Phase [deg]

150

100

50

0 0

0.1

0.2

0.3

0.4

0.5 Gain

0.6

0.7

0.8

0.9

1

Figure 9 - Total response of UTIAS, JAXA and TsAGI simulators in Sway

Heave 200


Phase [deg]

150

100

50

0 0

0.1

0.2

0.3

0.4

0.5 Gain

0.6

0.7

0.8

Figure 10 - Total response of UTIAS, JAXA and TsAGI simulators in heave


0.9

1

Surge to Pitch

Sway to Roll

300 250 200 150 100

200 150 100

50 0 0

TsAGI 0.63 rad/s 1 rad/s 3 rad/s

250 Phase [deg]

Phase [deg]

300


50

0.1

0.2

0.3

0.4

0.5 Gain

0.6

0.7

0.8

0.9

0 0

1

0.1

0.2

0.3

0.4

0.5 Gain

0.6

0.7

0.8

0.9

1

Figure 13 – Surge-to-Pitch and Sway-to-Roll cross-coupling Nichols plots for TsAGI simulator

Pitch to Surge

Roll to Sway

300 250 200 150 100 50 0 0


250 Phase [deg]

Phase [deg]

300


200 150 100 50

0.1

0.2

0.3

0.4

0.5 Gain

0.6

0.7

0.8

0.9

1

0 0

0.1

0.2

0.3

0.4

0.5 Gain

0.6

0.7

0.8

0.9

1

Figure 14 – Pitch-to-Surge and Roll-to-Sway cross-coupling Nichols plots for TsAGI simulator

VII. Discussion The above results demonstrate the motion fidelity criterion applied to four simulators, three of which included the total transfer function data. The influence of the three parameters is clear when they are combined. In The addition of transport delay and the mechanical system performance (frequency response) will only deteriorate the overall performance. The variations between research simulators are significant, and even more than initially expected. However, these systems also vary in geometry and application. The TsAGI and UTIAS simulators have been tuned through successive optimization and indeed demonstrate good performance. The small size of the UTIAS system prevents it from very high gains. The limited data of the NLR simulator limits comparison with others. The boundaries on the lower right of each plot identify the target in which the curve, or a significant portion thereof, should occur. While it is not feasible to require that the gain and phase be maintained at tight tolerances over a huge frequency range, it is however necessary to define what is relevant and acceptable. Therefore it is proposed that a relevant frequency range be established over which the boundaries apply. In the examples shown, the entire frequency range has been shown, however it would be reasonable to bound the upper frequency to a value related to the aircraft dynamics. We propose a limit that is 2.5 times the natural frequencies of the aircraft or class of aircraft, and ensure that the total cueing system response remains within the target boundaries for up to this value. This limit and the gain/phase boundaries must also be verified with other research simulators, and with training simulators in commercial operation, whose performance level is known. The The criterion appears to match the expectations – well-tuned motion systems fit reasonably well within the bounds. If the system is poorly tuned, it will likely fall outside the boundaries shown. A good example of this is the JAXA simulator as shown in this study. It is a properly engineered simulator with high-performance visual display system, 11 American Institute of Aeronautics and Astronautics

control loading and operates using well-proven math models and data. In its early stages of operation, it has not yet been tuned, nor fully accepted by pilots. The motion platform dynamics exceeded those of other systems, and the transport delay was not excessively high. However, the particular choice of motion drive algorithm and tuning parameters introduce significant phase and gain shift to the system. Therefore, following a rigorous tuning process, the JAXA FSCAT-A flight simulator will most likely perform at least as well as other research simulators. Practical Application of the Criterion While this criterion should give insight into the behaviour of the motion cueing system, it does not exclude other tests. Objective testing as currently carried out to determine motion system smoothness and turn around bump are not captured herein. Subjective testing of the motion system should also be done, with the specific goal of ensuring the system does not exceed its mechanical limits during maneuvering motions. These should include lateral and longitudinal maneuvers such as pitch doublets and side-steps, under varying flight conditions.

VIII. Conclusions and Recommendations The criterion described here is intended to demonstrate the possibilities and limitations of objectively qualifying motion cueing systems in flight simulators. This is not, however, intended to be a stand-alone motion system quality measurement and, hence, it must complement other objective techniques. It also does not completely eliminate subjective evaluation. However, inconsistency can be reduced significantly. Only the boundaries of the proposed criterion are still somewhat subjective and need further verification. The criterion should also be extended to encompass false cue effects caused, for example, by the motion system reaching the limits of its travel. These shall be discussed with the stakeholders to come to acceptable boundaries. In order to fully validate and regulate simulators according to such boundaries, additional investigation with research and training simulators must be conducted. After its full assessment on research simulators, the criterion should subsequently be tested with training simulators. During this process, known simulators, i.e. simulators that have well-documented characteristics and performance, should be evaluated. The boundaries for gain, phase and frequency range, as proposed by Advani and Hosman, should be validated with relation to the motion cueing characteristics of flight training simulators, and these should be generally acceptable to simulator manufacturers, regulators and training organizations. The initial goal is to incorporate the criterion in the regulations to identify and correct bad motion cueing. Later on, as research in this area continues, it may be possible to assess the complete motion cueing system through an automated process. In the meantime, limitations of use and shortcomings of the technique can be determined in order to understand how the criteria can be improved. The end result will be a better understanding of what is needed to produce more consistently tuned flight training simulators. It allows and encourages a next step in the innovation of simulators not only to make flight training more effective, but also to maintain the competitiveness in motion simulation. Without the objectivity that such a criterion offers, these will remain a hiatus in answering basic questions on the need for motion cueing in simulation.

Acknowledgements In addition to their own effort in leading this development for the ICAO International Working Group, the authors have received great support from the following persons and organizations. -

Professor Peter Grant of University of Toronto Institute for Aerospace Studies, Downsview, Canada. Mr. Kohei Funabiki of the Japan Aerospace Exploration Agency JAXA, Tokyo, Japan. Dr. L. Zaichik, Central Aerohydrodynamic Institute TsAGI, Zhukovsky, Moscow Region, Russia. Ir. R.C.J. Ruigrok, National Aerospace Laboratory, NLR, Amsterdam, the Netherlands.

These individuals, their affiliations, and their colleagues are thanked for their significant contributions to this exercise.


References 1

Anon (2003). Manual of Criteria for the Quantification of Flight Simulators. ICAO Doc 9625, AN/938. International Civil Aviation Organization, Montreal, Canada.

2

Hosman, Ruud and Henk Stassen, “Pilot’s perception in the control of aircraft motions”. Control Engineering Practices. Vol. 7, No. 11, pp. 1421-1428, 1999.

3

Meiry, J.L., (1965). The vestibular system and human dynamic space orientation. Massachusetts Institute of Technology, Man-Vehicle Lab. NASA CR-628, 1966.

4

Stapleford, R.L., R.A. Peters, and F.R. Alex, (1969). Experiments and a model for pilot dynamics with visual and motion inputs. NASA CR-1325.

5

Burki-Cohen, J., E.M. Booth, N.N. Soja, R.D. DiSario, T. Go, T. Longridge (2000). The effect of platform motion. Proceedings of the Royal Aeronautical Society Conf. on Flight Simulation – The Next Decade. London, May 2000.

6

Go, T.H., J. Burki Cohen, W.H. Chung, J.A. Schroeder, S. Jacobs, and T Longridge (2003). The effects of enhanced hexapod motion on airline pilot recurrent training and evaluation. AIAA Modeling and Simulation Technologies Conference. Austin, TX, 11-14 August 2003.

7

Burki-Cohen, J., T.H. Go (2005). The Effects of Simulator Motion Cues on Initial Training of Airline Pilots. AIAA Modeling and Simulation Technologies Conference. San Francisco, Ca, 15-18 August 2005.

8

Advani, S.K., and R.J.A.W. Hosman (2006). Revising Civil Simulator Standards – An Opportunity for Technological Pull. AIAA Modeling and Simulation Technologies Conference and Exhibit, Paper AIAA-20066248. Keystone, Colorado, Aug. 21-24, 2006

9

Advani, S.K., and R.J.A.W. Hosman (2006). Towards Standardising High-Fidelity Cost-Effective Motion Cueing in Flight Simulation. Royal Aeronautical Society Conference on: Cutting Costs in Flight Simulation. Balancing Quality and Capability. London, November 7-8, 2006.

10

Grant, P. R., “Motion Characteristics of the UTIAS Flight Research Simulator Motion-Base,” UTIAS technical note no. 261, July 1986.

11

Advani, S.K., Nahon, M., Haeck, N., "Optimisation of Six-Degrees-of-Freedom Flight Simulator Motion Systems". AIAA J. of Aircraft, Vol. 36, No. 5, Sept-Oct. 1999.
