EUROPEAN ORGANISATION FOR THE SAFETY OF AIR NAVIGATION EUROCONTROL EUROCONTROL EXPERIMENTAL CENTRE

Transcription

1 EUROPEAN ORGANISATION FOR THE SAFETY OF AIR NAVIGATION EUROCONTROL EUROCONTROL EXPERIMENTAL CENTRE COSPACE 2003 AIRCRAFT GUIDANCE MODEL BASED EXPERIMENTS EVALUATION OF AIRBORNE SPACING FOR TRAFFIC SEQUENCING EEC Report No. 391 Project AGC-Z-FR Issued: April 2004 The information contained in this document is the property of the EUROCONTROL Agency and no part should be reproduced in any form without the Agency s permission. The views expressed herein do not necessarily reflect the official views or policy of the Agency.

2

3 REPORT DOCUMENTATION PAGE Reference: EEC Report No. 391 Security Classification: Unclassified Originator: Cospace Project Sector, Safety and Productivity Business Area, EEC Sponsor: EUROCONTROL Air-Ground Co-operative Air Traffic Services (AGC) Programme and EUROCONTROL Experimental Centre (EEC) in conjunction with European Commission (EC) Directorate General for Transport and Energy (DG TREN) Trans-European Network for Transport (TEN-T) programme. TITLE: Originator (Corporate Author) Name/Location: EUROCONTROL Experimental Centre Centre de Bois des Bordes B.P.15 F Brétigny-sur-Orge CEDEX FRANCE Telephone: +33 (0) Sponsor (Contract Authority) Name/Location: EUROCONTROL Agency 96, Rue de la Fusée B-1130 BRUXELLES Telephone: COSPACE 2003 AIRCRAFT GUIDANCE MODEL BASED EXPERIMENTS EVALUATION OF AIRBORNE SPACING FOR TRAFFIC SEQUENCING Authors Dan Ivanescu (Steria), Chris Shaw, Eric Hoffman, Karim Zeghal Date 04/2004 Pages xiv + 55 Project CoSpace AGC-Z-FR Distribution Statement: (a) Controlled by: EUROCONTROL Project Manager (b) Special Limitations: None (c) Copy to NTIS: YES / NO Figures 36 Tables 9 Task No. Sponsor AGC-Z-FR-0000 Annexes 1 References 22 Period Descriptors (keywords): Airborne Separation Assistance System Air Traffic Management Aircraft Guidance Airborne Spacing Fast-Time Experiment Wind and Turbulence Mixed Aircraft Type Traffic Sequencing Automatic Dependent Surveillance Broadcast. Abstract: This report presents the results of aircraft guidance model based experiments conducted at the Eurocontrol Experimental Centre (EEC) in 2002 and These experiments fitted in with a series of human-in-the-loop and model based validation exercises aimed at investigating the use of spacing instructions (denoted airborne spacing) for sequencing of arrival flows. The human-in-the-loop experiments (pilots and controllers) enable an understanding of the impact of airborne spacing on human activity, effectiveness and safety. The model based experiments were intended to complement the validation of airborne spacing by enabling prototyping as well as the collection of a large amount of data under varying conditions. The series of model-based experiments described in this report aimed at: (i) prototyping spacing guidance algorithms for humanin-the-loop experiments; (ii) understanding the intrinsic dynamics of sequences of aircraft under normal and extreme operating conditions (varying entry conditions, wind turbulence and aircraft types); (iii) evaluating the effects of air-air surveillance transmission quality (e.g. ADS-B, Automatic Dependent Surveillance Broadcast, update rate, latency and accuracy) on the performance of airborne spacing. Two complementary operational airborne spacing applications were studied: Merge (aircraft on converging trajectories) and Remain (aircraft on same trajectory). From the three high level objectives introduced above specific objectives were derived to be studied separately. For the Merge application, different distance based guidance laws were developed and compared. The most robust was capable of merging multiple aircraft in descent under turbulent wind conditions. This guidance law was selected for studying the effects of initial distances and speeds on the ability of an aircraft to descend and establish a stable spacing behind another by a given merge waypoint point. For the Remain application the effects of ADS-B transmission quality, time based spacing criteria, mixed aircraft types and varying wind conditions on the ability of aircraft to maintain a given along-track spacing (distance or time) behind a descending lead aircraft were investigated. The results presented in the report, based on normal operating conditions, are consistent with the hypothesis that the airborne spacing Merge and Remain applications are robust. The results based on extreme conditions may be useful to evaluate the limits of applicability for Merge and Remain applications.

4

5 COSPACE 2003 Aircraft Guidance Model Based Experiments EUROCONTROL EXECUTIVE SUMMARY This report presents the results of aircraft guidance model based experiments conducted at the Eurocontrol Experimental Centre (EEC) in 2002 and These experiments fitted in with a series of human-in-the-loop and model based validation exercises aimed at investigating the use of spacing instructions (denoted airborne spacing) for sequencing of arrival flows. The human-in-theloop experiments (pilots and controllers) enable an understanding of the impact of airborne spacing on human activity, effectiveness and safety. The model based experiments were intended to complement the validation of airborne spacing by enabling prototyping as well as collecting a large amount of data under varying conditions. A series of model-based experiments was conducted using MATLAB/Simulink, with essentially an airborne perspective. A previous initial experiment allowed an understanding of the impact of speed and altitude profiles on airborne spacing. The subsequent experiments described in this report aimed at extending the scope as follows. Two complementary operational airborne spacing applications were studied: Merge (aircraft on converging trajectories) and Remain (aircraft on same trajectory). The three high level objectives below were translated in to specific objectives: (i) Prototyping spacing guidance algorithms for human-in-the-loop experiments Four spacing guidance laws for sequencing aircraft on merging trajectories were studied: a linear distance-based guidance, a linear speed-based guidance, a non-linear guidance based on trajectory prediction and a linear guidance based on the lead groundspeed profile. To evaluate these guidance laws, a set of performance criteria was established along with metrics of spacing distance error and airspeed difference at the merging waypoint and a set of operational scenarios was developed. The distance-based linear guidance produced both large spacing distance errors and large speed errors at the merging waypoint. The speed based linear guidance achieved near zero error spacing but with very large speed errors for all scenarios. The non-linear guidance produced the desired spacing with minor errors and with the similar aircraft speeds at the waypoint. However, when aircraft are in descent, this guidance becomes difficult to design (complex computations, long simulations requiring high memory resources). The best performing guidance was that based on the lead aircraft groundspeed profile. This guidance was able to merge multiple aircraft in descent under turbulent wind conditions with near-zero error spacing and with the similar aircraft speeds. Therefore the guidance based on lead aircraft groundspeed profile, in the loop with a spacing director and a human pilot model, was selected for further studies. (ii) Understanding intrinsic dynamics of sequences of aircraft under normal and extreme operating conditions (initial distances and speeds, wind turbulence and aircraft types) A first experiment was performed to study the effects of entry conditions (initial distances and speeds) on the ability of an aircraft to descend from 29,000 feet to 3,000 feet and establish a stable spacing (8 NM) behind another by a given merge point. Results for two and three aircraft at the same initial speed show how the possible initial spacing error envelope grew, when the initial distance of the lead aircraft to the merge waypoint was increased. The impact on initial spacing error of varying the difference in initial speed was slight. The effects of mixed aircraft types and wind conditions on the ability of an aircraft to maintain a constant time delay along-track spacing behind a descending lead aircraft were then investigated. An exact constant time delay spacing criterion based on lead aircraft position history was used to compare the spacing performance of all combinations of heavy and light aircraft for different wind conditions (with or without turbulence). Results show for both constant and turbulent winds that cross-track winds could have just as detrimental an effect on along-track time based spacing performance as along-track winds. Project AGC-Z-FR - EEC Report No. 391 v

6 EUROCONTROL COSPACE 2003 Aircraft Guidance Model Based Experiments Turbulent winds severely degraded time based spacing stability particularly in cross and head wind conditions. Tail winds were the least disturbing for time based spacing. Considering the aircraft type, it was established that a heavy aircraft following a light tended to produce the maximum spacing errors. The other three combinations of aircraft type resulted in similar maximum spacing error behaviour. (iii) Evaluating effect of air-air surveillance transmission quality (e.g. ADS-B, Automatic Dependent Surveillance Broadcast, update rate, latency and accuracy) on the performance of airborne spacing The effects of ADS-B transmission quality on the airborne spacing performance of a sequence of six aircraft in descent were investigated. Results show that constant time delay spacing remained more stable than constant distance spacing for degradations in ADS-B update rate, latency and accuracy. Transmission loss compensation terms added to the guidance produced significant improvements in spacing guidance stability and allowed similar performances as in the case without degradations. Conclusions and Future Experiments The results presented in the report are consistent with the hypothesis that the airborne spacing Merge and Remain applications should be robust under normal conditions. The results based on extreme conditions may be useful to evaluate the limits of applicability for Merge and Remain applications. Data collected from future human-in-the-loop experiments could be used to improve human pilot and spacing director models. In turn, the subsequent model based simulations should improve the spacing assistance provided to the flight crew for the pilot-in-the-loop simulations. With model based experiments the potential exists to study safety aspects by observing the impact of varying several parameters in a large number of trials (MonteCarlo simulations). Eventual interaction with an ACAS Airborne Collision Avoidance System needs investigation, especially for merge situations. vi Project AGC-Z-FR EEC Report No. 391

7 COSPACE 2003 Aircraft Guidance Model Based Experiments EUROCONTROL ACKNOWLEDGEMENTS This project is sponsored jointly by the EUROCONTROL Air-Ground Co-operative Air Traffic Services (AGC) programme, EUROCONTROL Experimental Centre (EEC) and European Commission (EC) Directorate General for Transport and Energy (DG TREN) Trans-European Network for Transport (TEN-T) programme. Project AGC-Z-FR - EEC Report No. 391 vii

8 EUROCONTROL COSPACE 2003 Aircraft Guidance Model Based Experiments TABLE OF CONTENTS LIST OF ANNEXES... IX LIST OF FIGURES... IX LIST OF TABLES... X REFERENCES... XI RELATED PUBLICATIONS... XII ABBREVIATIONS... XIV 1. INTRODUCTION PRINCIPLES BACKGROUND CONTEXT OF THE PROJECT PREVIOUS WORK OBJECTIVES AND DEFINITIONS OBJECTIVES DEFINITIONS Merge Application Remain Application ADS-B Transmission Quality GUIDANCE DESIGN MERGE APPLICATION Quality Criteria Linear Guidance Non-Linear ( Bang-bang ) Guidance Lead Groundspeed Based Guidance REMAIN APPLICATION Quality Criteria Linear Guidance APPARATUS FAST-TIME SIMULATION ENVIRONMENT ADS-B MODEL AIRCRAFT MODEL Aircraft Dynamics Model Autopilot Model Spacing Director Human Pilot Model ATMOSPHERE AND WIND MODELS viii Project AGC-Z-FR EEC Report No. 391

9 COSPACE 2003 Aircraft Guidance Model Based Experiments EUROCONTROL 6. METHOD MERGE EXPERIMENTS Comparison of Linear and Non-Linear Guidance Evaluation of Lead Groundspeed Based Guidance and Spacing Director Effect of Entry Conditions REMAIN EXPERIMENTS Effect of ADS-B Transmission Quality Evaluation of Time-Based Spacing Criteria Effect of Mixed Aircraft Types and Varying Wind Conditions RESULTS MERGE EXPERIMENTS Comparison of Linear and Non-Linear Guidance Evaluation of Lead Groundspeed Based Guidance and Spacing Director Effect of Entry Conditions REMAIN EXPERIMENTS Effect of ADS-B Transmission Quality Evaluation of Time-Based Spacing Criteria Effect of Mixed Aircraft Types and Varying Wind Conditions SUMMARY OF MAIN RESULTS SUMMARY OF MERGE RESULTS SUMMARY OF REMAIN RESULTS CONCLUSIONS FUTURE EXPERIMENTS FRENCH TRANSLATION (TRADUCTION EN LANGUE FRANÇAISE)...45 LIST OF ANNEXES Annex LIST OF FIGURES Figure 1: Merge, cockpit display of traffic information... 6 Figure 2: Merge example... 6 Figure 3: Remain, cockpit display of traffic information... 7 Figure 4: Remain example... 7 Figure 5: A speed-time graph Figure 6: Speed-time graph of bang-bang control: demand and response Figure 7: Closed loop control law system block diagram with human pilot model Figure 8: The wind model components Figure 9: Turbulent wind versus altitude - JAR-AWO model (20 knots wind at 30 feet AGL) Figure 10: Merge example three aircraft Figure 11: Merge example admissible entries of trail Figure 12: Comparison of absolute distance spacing error at the merge waypoint Project AGC-Z-FR - EEC Report No. 391 ix

10 EUROCONTROL COSPACE 2003 Aircraft Guidance Model Based Experiments Figure 13: Comparison of absolute magnitude of groundspeed difference at the merge waypoint Figure 14: a) Spacing distance error between aircraft b) Comparison of true airspeed and groundspeed of the lead and trailing aircraft Figure 15: Comparison between the desired CAS given by the guidance law and the demanded CAS from the spacing director (as input to the pilot model) Figure 16: Entry conditions initial spacing error bounds for a sequence of two aircraft having the same initial speeds Figure 17: Entry conditions initial spacing error for a sequence of two aircraft having different initial speeds Figure 18: a) Entry conditions initial spacing error for a sequence of three aircraft with the same speeds b) Initial position for a sequence of three aircraft with the same speeds Figure 19: Impact of ADS-B update rate on in-trail performances: maximum spacing error (%) as function of update period. Solid lines: basic guidance law, dashed lines: guidance with update rate compensation Figure 20: Engine throttle oscillations (%) function of probability of reception of ADS-B, guidance with update rate compensation Figure 21: Impact of ADS-B latency on spacing error, basic guidance law Figure 22: Impact of ADS-B latency on spacing error, guidance with latency compensation Figure 23: Impact of ADS-B accuracy on in-trail spacing performance Figure 24: Impact of ADS-B accuracy on in-trail performance Figure 25: ACTD - true airspeed behaviour Figure 26: CTD - true airspeed behaviour Figure 27: CTD and ACTD, time spacing error comparison for a continuous update rate Figure 28: Turbulent cross wind effects on largest of maximum spacing errors Figure 29: Turbulent head wind effects on largest of maximum spacing errors Figure 30: Turbulent tail wind effects on largest of maximum spacing errors Figure 31: Turbulent cross wind effects on mean of maximum spacing errors Figure 32: Turbulent head wind effects on mean of maximum spacing errors Figure 33: Turbulent tail wind effects on mean of maximum spacing errors Figure 34: a) Time spacing errors for constant head and tail winds b) Equivalent distance based spacing errors for constant head and tail winds Figure A-1: Simulation results check-case Figure A-2: Simulation results check-case LIST OF TABLES Table 1: Application chosen to study for each specific objective: Yes means a performed study, No means a potential future study... 5 Table 2: Navigation Accuracy Categories (NAC)- Position and Velocity (RTCA, 2002) Table 3: Mean wind speed increasing with altitude and function of wind measured at 30 feet AGL Table 4: Initial conditions for comparing scenarios Table 5: The wind parameters Table 6: ADS-B experimental parameters and perturbation ranges Table 7: Mixed aircraft types and winds: experimental parameters Table 8: Table 9: Current status of the fast-time simulations: Yes in the table means a performed study, No means a potential future study Remain : performance of improved spacing guidance for realistic ADS-B transmission x Project AGC-Z-FR EEC Report No. 391

11 COSPACE 2003 Aircraft Guidance Model Based Experiments EUROCONTROL REFERENCES [1] Abbott, T.S., (2002): Speed Control Law for Precision Terminal Area In-Trail Self Spacing, NASA Technical Memorandum , NASA Langley Research Center, Virginia, USA. [2] Federal Aviation Administration/EUROCONTROL, (2001): Principles of Operation for the Use of Airborne Separation Assurance Systems, FAA/EUROCONTROL Cooperative R&D. [3] Fokker, (1989): Flight Simulator Data for the Fokker F28 mk100, Report L , issue 2, September, Restricted. [4] Grimaud, I., Hoffman, E., Rognin, L. and Zeghal, K., (2003): EACAC 2001 Real-Time Experiment, EUROCONTROL Experimental Centre (EEC) Report No. 380 and 380-2, Vol.1 and 2, Released Issue: March 2003, EUROCONTROL Experimental Centre, France. [5] Hammer, J., (2000): Case Study of Paired Approach Procedure to Closely Spaced Parallel Runways, Air Traffic Control Quarterly, Vol. 8(3), pp [6] Hanke, J. and Nordwall, S., (1995): The Simulation of a Jumbo Jet Transport aircraft, NASA Report D , NASA Langley Research Center, Virginia, USA. [7] Hoffman, E., (1993): Contribution to Aircraft Performance Modelling for ATC use, EEC Report 258, EUROCONTROL Experimental Centre, France. [8] Joint Aviation Authorities, (1996): Joint Aviation Requirement-All Weather Operations, JAR 07/03-13, Netherlands. [9] Kelly, J.R., (1983): Effect of Lead-Aircraft Ground-Speed Quantization on Self Spacing Performance Using a Cockpit Display on Traffic, NASA Technical Paper 2194, NASA Langley Research Center, Virginia, USA. [10] Lambregts, A.A., (1983): Integrated System Design for Flight and Propulsion Control Using Total Energy Principles, AIAA Aircraft Design, Systems and Technology Neeting, AIAA Paper , Texas, USA. [11] Mathworks Inc., (2003): Matlab Users Guide, Mathworks, Natick, Massachusetts, USA. [12] Mathworks Inc., (2003): Simulink Users Guide, Mathworks, Natick, Massachusetts, USA. [13] McGruer, D., (1973): Aircraft Dynamics and Automatic Control, Princeton University Press. [14] NLR National Aerospace Laboratory, (2002): AMAAI Modelling Toolset for the Analysis of In-trail Following Dynamics, NLR Report CR , Netherlands. [15] Pritchett, A.R. and Yankosky L.J., (1998): Simultaneous Design of Cockpit Display of Traffic Information & Air Traffic Management Procedures, SAE Transactions - Journal of Aerospace, Paper [16] Pritchett, A.R. and Yankosky L.J., (2000): Pilot Performance at New ATM Operations: Maintaining In-Trail Separation and Arrival Sequencing, Proceedings of AIAA Guidance, Navigation, and Control Conference, Colorado, USA. Project AGC-Z-FR - EEC Report No. 391 xi

12 EUROCONTROL COSPACE 2003 Aircraft Guidance Model Based Experiments [17] RTCA, (1998): Minimum Aviation System Performance Standards for Automatic Dependent Surveillance Broadcast (ADS-B), RTCA Paper DO-242 / SC186, RTCA Inc., USA. [18] RTCA, (2000): Operational Concepts for Cockpit Display of Traffic Information (CDTI) Initial Applications, RTCA Paper DO-259 / SC186, RTCA Inc., USA. [19] RTCA, (2002): Minimum Aviation System Performance Standards for Automatic Dependent Surveillance Broadcast (ADS-B), RTCA Paper DO-242A / SC186, RTCA Inc., USA. [20] Sorensen, J.A. and Goka, T., (1983): Analysis of In-Trail Following Dynamics of CDTI- Equipped Aircraft, Journal of Guidance, Control and Dynamics, Vol. 6, pp [21] Vinken, P., Hoffman, E., Zeghal, K., (2000): Influence of Speed and Altitude Profile on the Dynamics of In-trail Following Aircraft, Proceedings of AIAA Guidance, Navigation and Control Conference, Colorado, USA. [22] Zeitlin, A.D., (2001): Safety Assessments of ADS-B and ASAS, 4 th USA / Europe Air Traffic Management R&D Seminar, New Mexico, USA. RELATED PUBLICATIONS Ivanescu, D., Hoffman E. and Zeghal K., (2002a): Impact of ADS-B Link Characteristics on the Performances of In-Trail Following Aircraft, Proceedings of AIAA (American Institute of Aeronautics and Astronautics), Guidance, Navigation, and Control Conference, Monterrey, California, USA. Ivanescu, D., Hoffman, E., Shaw C. and Zeghal K., (2002b): Analysis of Spacing Guidance for Sequencing Aircraft on Merging Trajectories, 21 st Digital Avionics Systems Conference, Irvine, California, USA. Hoffman, E., Ivanescu, D., Shaw, C. and K. Zeghal, (2003a): Analysis of Constant Time Delay Airborne Spacing Between Aircraft of Mixed Types in Varying Wind Conditions ; Proceedings of 5 th USA / Europe Air Traffic Management R&D Seminar, Budapest, Hungary. Hoffman, E., Ivanescu, D., Shaw, C. and K. Zeghal, (2003b): Effect of Mixed Aircraft Types and Wind on Time Based Airborne Spacing, Proceedings of AIAA Guidance, Navigation, and Control Conference, Austin, Texas, USA. Ivanescu, D., Shaw, C., Hoffman, E. and K. Zeghal, (2003): Effect of Entry Conditions on Airborne Spacing when Sequencing Multiple Converging Aircraft, Proceedings of AIAA Guidance, Navigation, and Control Conference, Austin, Texas, USA. Hoffman, E., Ivanescu, D., Shaw, C. and Zeghal K., (2003): Effect of Automatic Dependent Surveillance Broadcast (ADS-B) Transmission Quality on the Ability of Aircraft to Maintain Spacing in a Sequence, Air Traffic Control Quarterly Special Issue: Aircraft Surveillance Applications of ADS-B, Vol. 11(3). xii Project AGC-Z-FR EEC Report No. 391

13 COSPACE 2003 Aircraft Guidance Model Based Experiments EUROCONTROL CoSpace Project Reports: Aligne, F., Grimaud, I., Hoffman, E., Rognin, L. and K. Zeghal, (December 2003): CoSpace 2002 controller experiment assessing the impact of spacing instructions in E-TMA and TMA Volume I and II, EEC Report No. 386, Eurocontrol Experimental Centre. Hebraud, C., Hoffman, E., Papin, A., Pene, N., Rognin, L., Sheehan, C. and K. Zeghal, (February 2004): CoSpace 2002 flight deck experiments assessing the impact of spacing instructions from cruise to initial approach Volume I and II, EEC Report No. 388, Eurocontrol Experimental Centre. CoSpace Project Web Site: Project AGC-Z-FR - EEC Report No. 391 xiii

14 EUROCONTROL COSPACE 2003 Aircraft Guidance Model Based Experiments ABBREVIATIONS Abbreviation ACAS ACTD ADS-B AGL ASAS ATC ATM CAS CD CTD CDTI EEC FAA FL GS ISA JAR NAC NACp NACv NM PI RTCA TAS TIS-B TMA De-Code Airborne Collision Avoidance System Approximate Constant Time Delay Automatic Dependant Surveillance Broadcast Above Ground Level Airborne Separation Assistance System Air Traffic Control Air Traffic Management Calibrated Air Speed Constant Distance Constant Time Delay Cockpit Display of Traffic Information EUROCONTROL Experimental Centre Federal Aviation Administration Flight Level Ground Speed International Standard Atmosphere Joint Aviation Authorities Navigation Accuracy Categories NAC horizontal position NAC velocity Nautical Mile Proportional Integral Radio Technical Commission for Aeronautics True Air Speed Traffic Information Service Broadcast TerMinal control Area xiv Project AGC-Z-FR EEC Report No. 391

15 COSPACE 2003 Aircraft Guidance Model Based Experiments EUROCONTROL 1. INTRODUCTION This report presents the results of aircraft spacing guidance model based experiments conducted at the Eurocontrol Experimental Centre (EEC) in 2002 and These experiments fitted in with a series of human-in-the-loop and model based validation exercises aimed at investigating the use of spacing instructions (denoted airborne spacing) for sequencing of arrival flows. The human-in-theloop experiments (pilots and controllers) enable an understanding of the impact of airborne spacing on human activity, effectiveness and safety. The model based experiments were intended to complement the validation of airborne spacing by enabling prototyping as well as collecting a large amount of data under varying conditions. A series of model-based experiments is conducted using MATLAB/Simulink, with essentially an air perspective. A previous initial experiment allowed an understanding of the impact of speed and altitude profiles on airborne spacing. The subsequent experiments described in this report aimed at extending the scope of the initial experiment by: (i) (ii) (iii) prototyping spacing guidance algorithms for the human-in-the-loop experiments; understanding intrinsic dynamics of sequences of aircraft under normal and extreme operating conditions (varying entry conditions, wind turbulence and aircraft types); evaluating the air-air surveillance transmission quality (ADS-B update rate, latency and accuracy) on the performances of airborne spacing. The document is organised as follows: Section 2 introduces the principles, the background and the project context. Section 3 presents the objectives and briefly defines the operational context: airborne spacing applications Merge and Remain and ADS-B transmission quality. Section 4 describes the corresponding guidance laws. Section 5 describes the model based environment: aircraft, ADS-B, wind and atmosphere models, MATLAB/Simulink platform. Section 6 gives the experiment set-up. Section 7 presents the main findings with graphs and discussions. Section 8 summarises main results and makes recommendations. Section 9 presents the conclusions. French translation of executive summary, introduction, objectives, recommendations and conclusions. Appendix of aircraft model validation. A series of documented experiments with corresponding prototype guidance systems was produced addressing each of the above goals. Results of experiments were published at conferences where appropriate (see Related Publications section). Project AGC-Z-FR EEC Report No

16 EUROCONTROL COSPACE 2003 Aircraft Guidance Model Based Experiments 2. PRINCIPLES 2.1. BACKGROUND Air Traffic Management (ATM) concepts are currently being evaluated where air traffic controllers may have the possibility to give some spacing related instructions to the flight deck. For example, in zones of convergence, pilots could be instructed to establish and maintain a prescribed spacing with respect to an aircraft in front (lead) by monitoring the spacing and adjusting their own speed [FAA/EUROCONTROL, 2001]. Possible operational benefits could be fewer time critical instructions, more strategic ways of working and greater traffic awareness for pilots (see CoSpace real time experiments [Grimaud et al. 2003], and references therein). It is expected that the increased controller availability could lead to improved safety, which in turn could enable better efficiency and/or, depending on airspace constraint, more capacity. In addition, it is expected that flight crew could gain in awareness and anticipation by taking an active part in the management of the situation with respect to the concerned aircraft. In this context the air-to-air surveillance technology ADS-B is one of the potential key enablers required to support the implementation of such an application. Surveillance information, such as position and velocity of the lead aircraft, is transmitted from the lead aircraft to the trailing aircraft by ADS-B (or TIS-B) [RTCA, 1998, 2000, 2002] and may be displayed to the pilot on a Cockpit Display of Traffic Information (CDTI) as part of an Airborne Separation Assistance System (ASAS) CONTEXT OF THE PROJECT The objective of CoSpace is to assess the operational feasibility, benefits and limits of the use of spacing instructions ( airborne spacing ) as well as to identify the evolutions induced or required with respect to today s ATC (Air Traffic Control). CoSpace covers concept definition up to validation aspects including human-in-the-loop and model-based simulations. The focus is on the sequencing of arrival flows from cruise to initial approach (extended TMA TerMinal control Area ) and more recently down to final approach (TMA). Since its inception in 1998, the project follows an iterative process, in which every step helps defining the next one. Because of the inherent air-ground nature of the concept, both controller and pilot aspects have to be considered. After an initial air-ground human-in-the-loop simulation, two series of air and ground simulations were conducted: five ground simulations with in total 34 controllers from different European countries over 14 weeks; four air simulations with in total 31 test and airline pilots over 25 days. The consistency between the two series is ensured essentially by relying on the same concept (applications, procedures and phraseology), the same operational environment (type of airspace, scenarios) and a unified validation framework (experimental plan, metrics). Model-based simulations have been introduced to complement the validation of airborne spacing. A first series of aircraft guidance model-based simulations has been conducted, with essentially an air perspective. They rely on the modelling of aircraft, autopilot, spacing guidance and pilot behaviour. This series of simulations allows some understanding of the intrinsic dynamics of sequences of aircraft, the impact of airborne surveillance characteristics (e.g. ADS-B update rate), the effect of wind and of aircraft type. The data collected during the pilot-in-the-loop simulations allowed refinement of the pilot model. In turn, the subsequent model-based simulations allow improvements of the spacing assistance provided to the flight crew for the pilot-in-the-loop simulations, and to refine the experimental plan (e.g. selection of scenarios). 2 Project AGC-Z-FR EEC Report No. 391

17 COSPACE 2003 Aircraft Guidance Model Based Experiments EUROCONTROL A second series of model-based simulations was introduced which a controller perspective. The modelling investigated the impact of airborne spacing on controller workload in the extended TMA. The controller was modelled by defining standardised ATC tasks performed with and without airborne spacing. The allocation of these tasks during the simulations enabled the impact of spacing instructions on controller workload to be assessed. This modelling sought to provide objective workload results to extend the subjective results collected during the November 2002 Real Time Controller Simulation PREVIOUS WORK Sequencing aircraft applications were widely studied in the literature, first solutions being obtained in the 1980 s [Sorensen and Goka, 1983]. The study covered both analytical and experimental aspects. Three separation criteria were introduced (constant distance, constant time predictor, and constant time delay). To simulate strings of aircraft a mathematical model was used. In a further analysis ([Kelly, 1983] and the references therein) two separation criteria were investigated through pilot-in-the-loop experiments with spacing cues on the cockpit display. In the 1990 s, computer based models were used along with pilot-in-the-loop [Pritchett and Yankosky, 1998, 2000]. In recent years the impact of top of descent [Vinken et al. 2000] and speed reduction on in-trail following ( Remain ) aircraft performance was investigated through mathematical models, and more advanced guidance algorithms were proposed [Abott, 2002]. Merging applications seemed to be less well studied. [Grimaud et al, 2000] developed a detailed operational concept for a Merge application. Real-time simulations show that it is important that both air traffic controller and pilots have a clear understanding of whether the complete Merge manoeuvre is feasible at the time the instruction is issued. In [Hoffman et al, 2002] three different guidance controllers for merging pairs of aircraft at a fixed waypoint in level flight with no wind were investigated. Although standards are under development and initial recommendations are proposed [RTCA, 2002], [Zeitlin, 2001], the impact of ADS-B characteristics on the performances of the considered applications remains a key issue. An experimental study focussing on closely spaced parallel approaches has been carried out by [Hammer, 2000]. Concerning the realism of the past studies on sequencing aircraft guidance, most of them reviewed tended to assume perfect ADS-B transmission quality, similar aircraft performance [Pritchett and Yankosky 2000] and simple constant wind models. To add to the current body of knowledge on airborne spacing for sequencing and merging, our studies therefore addressed the following aspects: (i) Merge and Remain applications (ii) distance and time based spacing guidance (iii) mixed aircraft types (iv) different constant and turbulent wind conditions (v) different ADS-B update rates, latencies and accuracies (vi) different entry conditions. Project AGC-Z-FR EEC Report No

18 EUROCONTROL COSPACE 2003 Aircraft Guidance Model Based Experiments 3. OBJECTIVES AND DEFINITIONS 3.1. OBJECTIVES For each application, the three high level objectives introduced above were translated in to specific objectives as follows: High level objectives Specific objectives Prototyping of spacing guidance algorithms for the human-in-the-loop experiments Develop guidance law ( Remain and Merge ). Develop spacing director system ( Merge ). Develop time based spacing criteria ( Remain ). Increase understanding of intrinsic dynamics of sequences of aircraft under normal and extreme operating conditions Measure effect on spacing performance of initial position and speed ( Merge ). Measure effect on spacing performance of heavy and light aircraft types ( Remain ). Measure effect on spacing performance of constant and turbulent wind conditions ( Remain ). Evaluate effect of air-air surveillance transmission quality on the performance of airborne spacing Measure effect on spacing performance of ADS-B update rate, latency and accuracy ( Remain ). 4 Project AGC-Z-FR EEC Report No. 391

19 COSPACE 2003 Aircraft Guidance Model Based Experiments EUROCONTROL Representative cases were studied as follows: Table 1: Application chosen to study for each specific objective: Yes means a performed study, No means a potential future study. Application Develop guidance laws Develop spacing director Develop time based spacing criteria Measure effect on spacing performance of initial position and speed Measure effect on spacing performance of heavy and light aircraft Measure effect on spacing performance of constant and turbulent wind conditions Measure effect on spacing performance of ADS-B transmission quality Merge Yes Yes No Yes No No No Remain Yes No Yes No Yes Yes Yes 3.2. DEFINITIONS Merge Application The Merge application involves an air traffic controller asking a pilot to select 1 a neighbouring aircraft as a target on a CDT I (see Figure 1). An example of the phraseology developed is: Controller: DLH456, select target 1234 Pilot of DLH456: Selecting target 1234, DLH456 Once the target has been selected, the air traffic controller can then ask the pilot to Merge behind the target at a given waypoint ahead with a given spacing. An example of the phraseology developed is: Controller: DLH456, behind target, merge waypoint 8 NM behind Pilot of DLH456: Merging way point, 8 NM behind target, DLH456 where the spacing could be defined as a distance in nautical miles (NM) or time. 1 Note: The target is identified by a unique code other than the callsign to avoid confusion over the radio communication party-line. Project AGC-Z-FR EEC Report No

20 EUROCONTROL COSPACE 2003 Aircraft Guidance Model Based Experiments Figure 1: Merge, cockpit display of traffic information An example of the Merge application is illustrated in Figure 2. The two aircraft, the lead (target) and the trail aircraft, are flying direct to the same fixed waypoint. The solid arrows represent the current position and track angle of the aircraft, and the dashed arrows represent the desired positions of the two aircraft when the lead will reach the waypoint. At the moment the lead reaches the waypoint the spacing between aircraft must be equal to the desired spacing within a defined tolerance, and the aircraft should have similar speeds. After the waypoint the problem is similar to the in-trail following aircraft situation, i.e. each aircraft follows its own trajectory within a sequence maintaining the spacing between itself and the aircraft immediately in front. Mathematically, this concept can be expressed as follows: let d trail and d lead be the distances from the current positions of the respective aircraft to the waypoint. When the lead reaches the waypoint (d lead =0), the spacing error (d error ) is the difference between the remaining distance of the trailing aircraft to the waypoint and the desired spacing: d error = d trail d lead = 0) ( d (3.1) spacing Merge Behind target, merge WPT 8NM behind WPT D AFR DLH Figure 2: Merge example 6 Project AGC-Z-FR EEC Report No. 391

21 COSPACE 2003 Aircraft Guidance Model Based Experiments EUROCONTROL Remain Application Remain is the second airborne spacing application studied (Figure 3 and Figure 4). The Remain application involves an air traffic controller asking a pilot to select a neighbouring aircraft as a target on a CDTI as in Merge. An example of the phraseology developed is: Controller: DLH456, select target 1234 Pilot of DLH456: Selecting target 1234, DLH456 Once the target has been selected, the air traffic controller can then ask the pilot to maintain a given longitudinal distance or time behind the target. An example of the phraseology developed is: Controller: DLH456, behind target, remain 8 NM behind Pilot of DLH456: Remaining 8 NM behind target, DLH456 where the spacing could be defined as a distance in nautical miles (NM) or time. Figure 3: Remain, cockpit display of traffic information Remain Behind target, remain 8NM behi nd AFR DLH D Figure 4: Remain example Project AGC-Z-FR EEC Report No

22 EUROCONTROL COSPACE 2003 Aircraft Guidance Model Based Experiments ADS-B Transmission Quality Concerning the ADS-B transmission quality, the following three characteristics were considered: Update rate of reports: sustained rate at which periodic ADS-B reports were received. Latency of transmission: delay between the time when the ADS-B report was received and processed, and the time when position and velocity were measured. This included not only ADS-B latency, but also additional delays in the processing of the information. Accuracy of surveillance information: difference between the state vector transmitted by ADS-B and the true values. 8 Project AGC-Z-FR EEC Report No. 391

23 COSPACE 2003 Aircraft Guidance Model Based Experiments EUROCONTROL 4. GUIDANCE DESIGN 4.1. MERGE APPLICATION A first theoretical study was devoted to finding appropriate guidance for establishing a desired spacing and speed between aircraft when merging in flight. Various guidance spacing criteria and guidance laws were designed and tested during the period October 2002 March The main results are presented in the following sub-section. The most appropriate guidance design was chosen for further studies Quality Criteria Such guidance laws aim at establishing a given spacing at the waypoint to a lead aircraft through speed adjustments as a pilot or as cockpit automation might do. The guidance law receives surveillance data from the lead aircraft and feeds the desired target calibrated airspeed input ( CAStraildesired ) of the aircraft model. The target altitude (hdes) is fed independently, and depends on the top of descent scenario. The following criteria were considered when designing merge spacing guidance laws: i) Spacing error at the merging waypoint should be small. ii) iii) iv) The guidance should be robust against large initial spacing errors. The trailing aircraft should have a similar speed to that of the lead aircraft for a smooth transition in to a Remain application after the waypoint. The frequency of speed changes asked of the pilot in the trailing aircraft should be low enough to be operationally acceptable. v) The speed profile of the trailing aircraft should be smooth with minimal speed deviations because it may itself be a target leading another trailing aircraft behind Linear Guidance First linear guidance laws were developed starting from initial ideal conditions (i.e. no spacing error, or near zero). In a classical linear approach, common pre-defined spacing criteria can be employed to predict and characterise the desired spacing at the waypoint. A guidance law was then designed to control and minimise the spacing error. Based on such prediction, this spacing error was computed in two ways: Linear Guidance: Distance-Based Prediction The simplest solution seemed to be adapting the already existing criteria defined in [Pritchett and Yankosky, 1998], [Sorensen and Goka, 1983] for the in-trail concept. The use of the same criteria for both concepts (i.e. merging and in-trail ) facilitates the passage from one operational mode ( merging ) to the other ( in-trail ). Project AGC-Z-FR EEC Report No

24 EUROCONTROL COSPACE 2003 Aircraft Guidance Model Based Experiments If no prediction based on current speeds is considered, equation (3.1) can be approximated as the difference between the two aircrafts current distance to the waypoint taking into account the desired spacing: d error = d d d (4.1) trail lead spacing Linear Guidance: Speed-Based Prediction The distance-based prediction criteria can be enhanced using the current aircraft speeds (V trail, V lead ). Assuming that these speeds are constant, the predicted spacing at the waypoint can be approximated as follows: The predicted time needed for the lead aircraft to reach the waypoint is: t = leadtogo d lead V lead (4.2) In the meantime, the trailing aircraft covers a distance equal to: dist = t leadtogo V trail = d lead V trail V lead (4.3) Finally the spacing error is defined as: d error V = dtrail ( d trail lead ) dspacing (4.4) V lead Linear Guidance Law The spacing error d error guidance law had the following form: V cmd = K p d error was controlled by a corresponding linear guidance law. In both cases this ( ) ( K s s s ) 2 2 i ω + ω ζ (4.5) man man man s Details about the guidance can be found in [NLR, 2002], [Hoffman et al. 2003] Non-Linear ( Bang-bang ) Guidance Time and speed are inherently related therefore to meet simultaneous time and speed constraints a better performing guidance law could take the coupling into account. Speed-time graph analysis. The problem of meeting a simultaneous time and speed constraint at a waypoint can be more generally analysed using a speed-time graph (Figure 5), where speed is the groundspeed. 10 Project AGC-Z-FR EEC Report No. 391

25 COSPACE 2003 Aircraft Guidance Model Based Experiments EUROCONTROL The start point, end point and area under the graph are completely defined by the constraints: initial speed, initial time, final speed, final time and distance to go to waypoint (area under graph). final_speed distance_to_go = speed(t)dt (4.6) initial_speed The problem is reduced to finding the shape of the speed profile within the bounds of operational feasibility. Figure 5: A speed-time graph Maximum speed, minimum speed, acceleration and deceleration must be within the normal flight operating envelope and use of speed targets should be consistent with normal speed guidance e.g. constant CAS descent. A bang-bang controller was designed to provide the following speed profile characteristics (see Figure 6): Figure 6: Speed-time graph of bang-bang control: demand and response Depending on whether the trailing aircraft is late/early, an expedite acceleration/deceleration to a constant maximum/minimum speed was demanded for a certain duration followed by an expedite deceleration/acceleration to the final speed. The duration of the intermediate speed target was calculated from the given constraints using the formula: step_duration = D0 final_speed final_time speed_limit final_speed (4.7) Project AGC-Z-FR EEC Report No

26 EUROCONTROL COSPACE 2003 Aircraft Guidance Model Based Experiments Where D o is distance to go to the waypoint and speed_limit can be a maximum or minimum speed depending on whether the trailing aircraft is late or early. The formula is based on the area of the two rectangles under the target speed profile as an approximation to the actual distance (area under actual speed profile). The area calculation is simple and the errors in lag between acceleration and deceleration tend to cancel. Maximum and minimum speeds were chosen for expedite manoeuvres and assumed fixed for the duration of the capture Lead Groundspeed Based Guidance A third spacing guidance law was derived using the criteria in section 4.1.1: CAS traildesired = GS CAS conversion ( GStraildesired ) (4.8) where the desired CAS of the trailing aircraft CAS traildesired is derived by converting the desired groundspeed of the trailing aircraft GS traildesired to the equivalent CAS. The desired groundspeed of the trailing aircraft is based on the current groundspeed of the lead aircraft GS lead plus a corrective speed term derived from the spacing distance error d error divided by the time to go before the lead aircraft arrives close to the merging waypoint. d t leadtogo error traildesired = GSlead t (4.9) leadtogo GS + The spacing distance error is defined as the difference between the distance of the trailing aircraft to the merge waypoint dtrail, the distance of the lead aircraft to the merge waypoint dlead and the desired spacing distance : d spacing d error = d d d (4.10) trail lead spacing t leadtogo is given by: If dlead 2d spacing then t else leadtogo dlead dspacing = (4.11) GS lead d t leadtogo = GS spacing lead (4.12) 12 Project AGC-Z-FR EEC Report No. 391

27 COSPACE 2003 Aircraft Guidance Model Based Experiments EUROCONTROL 4.2. REMAIN APPLICATION Quality Criteria The following criteria were considered when designing remain spacing guidance law: i) Spacing error between two consecutive aircraft should be small. ii) iii) iv) The guidance should be robust against initial spacing errors or speed changes. All responses should exhibit well-damped dynamic behaviour. The speed profile of the trailing aircraft should be smooth with minimal speed deviations because it may itself be a target lead for another trailing aircraft behind Linear Guidance Spacing Criteria Again, pre-defined along-track spacing criteria were employed to characterize the desired spacing. To distinguish them from the Merge criteria y lead and y trail were used here as notation for alongtrack distances. Constant Distance (CD) Spacing Criterion: This criterion was based on the spacing distance between two aircraft. The spacing error (y errorcd ) was calculated from the difference between the lead aircraft s position (y lead ) and the trailing aircraft position (y trail ) taking into account the desired spacing (y spacing ): y errorcd = y lead y trail y spacing (4.13) Where positions were measured by along track distance from a common origin. Constant Time Delay Criterion (CTD): The CTD criterion was based on constant time spacing. Spacing error was defined as the difference between the elapsed time since the lead aircraft overflew the current trailing aircraft position, and the desired time spacing: t errorctd * * = t t ylead ( t ) = ytrail ( t ) t spacing (4.14) The CTD criterion was rewritten as an equivalent constant distance-based criterion: y errorctd ( t) = y ( t t ) y ( t) lead spacing trail (4.15) The re-formulation of the CTD spacing criterion as a constant distance criterion was used in the design and implementation of the guidance law controller. Approximate Constant Time Delay (ACTD) Spacing Criterion: The ACTD spacing criterion was based on the CTD spacing criterion, which, assuming the lead aircraft speed (V leadtas ) was constant, was approximated as: t erroractd = ylead y V leadtas trail t spacing (4.16) Project AGC-Z-FR EEC Report No

28 EUROCONTROL COSPACE 2003 Aircraft Guidance Model Based Experiments The ACTD spacing criterion was rewritten as an equivalent constant distance-based criterion: y erroractd = y y y (4.17) lead trail distactd where: y distactd = V t (4.18) lead TAS spacing The re-formulation of the ACTD spacing criterion as a constant distance criterion was used in the design and implementation of the guidance law controller Spacing Guidance The linear guidance law defined in sub-section was used: V cmd = K p y error 2 2 ( ) ( Ki + s) s + 2 ω s + ω ζ (4.19) man man man s Transmission Loss Compensation The basic guidance law presented above contains no compensation for imperfect ADS-B messages but was used to find the minimum level of performance. The following compensation terms were then added to the guidance law for update rate, latency and accuracy to cope with a realistic imperfect ADS-B transmission model. Update Rate In the basic guidance law, the lead position was assumed constant between ADS-B updates. An update rate compensation term was designed containing the following extrapolation, to better approximate the lead s position between consecutive ADS-B updates: ylead _ predicted ( t + τ ) = ylead ( t) + V ( t) τ (4.20) lead TAS where τ was the time between two updates. The above was found to be a good estimate of the lead aircraft intermediate trajectory when the true air speed (TAS) was varying slowly. Latency Compensation was added for only the constant delay component of latency. The constant delay component could be calculated if the time measurement was assumed to be included in the ADS-B report. To do this, the guidance law was slightly adjusted: the control error term contained a third parameter which was a correction depending on the amount of constant latency (t lat ). y lead _ ( t) = y ( t) + V ( t) t (4.21) adjusted lead lead TAS lat Accuracy Accuracy of the surveillance information was modelled using white noise characterised by a mean and deviation. A first order Butterworth filter was added to the speed to compensate for lack of accuracy and hence improve the constant time delay spacing performance. The band pass of this filter was between 0.01 rad/s and 1 rad/s and was chosen so that the bandwidth of air speed control was centred in this interval. 14 Project AGC-Z-FR EEC Report No. 391

29 COSPACE 2003 Aircraft Guidance Model Based Experiments EUROCONTROL 5. APPARATUS 5.1. FAST-TIME SIMULATION ENVIRONMENT All aircraft models, spacing guidance, wind and atmosphere models were simulated in MATLAB (version 6, release 12) and Simulink (version 4, release 12) for Windows/Personal Computer (PC). All the fast-time simulations were performed on a Pentium IV based PC (1.5 GigaHertz, with 256 MegaBytes internal Random Access Memory (RAM)). MATLAB is a high-performance language tool for technical computing. It integrates programming, computation and visualisation in an easy-to-use environment. MATLAB consists of five main parts: The MATLAB language (a high level language for technical computing). The MATLAB working environment (set of tools to facilitate the work of user or programmer). Handle graphics (high-level commands for 2-D and 3-D data visualization). The MATLAB mathematical function library (large collection of computational algorithms). The MATLAB Application Program Interface (allows interaction with C and Fortran). Simulink is a widely used software package for modelling, simulating and analysing dynamic systems. It supports linear and non-linear systems, modelled in continuous or sampled time. Simulink provides a Graphical User Interface (GUI) for building models as block diagrams, using click-and-drag mouse operations. MATLAB and Simulink are integrated so the models can be simulated and revised in either environment at any point ADS-B MODEL Each leading aircraft sends a subset of ADS-B surveillance information: position, velocity, position accuracy, velocity accuracy (RTCA Navigation Accuracy Categories) and a time of measurement. This state vector information was transmitted through ADS-B reports. The ADS-B model was composed of an ADS-B transmitter model and ADS-B receiver model. To simulate more realistic (i.e. imperfect) ADS-B transmissions, the following characteristics were modelled: Update rate of reports: initially, a perfect update rate was considered, i.e. the probability of reception was 100 %, then probabilities of reception less than 100 % were considered. Latency of transmission: assumed to be the same for all information. Furthermore, this latency consisted of a mean time delay with a stochastic variation (to model jitter). The standard deviation of this stochastic variation determined the amount of jitter. Accuracy of surveillance information: accuracy of the horizontal position and velocity was characterised by gaussian distributions with zero means and respective standard deviations sigma σ hp and σ hv (Table 2). Project AGC-Z-FR EEC Report No

30 EUROCONTROL COSPACE 2003 Aircraft Guidance Model Based Experiments Table 2: Navigation Accuracy Categories (NAC)- Position and Velocity (RTCA, 2002) NAC for Position Horizontal. Position Error (2σ hp 95 %) NAC for Velocity Horizontal Velocity Error (2σ hv 95 %) 0 Unknown 0 Unknown 1 < 10 NM 1 < 10 m/s 2 < 3 m/s 7 < 0.1 NM 3 < 1 m/s 8 < 0.05 NM 4 < 0.3 m/s 9 < 30 m 10 < 10 m 11 < 3 m 5.3. AIRCRAFT MODEL For the purpose of this study, a model was required with realistic behaviour of an aircraft along typical descent profiles, including speed changes and intermediate altitude steps. The aircraft model was composed of an aircraft dynamics model, autopilot model, spacing director model and human pilot model: Aircraft Dynamics Model For the aircraft dynamics the following general assumptions were made: Flat, non-rotating earth. Standard atmosphere. Fully co-ordinated flight. It was assumed that the sideslip angle β was always zero and there was no side force. The equations of motion used for the aircraft model were based on the three-dimensional pointmass differential equations, as found in many references [McGruer, 1973], [Hanke and Norwall, 1995]. The total set of differential equations resulted in 7 state variables, [γ V h ϕ ψ x east x north ], where: γ is the flight path angle, V the TAS, h vertical distance or altitude, ϕ is the bank angle, ψ the heading angle, x east the east position and x north the north position and m the aircraft mass. Because the aircraft mass was not considered to be constant, the equations of motion were complemented by an eighth equation, describing the loss of mass due to fuel flow (Q) of the aircraft. The final set of equations was: L + T sin α g & γ = cos ϕ cos γ (5.1) m V V T cosα D V & = g sinγ (5.2) m h & = V sin γ (5.3) 16 Project AGC-Z-FR EEC Report No. 391

31 COSPACE 2003 Aircraft Guidance Model Based Experiments EUROCONTROL ϕ& = p (5.4) g tan ϕ ψ = V & (5.5) x& = V cosγ cosψ cos χ (5.6) east V wind north V wind wind x& = V cosγ sinψ sin χ (5.7) wind m& = Q (5.8) Here, D is the drag, T the engine thrust, α angle of attack, χ wind and V wind are the wind direction and speed, L is the lift, p is the roll rate and g is gravity. Only normal flight regimes were considered in this study, therefore α was relatively small, and in (5.2) cosα was approximated to 1. In (5.1), the term T sinα was considered as negligible in comparison with the lift contribution. This simplified (5.1) and (5.2) to: L g & γ = cosϕ cosγ (5.9) m V V T D V & = g sin γ (5.10) m Differential equations (5.3) to (5.10) constituted the basic equations of motion of the aircraft model. The aerodynamic forces were based on an aerodynamic model, using an estimate of the aircraft trimmed aircraft polar, with an extension to model the effects of Mach-drag rise. The Mach-drag rise component was a function of Mach number and lift coefficient. A 2-dimensional look-up table was used to model the aircraft polar. The thrust was computed from a given thrust to weight ratio for a given aircraft, by multiplying this ratio by a percentage thrust command and the maximum take-off mass of the aircraft type. The thrust to weight ratio was calculated from a two-dimensional look-up table, as a function of Mach and pressure altitude. The thrust characteristics used in the model were typical for high by-pass turbofan aircraft. Due to the fact that the thrust was calculated as a dimensionless thrust to weight ratio, the thrust model could be adapted easily to various aircraft types, without significant changes to the thrust model. By using a calibration factor (ranging from plus or minus 20 %) the model could easily be adapted to any aircraft type Autopilot Model The autopilot allowed the aircraft to follow the reference targets (desired air speed and altitude). The principle used to design the autopilot was based on the total energy rate [Lambregts, 1983], [Hoffman, 1993]. It is beyond the scope of the present paper to go into the details of the actual implementation of the control system. The tuning of the parameters of the pilot model and the validation of the overall resulting trajectories were performed using two references: (i) a fixed base cockpit simulator at the Eurocontrol Experimental Centre, based on a high fidelity 6 degree of freedom Boeing 747 and Airbus A320 aircraft models; (ii) at the National Aerospace Laboratory of the Netherlands a high fidelity 6 degree of freedom Fokker 100 simulator (see Annex). Project AGC-Z-FR EEC Report No

32 EUROCONTROL COSPACE 2003 Aircraft Guidance Model Based Experiments Spacing Director The desired CAS of the trailing aircraft CAS traildesired is passed though a quantiser to reduce the variations before presentation to the pilot model. This spacing director logic takes the form: if CAS out t 1 CAS in t 2knots then CAS () t CAS () t out = else CASout () t = CASout ( t 1) in (5.11) (5.12) The final output is rounded to the nearest knot before presentation to the pilot model Human Pilot Model Pilot reaction to the above demand from the spacing director ( CAS traildemand ) is modelled by a 3 s constant time delay. This value has been compared with pilots reaction time in real-time simulations. Figure 7: Closed loop control law system block diagram with human pilot model 5.4. ATMOSPHERE AND WIND MODELS The basis for the atmosphere and air data model is an implementation of the International Standard Atmsophere (ISA) model. The Simulink model calculates static temperature, speed of sound, static air pressure and air density for a given altitude under ISA conditions, but also for deviations from ISA conditions in terms of temperature (covering at least delta ISA temperature from -10 to + 30 C) and static pressure. The wind model was based on that of the Joint Aviation Requirements All Weather Operations (JAR-AWO) autoland certification process [JAR-AWO, 1996]. Figure 8 presents the wind model as implemented in MATLAB. The turbulence provided has a gaussian distribution, conforming to the so-called Dryden spectrum. The turbulence provides disturbances of the true airspeed and angleof-attack. 18 Project AGC-Z-FR EEC Report No. 391

33 COSPACE 2003 Aircraft Guidance Model Based Experiments EUROCONTROL Figure 8: The wind model components Altitude (feet) Mean wind speed Turbulence Wind speed (knots) Figure 9: Turbulent wind versus altitude - JAR-AWO model (20 knots wind at 30 feet AGL) In this model the mean wind speed was altitude dependent, and directly associated with the wind as measured at 30 feet AGL (Above Ground Level). The mean wind speed determined the turbulence intensity, and the wind speed increased with altitude (Figure 9). The magnitude of the mean wind speed increasing with altitude is defined by the following expression: V mean ( h) = V 30 h (5.13) Project AGC-Z-FR EEC Report No

34 EUROCONTROL COSPACE 2003 Aircraft Guidance Model Based Experiments where Vmean is the mean wind speed (knots) measured at h metres AGL and V30 is the mean wind speed (knots) at 30 feet AGL. Table 3 summarises different values for the mean wind speed varying with altitude for a given initial value measured at 30 feet AGL. Table 3: Mean wind speed increasing with altitude and function of wind measured at 30 feet AGL Altitude AGL (feet) Mean Wind Speed (knots) , , , , , , Project AGC-Z-FR EEC Report No. 391

35 COSPACE 2003 Aircraft Guidance Model Based Experiments EUROCONTROL 6. METHOD This section describes the evaluation methods used to perform the experiments. For each experiment, an appropriate method was used, involving three aspects: an operational scenario, specific metrics and experimental parameters. The main target spacing parameter was based on a distance or time comfortably greater than the separation standard e.g. 7 or 8 NM target spacing for a 5 NM separation standard MERGE EXPERIMENTS Comparison of Linear and Non-Linear Guidance Operational Scenarios To compare the three guidance designs: (i) Linear Distance-Based Prediction Guidance (section ); (ii) Linear Speed-Based Prediction Guidance (section ); (iii) Non-linear Bang-Bang Guidance (section 4.1.3). the following scenarios were simulated: Table 4: Initial conditions for comparing scenarios Initial Condition Scenario 1 Scenario 2 Scenario 3 Lead initial speed (knots CAS) Trailing initial speed (knots CAS) Initial distance between lead and the waypoint (NM) Initial distance between trailing and the waypoint (NM) Cause of initial error Trailing aircraft late Trailing aircraft early, high relative speed Trailing aircraft late, low relative speed In all scenarios, the lead aircraft was flying at constant altitude FL290 and constant speed 272 knots CAS. The desired spacing between aircraft at the merge waypoint was 7 NM. For the bangbang controller the maximum and minimum speeds were fixed at 330 knots CAS (~490 knots TAS) and 230 knots CAS (~357 knots TAS) Metrics The following metrics were used to compare the performances of the three guidance laws: Spacing distance error in NM at the waypoint. Groundspeed difference in knots between aircraft at the merge waypoint. Project AGC-Z-FR EEC Report No

36 EUROCONTROL COSPACE 2003 Aircraft Guidance Model Based Experiments Evaluation of Lead Groundspeed Based Guidance and Spacing Director Operational Scenario To validate the guidance law the following scenario was simulated: The lead and trailing aircraft were in descent from 29,000 feet to 3,000 feet, flying direct to the same fixed waypoint. The lead aircraft track angle was due North and the trailing aircraft track angle -25 (i.e. pointing ~North North West towards the merge waypoint). At the initial time the lead aircraft was at 80 NM from the merge waypoint and constant speed 272 knots CAS (7 NM/minute). The trailing aircraft was initialised at 96 NM from the merge waypoint with a speed of 232 knots CAS. The scenario was conducted under realistic atmospheric conditions with turbulent cross winds from -45 (i.e. from the North-East), of mean strength varying from 20 knots at 30 feet altitude up to 50 knots at 29,000 feet. Note on lateral guidance For the above and following Merge scenarios the aircraft were assumed to fly directly over the merge waypoint i.e. there was no turn anticipation of the merge waypoint Metrics The same metrics as above were used to evaluate the performance of the guidance law. The intermediate speed profiles were also checked for operational acceptability as detailed in the four guidance design criteria (see section Spacing Guidance Design ) Effect of Entry Conditions Findings from Real-Time Experiments From Co-Space real time simulations it was observed that air traffic controllers frequently used heading instructions to spread aircraft out on diverging stretched paths (creating spacing) before merging them again on converging tracks. This often led to situations where the Merge instruction was used for several consecutive aircraft on multiple converging tracks. Figure 10 shows a common case where three aircraft have to merge to the same waypoint from three different directions. In this case the first trailing aircraft (Aircraft N 2) becomes a lead for the second trailing aircraft (Aircraft N 3). Waypoint Desired Spacing between N 1 and N 2 Aircraft N 1 Aircraft N 2 Aircraft N 3 Figure 10: Merge example three aircraft 22 Project AGC-Z-FR EEC Report No. 391

37 COSPACE 2003 Aircraft Guidance Model Based Experiments EUROCONTROL Operational Scenarios Three aircraft descended along different converging trajectories (aircraft tracks for N s 1 and 2 as above plus aircraft N 3 track at -45 i.e. from South East) from 29,000 feet to 3,000 feet to the same fixed merge waypoint. The operational scenarios were conducted under realistic atmospheric conditions with cross winds of 20 knots at 30 feet altitude and 50 knots at 29,000 feet altitude and medium turbulence. At 25,000 feet, a speed change was commanded for the lead profile from 272 knots CAS to 232 knots CAS Metrics A given set of entry conditions was deemed valid if they resulted in both the following two criteria being met: 1. Spacing criterion: Spacing distance error at the waypoint had to be less than ±0.4 NM (i.e. 5 % of spacing distance error). See spacing guidance design criterion (i). This value is consistent with pilots in real time simulations being able to manually maintain spacing within 0.5 NM. 2. Speed criterion: Groundspeed difference between two consecutive aircraft at the waypoint had to be less than 10 knots. See spacing guidance design criterion (ii). Too close Admissible entries Lead Too far Figure 11: Merge example admissible entries of trail Experimental Parameters The following parameters were varied as follows. For a sequence of two aircraft: 1. Initial aircraft N 1 distance to go to the waypoint was varied between the values [0, 10, 20, 40, 70, 100] NM. 2. For each initial distance of aircraft N 1 above, initial spacing error of aircraft N 2 relative to aircraft N 1 was varied to find lower and upper bounds somewhere in the range 22 to +25 NM. 3. For the following initial distances of aircraft N 1 [20, 40, 70, 100] NM, initial spacing error and speed of aircraft N 2 relative to N 1 were varied to find lower and upper bounds in the ranges 24 to +26 NM and 60 to +50 knots respectively. Project AGC-Z-FR EEC Report No

38 EUROCONTROL COSPACE 2003 Aircraft Guidance Model Based Experiments For a sequence of three aircraft: 1. Initial aircraft N 1 distance to go to the waypoint was varied between the values [40, 70, 100] NM. 2. For each initial distance of aircraft N 1 above, initial spacing errors of aircraft N 2 relative to aircraft N 1, and aircraft N 3 relative to aircraft N 2 were varied to find lower and upper bounds. These bounds were within the ranges -22 to +25 NM (aircraft N 2) and -31 to +30 NM (aircraft N 3). Due to the use of a turbulent wind model, a large number of random trials was employed to find the lower and upper bounds. Table 5: The wind parameters Wind Conditions Speed Range Direction Type 0 to 87 knots at 30 feet (0 to 235 knots at 29,000 feet) Cross from West Head Tail Constant Turbulent A Boeing 747 model (initial mass 271,472 kg) was used for the heavy aircraft type and a Fokker 100 model (initial mass 37,919 kg) for the light aircraft type REMAIN EXPERIMENTS Effect of ADS-B Transmission Quality Test Scenario The following operational scenario was used for all trials. A lead aircraft followed its own planned profile and a sequence of five trailing aircraft of the same type adjusted their own speed to maintain the desired spacing to each respective preceding aircraft. All aircraft started at FL290 and 7 NM/minute TAS (272 knots CAS) and descended to FL100. All aircraft started their descent at the same location. The lead aircraft reduced speed from 272 knots to 230 knots (CAS) starting at FL150. Desired spacings between aircraft were 7 NM (for CD) and 60 s (for ACTD). At initialisation, all aircraft (Boeing ) were at the desired spacing, as well as at the same speed and altitude. Wind was zero for all altitudes Metrics Three indicators were used to assess the performance: the spacing (distance or time) between aircraft, the speed variation for each aircraft, and the throttle activity. 24 Project AGC-Z-FR EEC Report No. 391

39 COSPACE 2003 Aircraft Guidance Model Based Experiments EUROCONTROL Experimental Parameters Spacing Criteria Each test scenario was run using both CD (7 NM) and ACTD (60 s) spacing criteria. Guidance Law Compensation Each test scenario (with the exception of update rate with probability of reception) was run using basic guidance and compensated guidance laws. Update Rate The above test scenario was run for several different update rates, with 100 % probability of ADS-B message reception, from an update period of 0.1 s until the spacing metrics exhibited unstable behaviour (Table 6). The trials were repeated for probability of reception rates less than one until the spacing metrics exhibited unstable behaviour. Latency The above test scenario was run for several different constant delays increasing step-wise from 0 s until the spacing metrics exhibited unstable behaviour (Table 6). The trials were repeated for different latencies with jitter. For each trial, a maximum admissible jitter was considered. Accuracy The above test scenario was run for several different horizontal position and velocity accuracies increasing step-wise from 3 m and 0.3 m/s respectively until the spacing metrics exhibited unstable behaviour (Table 6). Note that update rate with probability of reception, latency with jitter and accuracy involved many trials therefore Monte-Carlo techniques were employed to identify those trials with the minimum spacing performance. Table 6: ADS-B experimental parameters and perturbation ranges Range Constant Period Update Period Latency Accuracy Probability of Reception Constant Delay Maximum Jitter Delay Horizontal Position Horizontal Velocity Minimum 0.1 s 0.1 to s 0.0 s < 3.0 m < 0.3 m/s Maximum 30.0 s 0.1 to s < 24.0 s < 0.1 NM < 1.0 m/s Evaluation of Time-Based Spacing Criteria Operational Scenario The following scenario was used: a lead aircraft followed its own descent profile and the trailing aircraft adjusted speed to maintain the desired time spacing (60 s). Both aircraft started at FL290 and 7NM/minute true airspeed (TAS) (272 knots CAS) and descended to FL100. The aircraft started their descent at the same location after 10 NM of flight (fixed top of descent). After 5 minutes the lead aircraft reduced speed from 272 knots to 230 knots. At the initial time, the aircraft (Boeing ) were at the desired spacing. Wind speed was zero for all altitudes Metrics Two indicators were used to assess the differences between the CTD and ACTD criteria: speed behaviour and spacing error. Project AGC-Z-FR EEC Report No

40 EUROCONTROL COSPACE 2003 Aircraft Guidance Model Based Experiments Parameters The experimental parameters under investigation were exact CTD spacing criterion and approximate CTD spacing criterion Effect of Mixed Aircraft Types and Varying Wind Conditions The following tests involved a pair of aircraft: a trailing aircraft following a lead aircraft in descent Operational Scenario The lead aircraft followed a predefined flight plan from cruise to descent and the trailing aircraft adjusted speed to maintain the desired time spacing (60 s). The aircraft tracks were due North. Both aircraft started at an altitude of 29,000 feet and 7 NM/minute true airspeed (272 knots CAS). The aircraft started their descent to 3,000 feet at the same location (10 NM from start at fixed top of descent). At 25,000 feet the lead aircraft reduced CAS from 272 knots to 242 knots. At 15,000 feet the lead aircraft performed a second CAS reduction from 242 knots to 212 knots. The aircraft were initialised at the desired spacing. After the first 5 NM of flight without wind, a JAR-AWO wind was introduced Metrics Constant wind One trial was run for each combination of wind speed and direction. The maximum time delay spacing error encountered during each trial was recorded and used as a metric. Turbulent wind 500 trials were run for each combination of wind speed and direction. The maximum time delay spacing error encountered during each trial was recorded. The largest value and mean value of these 500 maximum spacing errors were used as metrics Experimental Parameters A mixture of aircraft types and wind conditions were varied as indicated in Table 7: Aircraft Type Mix Wind Conditions Table 7: Mixed aircraft types and winds: experimental parameters Parameters Values Lead Heavy Light Heavy Light Trailer Heavy Heavy Light Light Speed Range 0 to 87 knots at 30 feet (0 to 235 knots at 29,000 feet) Cross from West Direction Head Tail Type Constant Turbulent A Boeing 747 model (initial mass 271,472 kg) was used for the heavy aircraft type and a Fokker 100 model (initial mass 37,919 kg) for the light aircraft type. Results for wind with turbulence involved Monte Carlo techniques to identify those with the minimum spacing performance. 26 Project AGC-Z-FR EEC Report No. 391

41 COSPACE 2003 Aircraft Guidance Model Based Experiments EUROCONTROL 7. RESULTS 7.1. MERGE EXPERIMENTS Comparison of Linear and Non-Linear Guidance Results of the fast-time simulations of the three guidance laws ( Distance based, Speed based and Bang-Bang ) for the three different scenarios described in section are shown in Figure 12 and Figure 13 show the absolute values of spacing error at the merge waypoint. Figure 12: Comparison of absolute distance spacing error at the merge waypoint Figure 12 shows the absolute values of the speed difference between aircraft at the merge waypoint. Figure 13: Comparison of absolute magnitude of groundspeed difference at the merge waypoint These results could be explained by considering merging at the waypoint as meeting a time and speed constraint. The linear distance based guidance laws appear not to recognize these constraints, and these guidance laws can perform satisfactorily only for ideal initial conditions (i.e. no initial spacing error). The non-linear bang-bang guidance took into account both time and speed constraints, and achieved the desired spacing with minor errors, therefore the first two criteria i) and ii) in paragraph were met. However, the non-linear guidance was based on an area under speed-time graph computation (Figure 6). When the aircraft was flying at constant altitude, Project AGC-Z-FR EEC Report No