AIAA Infotech@Aerospace Conference <br>and<br>aiaa Unmanned...Unlimited Conference 6-9 April 2009, Seattle, Washington AIAA 2009-1864 Trajectory Assessment Support for Air Traffic Control G.J.M. Koeners 1 Delft University of Technology, Delft, 2600GA, The Netherlands Future Air Traffic Management (ATM) systems may lead to a shift from the state based operation that is used today to a trajectory based operation where automation generates new trajectories to resolve conflicts and pilots can sent requests in the form of trajectories. If all aircraft participate in the trajectory based operation and all possible situations are taken into account, monitoring and re-planning the trajectories could be automated. However in the near future not all aircraft will be capable of trajectory based operations and it will be difficult to design an automated system that can cope with all situations. Furthermore a human will probably always remain responsible. As a consequence a controller in a (semi) trajectory based operation must be able to monitor the automation and in unforeseen situations come up with a solution. To be able to perform these tasks, the controller needs tools to asses the different trajectories to detect unforeseen unsafe situations and subsequently generate a solution. The paper describes an experiment to get data on the usability of current tools to asses a new trajectory and to explore the potential benefits of new tools that would make the separation margin more explicit. I I. Introduction N earlier research, it has been shown that air traffic controllers can benefit from support tools when performing their conflict detection and resolution task 1,2,3. The support tools used in this research vary from alerting the controller when a conflict is detected to automatically generating a conflict resolution. In the latter case the automation proposes a single maneuver to be executed by an aircraft to resolve the upcoming conflict 1.The air traffic controller has to evaluate the new trajectory using his knowledge and experience before communicating the solution to the pilot. A similar process of trajectory evaluation is expected when in the future controllers will receive flightplan requests by pilots 2. In this concept pilots use onboard systems to generate a conflict free path optimized for their flight and send it by datalink to air traffic control. The controller evaluates the requested trajectory and accepts, rejects or modifies the request. In Ref. 3, an experiment is described in which the trajectory negotiation process between pilot and controller was evaluated. It showed that when pilots requested a flight path change which was conflict-free according to the flight deck conflict detection and resolution system, controllers still rejected some of these trajectories. A possible explanation is that controllers use their own heuristics to assess whether a trajectory will be safe or not; as stated in Ref. 3, a safe route by controller standards may require more than missing other aircraft by the minimum separation requirement. This brings up the question whether the air traffic controllers reject conflict free trajectories because using their own heuristics they want to maintain a larger separation margin then strictly required or that the tools currently used will not give them the insight in the separation margins a new trajectory actually has. A. Trajectory Based Operations The earlier mentioned future vision where automation generates new trajectories to resolve conflicts and where pilots send requests in the form of trajectories leads to a shift from the state based operation that is used today to a trajectory based operation. An important difference between the two operations is predictability. With state based operations the current state of aircraft is used to predict the future state. This information is used by the air traffic controller to predict future conflicts. If a future conflict is detected, the current state of the involving aircraft is changed (e.g. change altitude). In a trajectory based operation not only the current state is known but also the trajectory that the aircraft intends to fly. If time information is included, this 4D trajectory describes where an aircraft plans to be as a function of time. The aircraft will use the trajectory as a control reference 4 and as a result the future location of the aircraft is not a prediction but an agreement. In contrast to prediction where the future position 1 Researcher, Electronic Navigation Systems Group, AIAA Member. 1 Copyright 2009 by the, Inc. All rights reserved.
is based on an extrapolation of the current state, now the future position is computed from the trajectory and the future time. As a result, the error between where the aircraft will actually be and is computed to be becomes independent of look-ahead time. If the trajectories of all the aircraft are known, future conflicts can be detected by testing for 4D intersections and resolved by changing trajectories. This change of trajectory is also needed if disturbances prevent an aircraft from following the 4D trajectory. B. Role of the Controller If all aircraft participate in the trajectory based operation and all possible situations are taken into account, monitoring and re-planning the trajectories could be automated. However in the near future not all aircraft will be capable of trajectory based operations and it will be difficult to design an automated system that can cope with all situations. Furthermore a human will probably always remain responsible. As a consequence a controller in a (semi) trajectory based operation must be able to monitor the automation and in unforeseen situations come up with a solution. To be able to perform these tasks, the controller needs tools to asses the different trajectories to detect unforeseen unsafe situations and subsequently generate a solution. These tools should minimize the chance of a misdetection but at the same time do not cause unnecessary changes to the trajectories. As mentioned before in earlier studies 3 controllers rejected trajectories that according to the automation were conflict free. The question is if the unnecessary interventions are a result of the earlier mentioned intentional extra separation preferred by the controllers or the inability to asses the trajectory s separation, the effect of a trajectory change and the associated margins. The answer to this question partly depends on which tools are currently available. II. Trajectory Assessment Tools A. Current Tools As a case study we looked at the tools currently available in the Royal Netherlands Air Force military air traffic control system. Military air traffic controllers perform conflict detection by estimating whether extrapolated aircraft tracks will result in a loss of separation. The air traffic control system functionality is able to extrapolate aircraft tracks to selectable time horizons, ranging from 1 to 5 minutes, thus providing support for the conflict prediction task. It is also possible to draw a bearing line between two aircraft or from an aircraft toward a fixed location. This bearing line can assist the controller with his task of conflict detection; when the bearing line between two aircraft remains in the same direction, a conflict may occur. Furthermore a tool can be used to calculate the estimated time of arrival. Based on current speed and a selectable location, the tool will calculate in how many minutes the selected aircraft will be at the selected location. Controllers can use flightplan information to estimate the intended route. Besides the given speed of an aircraft, history dots give a graphical indication of the speed of the different aircraft. The history dots show the last three positions picked up by radar. Finally a distance measurement tool is available. In the current system, no support tools are implemented to assist the air traffic controller with his task of conflict resolution. B. New Tools New tools can assist the controller in a trajectory based environment with the detection of unforeseen unsafe situations and aid the controller with the generation of a solution. When performing the detection task, the controller has to estimate if aircraft maintain enough separation, just as today s operation. However, the current tools are designed to assist the controller in a state based operation and might not be suited to evaluate a set of trajectories. To assist the controller, a new tool could compute and display the minimal distance. If the trajectories of the aircraft are known, the reliability of this calculation is increased. This tool, that shows the separation at the Closest Point of Approach (CPA), is already implemented in some ATC systems 5. 2
If a future conflict is detected the controller has to come up with a solution. Many studies have looked at automation that generates new conflict free trajectories 6-9. A parameter that is rarely used to solve a conflict is speed. Because of uncertainties in a state based operation, conflicts are solved within a short time frame. Since typically the speed range of an aircraft is limited, the control margin to change the separation within this short time frame is not sufficient. In a trajectory based operation the predictability is increased and future conflicts can be detected earlier. This might lead to a situation where the time to conflict is long enough so that the control margin of a speed change will be sufficient to solve the conflict. Solving the conflict by only changing speeds is an advantage in operations where a trajectory change is not desirable. A tool could generate a speed advisory to solve the conflict, however the solution is a balance between the extra separation needed and the available control margin. This margin can be affected by for example aircraft performance. If this can be made explicit, the controller has the final decision in this balance. A new tool was developed to show the required speed change and available control margin. In Fig. 1 the speed information for one aircraft is displayed. The white circle indicates the CPA with the current speed. The magenta circles indicate the two speeds where the CPA has a separation of the required 5 nautical mile (NM). The speed range between the two magenta circles will result in a CPA with a separation below 5 NM. The green bar indicates the available control margin. Since the current CPA (the white circle) is in between the two magenta circles, the separation is below 5 NM. The controller can solve the conflict by modifying the speed. In this case the new speed must be either below 182 knots or above 285 knots. The green bar indicates that a speed of 182 knots or lower is not possible and that only a speed of 285 knots or higher is a possible solution. The amount of control margin that is used to increase the separation is now up to the controller. A trial speed can be used to evaluate a possible solution. The yellow circle indicates the CPA if the aircraft would fly this trial speed. The separation at this CPA is indicated in yellow text, next to the yellow circle. In this case a speed of 300 knots will give a CPA with a separation of 6.2 NM. Since this speed is located within the green bar, the aircraft is able to fly this speed and is therefore a valid solution. To get data on the usability of current tools to asses a new trajectory and to explore the potential benefits of new tools that would make the separation margin more explicit, an experiment was conducted. The subjects had to evaluate a new trajectory in a simulated air traffic environment. In the baseline condition the subjects used current air traffic control tools to asses the new trajectory. In other conditions they used the new trajectory evaluation tool that was developed and showed the separation margins the new trajectory had with other traffic. The amount of separation the new trajectory had was varied. III. Experiment Setup Figure 1. New speed and separation evaluation tool. The first goal of the experiment was to evaluate the usability of the current- and new tools in assessing a new trajectory. The traffic situations that the controllers had to evaluate consisted of two aircrafts with a known trajectory. The two trajectories always had a 3D conflict. Depending on the speed of the aircrafts the 3D conflict could result in a loss of separation (distance between aircraft below 5 NM). The controllers were instructed that one of the two trajectories was a new trajectory and that their task was to asses this new trajectory using the available tools. After this assessment the controllers had to decide if they would accept this new trajectory. A. Conditions The geometry of the conflicting trajectories can have an influence on the usability of some of the trajectory assessment tools. To simplify the experiment the vertical dimension was not taken into account. Now a contributing factor to the usability of assessment tools could be the existence of turns in the trajectories before the conflicting crossing. To obtain a finite number of geometries that can be used in the experiment the following three geometries were defined: 3
Figure 2. Different geometries used in experiment a) No turns in either trajectory (Fig. 2A) b) One turn in one of the trajectories before the crossing (Fig. 2B) c) One turn in both trajectories before the crossing (Fig. 2C) Every run the controllers were confronted with one of the three conflict geometries. To simulate the process of receiving a new trajectory, one of the aircrafts and its trajectory appeared on screen 10 seconds after the session had started. The minimal distance the two aircraft had was varied. The required separation the controllers had to maintain was 5 NM. To get enough sample points but at the same time limit the number of runs a limited set of minimal separations was defined: 2.5 NM 5 NM 7.5 NM 10 NM 15 NM The different separations were realized by varying the speed of the aircraft with the new trajectory. To evaluate the different tools that support the controller with the trajectory evaluation task. The following conditions were defined: 1) Current tools 2) Separation at CPA as a number 3) Separation at CPA as a number + speed evaluation tool All conditions were varied within subjects. This made the total number of trajectory situations that a controller was asked to evaluate: 3 x 5 x 3 = 45. After evaluation the controller could take the following actions: Accept the new trajectory Reject the new trajectory Accept the new trajectory with a different speed The last option was added to achieve the second goal of the experiment: to evaluate the usability of the different tools in generating a new speed that would solve a future loss of separation. Furthermore, it will give an indication of the desired separation when the speed trial tool is used. Two extra scenarios were added to evaluate the effect of providing information about the amount of control margin, in this case the capability to increase the separation by changing the speed. The two scenarios are identical in geometry and separation, only the amount of control margin left to adjust the separation is different. In one scenario the minimal separation is just enough but there is still room to increase the separation by adjusting the speed (Fig. 3). The other scenario is identical, but now there is no room left to increase the separation by adjusting the speed (Fig. 4). Figure 3. Scenario with control margin to increase separation. Figure 4. Scenario without control margin to increase separation. 4
B. Measures An indication of the usability of a tool to evaluate a set of trajectories is the number of unnecessary interventions (false alarms) by the controller. Such a false alarm was recorded if a new trajectory with a separation between 5 NM and 15 NM was rejected or the speed was modified. In the conditions where the new trajectory had a separation with the existing trajectory of 2.5 NM, the controller had to come up with a solution. The proposed solution is an indication of the usability of the tools for generating a solution using speed. The controller could change the speed in such a way that the separation was 5 NM or larger. If the controller changed the speed, but the separation was still below 5 NM, an incorrect speed was recorded. If the controller wasn t able to come up with a new speed, the controller had to reject the new trajectory. An acceptance of a new trajectory that had a separation of 2.5 NM was classified as a missed detection. Figure 5. Research traffic management tool. C. Apparatus The tools currently used and the new tools were integrated in an existing research traffic management tool (Fig. 5). A total of 10 executive controllers from the Royal Netherlands Air Force 711 Military Air Traffic Control Centre (MilATCC) Squadron took part in the experiment. IV. Results A. Detection Figure 6 shows the percentage of false alarms, i.e. trajectory situations where the controller unnecessarily intervened. In these cases the controller took action while the separation of the two trajectories was 5 NM or more. Although the separation was sufficient the controller either rejected the new trajectory or modified the trajectory speed. By adding separation information the number of false alarms was greatly reduced. When a speed evaluation tool was added the number of rejects was further reduced. However the number of unnecessary changes increased compared to only showing the separation information. 5
40% 35% 30% 25% 20% Unn.Change Reject 15% 10% 5% % Current CPA CPA and Speed Figure 6. Percentage of false alarms. Figure 7 shows the trajectory rejections from Fig. 6, but now for the different separations the trajectories in actuality had. When controllers used the current tools, trajectories were rejected in all separation conditions. From 6% in the 15 NM condition up to more then 40% in the 5 NM condition. When using the tool that calculates the separation at the CPA and speed evaluation tool only trajectories were rejected when the separation was 5 NM. 50% 45% 40% 35% 30% 25% 20% 15% 10% 5% % 5 7.5 10 15 Figure 7. Percentage of route rejections. Figure 8 shows the unnecessary speed changes from Fig. 6, but now 50% for the different separation 45% conditions. When controllers used the current tools, the speed was 40% 35% unnecessary modified in all 30% separation conditions. From 6% in the 15 NM condition up to 24% in the 5 NM condition. When using the computed separation at the CPA, 25% 20% 15% 10% 5% unnecessary speed changes where % only made in the 5 NM condition. 5 7.5 10 15 When the speed and separation evaluation tool was used, 10% of the speeds were modified in the 7.5 NM Figure 8. Percentage of unnecessary speed changes. condition and over 45% in the 5 NM condition. Current Tools CPA CPA and Speed Current tools CPA CPA and speed B. Resolution In the scenarios where the separation was 2.5 NM the controller had to come up with a resolution. Figure 9 shows the different resolutions the controllers entered. In the condition where the controller used the computed separation at CPA and the speed evaluation tool, almost 90% of the solutions were a correct speed change. If only the computed separation at CPA was available, the percentage of correct speed changes was reduced to 30%. The rest was either rejected or the speed change did not solve the conflict (incorrect speed). Only in the condition were the controller used the current tools, 20% of the trajectories where accepted without a speed change. Because the actual separation was 2.5 NM these were labeled as missed detection. 6
100% 90% 80% 70% 60% 50% 40% Reject Correct Speed Missdetect Incorrect Speed 30% 20% 10% % Current CPA CPA and Speed Figure 9. Resolutions entered by controllers. C. Speed Change Effect Figure 10 shows the effect of a speed change made by the controller on the separation. The speed changes were made in all conditions and include the unnecessary changes. When the controller used the current tools, the separation after a speed change varied from 24 NM to 0.4 NM. Three speed changes resulted in a separation below 5 NM. In the condition where only the computed separation at CPA was available, the new separation varied between 16 NM and 1.6 NM. The 1.6 NM separation was a result 30NM 25NM 20NM 15NM 10NM 5NM NM CPA and Speed Current Tools CPA CPA and Speed Figure 10. Separation as a result of a speed change. of an unnecessary speed change! In the scenarios with the speed evaluation tool, the separation varied between 11 and 5 NM. D. Control Margin The controllers had to evaluate two conditions where only the control margin was varied. Six controllers accepted the trajectories in both conditions. The other four controllers changed the speed in the condition where there was enough control margin, three rejected the trajectory if there was no control margin and one controller accepted the trajectory in this condition. E. Trajectory Geometry The different trajectory geometries did not affect the detection and resolution task in the conditions where only the separation at the CPA was computed and in the conditions where also the speed tool was available. There was a small difference in the detection task when controllers only had the current tools. The number of unnecessary rejections increased from 18% if both trajectories did not have a turn to 23% if there were turns in the trajectory. There was no difference between the conditions with only one trajectory having a turn and both trajectories having a turn. 7
V. Discussion The experiment shows that the current tools do not give enough insight in the separation margins a new trajectory actually has. In a trajectory based environment this may lead to unnecessary interventions by the controller. When new tools were added the number of unnecessary interventions was significantly reduced. However when the separation at CPA was calculated and shown to the controller, a number of controllers still rejected or modified trajectories when the separation was 5 NM. This points to the earlier mentioned explanation that controllers use their own heuristics to assess whether a trajectory will be safe or not. This was confirmed by some of the comments made by the controllers during the experiment. The speed evaluation tool improved the ability to generate a solution based on a speed change. However, by showing the speed tool, the number of unnecessary changes was increased. The increase in speed modifications was not only when the separation margin was just sufficient (5 NM) but also when the margin was larger then the separation requirement (7.5 NM). Apparently the speed tool encouraged controllers to increase the separation as long as it is within the available control margin, indicated by the green bar in the speed evaluation tool. Since there is no penalty for changing the speed, from a controller s point of view the situation will improve if the separation is increased. However, a trajectory change (in this case speed) can have a negative effect on the traffic flow at another location within the ATM system and can trigger the modification of more trajectories in a trajectory based environment. To allow the controller to balance the local solutions versus the total optimization, future research should look at a way to indicate to the controller what the effects of a trajectory change are. VI. Conclusion Future Air Traffic Management systems may lead to a shift from the state based operation that is used today to a trajectory based operation. If all aircraft participate in the trajectory based operation and all possible situations are taken into account, monitoring and re-planning the trajectories could be automated. However in the near future not all aircraft will be capable of trajectory based operations and it will be difficult to design an automated system that can cope with all situations. Furthermore a human will probably always remain responsible. As a consequence a controller in a (semi) trajectory based operation must be able to monitor the automation and in unforeseen situations come up with a solution. To be able to perform these tasks, the controller needs tools to asses the different trajectories to detect unforeseen unsafe situations and subsequently generate a solution. To evaluate the usability of tools that are used today to asses a new trajectory and to explore the potential benefits of new tools that would make the separation margin more explicit, an experiment was conducted. The results show that the current tools do not give enough insight in the separation margins a new trajectory actually has. Adding a tool that calculates the separation at the CPA significantly reduces the number of unnecessary interventions. However, still a number of trajectories with a separation of 5 NM were rejected. This indicates that a safe trajectory by controller standards may require more than missing other aircraft by the minimum separation requirement. Adding the speed evaluation tool helped the controller with generating a resolution, but increased the number of unnecessary speed changes. The speed tool encouraged controllers to increase the separation using the available control margin. This local benefit could have negative effects on other locations within the ATM system. To allow the controller to balance the local solutions versus the total optimization, future research should look at a way to indicate to the controller what the effects of a trajectory change are. Acknowledgments The author would like to thank the air traffic controllers from the Royal Netherlands Air Force 711 Military Air Traffic Control Centre (MilATCC) Squadron for there participation in the experiment and Erik Theunissen for his input in developing the graphical user interface. References 1 Erzberger, H., 2006, Automated Conflict Resolution for Air Traffic Control, 25th International Congress of the Aeronautical Sciences (ICAS), Hamburg, Germany 2 AATT Project Office, 1999 NASA Ames Research Center, AAT Concept Definition for Distributed Air/Ground Traffic Management (DAG-TM), Version 1.0, Moffett Field CA. 3 Smith, N., P. Lee, T. Prevot, J. Mercer, E. Palmer and V. Battiste, 2004, A Human-in-the-Loop Evaluation of Air-Ground Trajectory Negotiation, AIAA 4th Aviation Technology, Integration and Operations (ATIO) Forum, Chicago, Illinois. 4 Wichman, K., Lindberg, L., Kilchert, L., Bleeker O., AIAA Guidance, Navigation, and Control Conference and Exhibit, Providence, Rhode Island, AIAA-2004-5413, Aug. 16-19, 2004 8
5 Eurocontrol, Towards a controller-based conflict resolution tool a literature review, 2002. URL: http://www.eurocontrol.int/eatm/gallery/content/public/library/litreview.pdf. 6 Chiang, Y.-J., Klosowski, J. T., Lee, C., and Mitchell, J. S. B., Geometric Algorithms for Conflict Detection/Resolution in Air Traffic Management, 36th IEEE Conference on Decision and Control, IEEE Control Systems Society, Piscataway, NJ, 1997, pp. 1835 1840. 7 Vivona, R., Karr, D., and Roscoe, Pattern-Based Genetic Algorithm for Airborne Conflict Resolution, AIAA Guidance, Navigation, and Control Conference and Exhibit, AIAA-2006-6060, August 2006 8 Kosecka, J., Tomlin, C., Pappas, G. and Sastry, S., Generation of conflict resolution maneuvers for air traffic management, in Proc. 1997 Int. Conf. Intell. Robot. Syst., Grenoble, France, Sept. 1997, pp. 1598 1603. 9 Tomlin, C., Pappas, G. and Sastry, S., Conflict resolution for air traffic management: A study in multi-agent hybrid systems, IEEE Trans. Automat. Contr., vol 43, pp 509-521, Apr, 1998 9