Route Analysis and Trajectory Synthesis
Final Approach Spacing Tool (FAST) depends on the accurate estimates of arrival times for all aircraft. These arrival times are used by FAST for sequencing and scheduling aircraft to the runway threshold. CTAS's Route Analyzer (RA) and Trajectory Synthesizer (TS) provide the rapid update and accurate calculation of Estimated Times of Arrival (ETAs) based on radar track or flight plan data.
The set of ETAs that the RA/TS computes represents ranges of possible arrival times given an aircraft's predicted route of flight combined with possible variations in degrees of freedom along those routes. Typical degrees of freedom include speed, horizontal maneuvers, and vertical maneuvers. Upon receipt of a flight plan or radar track data update (x-coordinate, y-coordinate, altitude, ground speed), the RA ''categorizes'' each aircraft's situation for each potential landing runway in terms of destination airport, airport configuration, geographical sections of airspace, engine type (jet, turbo-prop, or piston), approach segment (downwind, final, base, etc.), and aircraft states (x-coordinate, y-coordinate, altitude, heading, speed). Each situation category has a name and a complete description of route/degree of freedom combinations that are possible for that aircraft. The RA uses this site-adaptable data for each situational category to build a series of one or more routes for each aircraft, apply degrees of freedom to those routes, and finally request the TS to compute ETAs for each route/degree of freedom combination.
The inputs to the TS are the aircraft state, winds aloft, temperature and pressure profiles, a series of waypoints depicting the expected route of flight for an aircraft, and vertical and speed constraints on the predicted route. The outputs from the TS include a complete time-based (4D) trajectory along the expected path including all data pertinent for resolving conflicts and estimating times of arrival at points along the path.
As the aircraft flies through the arrival airspace and descends to the runway, it will change situation categories as it transitions from one flight segment to the next, producing stable sets of ETAs. These sets of ETAs form the basis for the sequencing and scheduling process. Once the sequencing and scheduling process is completed, the same set of RA/TS trajectories will be used for conflict resolution. Finally, they will be used as a reference in computing expected delay for aircraft in the runway allocation process.
Sequencer and Scheduler
Air traffic controllers must construct a plan about how aircraft will merge together and land safely. The Profile Selector (PFS) assists controllers by sequencing and scheduling aircraft at various points along their flight path while maintaining proper spacing and avoiding conflicts. This process is described below.
A trajectory is made up of a set of time steps at defined intervals. Each time step contains a predicted x-coordinate, y-coordinate, altitude, speed, and heading of an aircraft at a future time. A trajectory segment is a group of time steps that fall within a predefined segment of flight. The figure at right shows an aircraft and its trajectory broken into four trajectory segments called LONG_LEFT, DOWNWIND_LEFT, BASE_LEFT, and FINAL.
The first step in producing a plan or sequence is to determine an order in which to land the aircraft. Ordering is the process of both creating a relative sequence within each trajectory segment and combining those sequences into a consistent global ordering for each runway.
At the beginning of each sequencing cycle, FAST builds a new trajectory for each aircraft at its current position. The sequencing algorithm uses these trajectories to calculate the order of aircraft relative to one another. The previous sequence is also considered to provide stability to the new sequence.
The diagram above depicts a situation where seven aircraft merge together to land on the same runway. First the aircraft on each individual segment are ordered. Thus, each segment shown in the diagram at right may be ordered as follows.
Next, starting with the outermost segment, segments are merged such that the relative order in the previous step are preserved. In this example, the following sequences would be built.
|Merging Segments||Merged Sequence|
|C B G A|
|D C F E B G A|
Notice that in computing the merge of LONG_LEFT, and OVER_THE_TOP_RIGHT_TO_LEFT, onto DOWNWIND_LEFT, the aircraft maintain their order relative to other aircraft on the same segment. For example, even though aircraft G is inserted between aircraft B and A, aircraft B and A are in the same order relative to each other as when the sequence was computed for the LONG_LEFT segment alone. The result of merging the aircraft onto DOWNWIND_LEFT is used when merging DOWNWIND_LEFT, and BASE_LEFT onto FINAL. The resulting sequence on FINAL is the overall landing order for this runway.
Finally, with the sequence on the FINAL trajectory segment in
hand, the Scheduled Time of Arrival (STA) of
each aircraft is computed. The STA for the number one
aircraft is set to the aircraft's nominal arrival time. STAs
for the remaining aircraft are calculated according to:
STA of Tail Aircraft = Max[(STA of lead aircraft + required separation), Aircraft's Nominal ETA]The algorithm subsequently computes a delay for each aircraft by subtracting the STA from the aircraft's nominal time.
All trajectory segments for an aircraft are checked for conflicts with other aircraft within the same segments. If there is no conflict for an aircraft, it will be assigned its nominal trajectory. When a conflict is predicted, one or both aircraft trajectories must be manipulated to resolve the conflict. Because the aircraft have already been ordered, PFS knows which aircraft is ahead and which aircraft is behind. PFS will add delay to the trailing aircraft in order to resolve the conflict. The PFS accomplishes this by searching the trajectory for degrees of freedom which will help to resolve the conflict. The magnitude of the conflict is measured and translated into a required delay for the aircraft before it reaches the conflict point. Because the PFS knows which degrees of freedom will help to resolve the conflict, it can combine this knowledge with the route/degree of freedom/ETA values it received from the RA/TS to bound and then begin the iterative process for resolving the conflict.
The process of resolving conflicts contains a number of complicated situations. It may seem that for each aircraft added to the sequence, FAST only has to resolve violations that are with an aircraft ahead in the final sequence. Unfortunately, there are a number of cases where the situation is more complicated. In the figure above, for example, aircraft A will merge with aircraft B on DOWNWIND_LEFT and then another C on FINAL. Assume that the final ordering is B, C, then A. The idea of resolving only the conflicts A had with C could leave a DOWNWIND_LEFT conflict unresolved. The problem with checking for conflicts only with the aircraft ahead on FINAL is that a merge could be missed. FAST will search for and resolve conflicts on all trajectory segments between a given aircraft and the aircraft sequenced ahead of it.
In general, aircraft are vectored from Center airspace into a TRACON over a feeder gate or metering fix. Aircraft engine types (e.g. turbo-jets, turbo-props, piston) and feeder gate assignment map to a preferred runway which is typically the closest runway to that feeder gate. Depending on the procedures at a given TRACON, an aircraft may be eligible to change to secondary or alternate runways. Situations which would influence a controller-initiated runway change include excessive or unbalanced delay buildup on the preferred runway, controller workload for a given runway, and airline or control tower preferences for ease of ground traffic movement. A controller will select which aircraft to change to an alternate runway based on a number of considerations such as separating aircraft of a dissimilar engine type or weight class from the other aircraft in a busy stream of traffic, avoiding potential conflicts in current streams of traffic, or avoiding potential conflicts in merging streams of traffic. Ideally, a controller would like to change the runway early in the traffic flow (i.e. near the feeder gates), but because of the uncertainties of making such a decision early in the flow, changes are commonly held off until the last possible moment. This can cause an undesirable increase in workload for the pilots of arriving aircraft because of the late changes in selecting navigation frequencies and configuring the aircraft for an approach.
The strength of an automation system such as FAST is its ability to assign runways based on accurate estimations of delay savings and workload benefits at an early stage of the arrival process. The runway allocation algorithm employed in FAST attempts to meet three primary objectives:
The algorithm is heuristically-based and site-adaptable. The approach is to define the preferred runway for all aircraft in the landing sequence and then to select the set of all aircraft which are eligible for reassignment, apply criteria to narrow this set to a most likely aircraft to be reassigned, to test the aircraft's new runway in a full sequencing and conflict resolution cycle with all other aircraft, and finally to apply detailed criteria to this solution set for all aircraft.
The set of aircraft eligible for runway reassignment is defined by a runway allocation time window for each runway. The time window begins with a ``start testing runway allocation time horizon'' measured in expected flight minutes from a given runway and ends with a ``freeze runway allocation time horizon'' also measured in expected flight minutes from the runway. Any aircraft with an estimated time of arrival for a runway which falls within the runway allocation time window are deemed eligible for allocation to that runway. Once the set of eligible aircraft are determined for all arrival runways, the system builds an estimated schedule and its associated delays for each aircraft to their currently assigned runway and to any available alternate runways. The system then selects those eligible aircraft which pass a set of runway allocation heuristics. This selection process is based on a site adaptable decision tree file which incorporates facility procedures, delay reduction, and controller heuristics.
After all eligible aircraft have passed through this decision tree and thus narrowing the list of all eligible aircraft to a smaller set, FAST then selects a single aircraft which appears to have the greatest delay benefits to the overall arrival system. In some cases, there may not be any aircraft which pass these criteria and in this case, the runway allocation algorithm will not consider any aircraft for that update cycle. Once an aircraft is selected, it is then placed in an alternate runway sequencing and conflict resolution cycle. The entire arrival airspace sequencing problem is solved with this aircraft placed on its alternate runway. This allows the software to evaluate all aspects of the particular runway allocation. Full trajectory solutions are obtained for each aircraft which in turn give accurate sequences, expected delay, and conflict detection for the entire airspace. At this point, a new and more detailed set of criteria are applied. These criteria examine trajectory based issues such as potential conflict resolution problems and exhaustion of critical degree of freedom limits. They are applied to the alternate solution set in order to make the final determination as to whether or not to change the aircraft to the alternate runway.
Once an aircraft has been switched away from a given runway, that runway is blocked off from further consideration for that aircraft. A more optimal solution would be to allow allocation of this aircraft back to its original runway if a situation warrants, but this was found to be unacceptable to controllers. Controllers always have the final authority in the runway assignment, and if a controller directs FAST to assign a given aircraft to a runway through a keyboard entry, the system will freeze that decision and no longer consider that aircraft for any other runway. Finally, once an aircraft's ETA falls below a runway's freeze time horizon, that runway will be blocked off from further consideration. After all but one runway has been blocked off, the runway assignment advisory is frozen for the remainder of the flight. In nearly all cases, the aircraft has a frozen runway assignment before twelve minutes of flight time from the runway.
The development of the controller interface has focused on implementing FAST on two different controller interface platforms. The first interface platform is the current controller interface in operation at Dallas/Fort Worth called the Full Digital ARTS Display (FDAD). The FDADs employ a monochrome digital display with trackball, keyboard and analog input devices. The FDADs will be used as the controller interface in the initial field implementation of FAST. The second interface platform is a Sun workstation color monitor with mouse/trackball and keyboard input devices. This color workstation was the initial development platform for FAST and was used primarily before the FDADs became available for testing at NASA.
To assure an effective controller interface, the FAST development team used active air traffic controllers throughout the interface design and four key guidelines evolved:
The first line indicates the aircraft identification or call sign. The second line contains two data fields. The first data field contains the current reported altitude (in hundreds of feet) time-shared with a facility scratch pad (typically containing the current runway assignment), and the second data field contains the aircraft's current ground speed (in tens of knots) time-shared with the aircraft type (e.g. Boeing 727 is displayed as ``B727''). The third line shown is the FAST information data line and is not currently displayed operationally in TRACONs. This line also contains two data fields. The first data field contains the FAST relative sequence number to the runway time-shared with the FAST runway assignment advisory. The second data field can contain both indicated airspeed and heading advisory information. If the data is indicated airspeed information, it is shown in tens of knots, and if it is an advised heading, it is show in tens of degrees (magnetic North) preceded by an ``H'' for heading.
Heading advisories are displayed as a location to issue the turn (shown by a graphical marker), a magnetic heading in degrees next to the marker, and a turn arc depicting the projected aircraft path taking into account its speed, heading and the winds aloft. The graphical data is color coded based on arrival feeder gate for the aircraft. The aircraft flight data block changes color to match the graphical advisory. When the aircraft executes or passes the advisory, the flight data block reverts back to its nominal color.
Simulation and Field Testing
The planning and development for field testing and implementation has been ongoing for several years. Initially, simulations were conducted with controllers from all over the U.S., but as the development progressed, simulations focused exclusively on and with controllers from the Dallas/Fort Worth TRACON in preparation for field testing at that facility.
The simulations focused on a number of issues ranging from validation of the algorithms to an evaluation of human factors issues. They assisted both in the development of the system and the planning of the field deployment. The simulations had the following objectives:
Initial information on the potential benefits was obtained in a simulation evaluation of FAST operating on a color display in a single runway, Instrument Flight Rules (IFR) configuration (Davis et al, 1991). This simulation demonstrated efficient use of airspace, increased landing rates, and controller acceptance of the system. Similar results were obtained in an independent study (Credeur et al, 1993).
More recently, simulations of FAST designed for the Dallas/Fort Worth TRACON have demonstrated the system's performance in more complex operations with multiple runways in both IFR and Visual Flight Rules (VFR). These simulations have included parallel simultaneous and staggered approaches, as well as converging approaches. The simulations have been conducted on the FDAD's with traffic scenarios based on live traffic samples from the Dallas/Fort Worth TRACON. The primary results of the tests are discussed below.
First, the controllers have reported that detection of the speed and heading advisories on a monochrome display is difficult and requires additional workload. This is largely because of screen clutter from non-arrival air traffic. In addition, controllers sometimes have difficulties associating advisories with the correct aircraft on the monochrome display. The controllers stated that the color display mitigates these problems substantially.
Second, the controllers felt that sequence numbers and runway assignment advisories would provide substantial benefits even on the monochrome displays. The controllers report that these advisories often improved on their own decisions. The best use of the sequence numbers was found to be in sectors where the controllers merged streams of traffic. Runway assignment advisories were found to match or improve the controller decisions in most cases. Occasionally, controllers had a tendency to doubt runway allocation advisories because FAST could ``see'' aircraft that are out of the controller's view and thus make an accurate assessment at an earlier stage. However, the ``doubtful'' runway assignment advisories were nearly always shown to be correct.
Passive FAST vs. Current Operation
Passive FAST Operational at Dallas/Fort Worth TRACON
On February 15, 1999, the Passive Final Approach Spacing Tool (pFAST) began sixteen hour per day operational use at the Dallas/Fort Worth (D/FW) TRACON. This was a major milestone for the pFAST Free Flight Phase One (FFP1) effort as it demonstrated that pFAST could be independently adapted and run operationally by an FAA contractor (Sterling Software) without substantial NASA oversight or input. In the two weeks of operation, over 80% of the possible arrival rushes were run using pFAST. Controller acceptance of the runway advisories was 96.9%, and acceptance of the sequence advisories was also high, although not specifically tracked. Air traffic controllers and D/FW TRACON management reported the first week of operations as being ``very positive.'' Controllers in the D/FW Tower reported that airport traffic balance was at an all time high, and that rushes seemed to begin and end earlier, potentially pointing to reduced delays. The D/FW Tower also reported an increase in surface traffic operational efficiency due to the improved overall balance of traffic on the ground. To date, only one software change to address improved sequencing of low speed general aviation aircraft has been required and the resulting change has been transferred to D/FW for operational use with positive results. The D/FW TRACON continues to use pFAST on a daily basis and analytical studies to assess the benefits impact will be initiated following an initial 1-2 month ``burn-in'' period.
An automation system for assisting terminal area air traffic controllers in efficiently managing and controlling arrival traffic has been developed and tested in simulation and in the field. The automation system, referred to as the Final Approach Spacing Tool (FAST), was developed through thousands of hours of real-time simulations and field tests with active air traffic controllers. Results from these tests show benefits in efficient airspace utilization, reduced controller workload, increased runway capacity, and reduced air traffic delays.
The FAST system is designed to operate either independently or in direct coordination with the other CTAS tools. However, simulation results indicate that the arrival air traffic management process will receive the most benefits by utilizing all tools sets in CTAS. Integrated tests of all CTAS tools are planned for the Dallas/Fort Worth and Denver airports.
Adapted from NASA CTAS site.
You Decide Intro
You Decide Scenario
You Decide Decision-Making Process