ATC activities

Correlation, Sensor data-track Join

One of the imperatives for ATC is the best estimate of each aircraft vector whether cooperative or non-cooperative. We must answer the question, where is each aircraft and where is it going? The controller for redirecting flight uses this information. It is used by the system in determining potential air-to-air conflicts, by the system to determine conflict with terrain, by the system in suggesting changes in flight, to optimize time to destination or minimize fuel consumption. It should be obvious the track each aircraft is following is highly significant.

The track is developed from the set of sensor reports that are received at the controlling station. The reports may be from primary or secondary radar or linked data from the aircraft that provides GPS position and velocity data. Some other new sensors are now in initial stages of employment and can easily be accommodated.

What is the process followed? During a short period (e.g. 0.5 seconds) a database of reports is collected. In the system database a set of tracks are developed from these reports. A join of the report DB and the track DB implement this process called report to track correlation.

The join is made using a small (1-sigma) error box about both the report position and the current track position. (The box is 2-D or 3-D depending on the track and report characteristic. A 2-D box can intersect wit a 3-D track, but a 3-D report cannot intersect with a 2-D track.) The intersection of the report box and a track box is considered a hit, and the report is entered into the track record for track smoothing and prediction and the report is marked “correlated”. If one report is found to correlate with two tracks an “ambiguity” is declared and both tracks are marked and excluded from correlation this period. They are updated using current data. When a report does not correlate with any track the report error box is enlarged and correlation is again tried. This is done to consider the possibility there is more error in the report or that the track is in the process of turning. Reports that do not correlate during this second trial are given a further enlarged box and the process is repeated. Reports that do not correlate on the third pass are entered into an empty track record and declared to be new “tentative” tracks.

This process continues until all the reports received during this period are handled. Nominally there are about 1.4 report trials per received report. The scheduled correlation is complete, and the next scheduled process of track smoothing and prediction process is initiated when its startup time is reached. This process is of time complexity function (TCF) O(n) where n is the number of reports.

Track Smoothing and Prediction

In this operation all tracks that have had a “hit” in the correlation process are selected and the differences between the report and track positions are developed. The differences are modified by a weighting factor that is a function of track quality, and is developed for each dimension independently. Weights are large for newly initiated or turning tracks and become small as the track follows a straight line. The modified differences are added to the previously predicted position. This is called the smoothed position.

Similarly multiplying the position differences by a velocity weight and adding to the previous predicted velocity develops a smoothed velocity. The smoothed velocity is added to the smoothed position to produce the next predicted position that is used during the next correlation period.

These processes proceed simultaneously over the set of tracks and therefore are TCF O(1).

Display Update

Controllers and managers need data updated to the last second. Passing data to the controllers requires selecting data needed formatting it for the controller’s displays and then transmitting this to the system bus for selection and scale change at the display processor. Special marks or indications to a particular control station are included with the data for each controller. This process is scheduled every second.

Conflict detection and resolution

In this operation each track is predicted into the future in each of the three position dimensions against time. Each future envelope consists of upper and lower bounds, which are a function of one half of the minimum clearance in each dimension plus an uncertainty, which is related to track quality. All tracks future path envelopes are computed simultaneously. Then a future path envelope is taken from the first track – “the trial track”, and tested against all other tracks. This trial track is then deselected for future trials.

The test is done simultaneously in each dimension for the trial track against all other tracks. The first time of intersect and the last time of intersect of the future paths are recorded for each dimension with each track. There is a minimum time and a maximum time for each dimension recorded with each track. Then each dimension is tested for intersect time with each other dimension. If an intersect is found for all dimensions the track is marked for possible conflict with the trial track. The conflict mark for IFR traffic appears to the responsible controller during display update.

When a conflict is detected the resolution operation is applied to determine a minimum change in the trial track that will resolve the conflict. This change may be a small change in direction or a small change in altitude. Each resolution maneuver must be tried for to ascertain the maneuver will not cause conflict with any other track while assuring the maneuver will correct the conflict found. When a correct maneuver is found the maneuver is sent with the conflict mark as a suggestion to the controller.

VFR Automatic Voice Advisory

Automatic Voice Advisory (AVA) is similar to conflict detection above. Each participating track is simultaneously projected against all other tracks and the process informs participating VFR traffic of traffic in proximity of their flight. Participation involves each VFR flight being given an AVA secondary radar code and having coded messages that are converted to voice in the cockpit for the pilot. The message will point out possible conflicts in the near space – three to five miles – and the pilot can then make a necessary maneuver to avoid the conflict. A typical voice advisory might be “NR 1234 traffic at 3 miles at your 2 o’clock direction.” Altitude can be provided if known. This process is scheduled every four seconds, and is O(n) where n is the number of participating aircraft.

Terrain Avoidance

In this operation each terrain item is tried against each projected flight vector for some future time. This time may be, for example, one or two minutes. Any intersect within this time is marked in the track with terrain altitude for the controller or for VFR traffic in AVA.

The process is repeated every 8 seconds and is O(t) where t is the number of terrain elements.

Terminal Approach Optimization

At the present time this operation is left to controllers at the terminal control center. Aircraft for a given terminal are handed off from the center or from another terminal. Then the flights may be put into a holding pattern for some time before they are directed to a specific runway. Terminal approach optimization is intended to minimize the holding pattern by developing a landing order before they enter the terminal area.

First all the scheduled flight operations, arrivals and departures, for a terminal are collected for a period (say an hour). The arrivals are selected from the flight plans at this time. Then each flight is projected forward using its present flight characteristics to a handoff time at the terminal. The times are then ordered and interspersed with takeoff times to look for periods when the runways are not in optimal use or appear to be overloaded. Then when too many arrivals exist in an interval, say an hour, some flights can be tested by reducing speed and or increasing speed in order to optimize runway use.

Aperiodic requests

Aperiodic requests must be collected during an interval and then treated as scheduled. It is necessary to determine a maximum number of such events during the scheduled interval. What are aperiodic events? They maybe such things as a terminal update, for example, a change in wind conditions or runway direction. A request by a controller for the runway length required for a particular aircraft type. Or a change is entered in a flight plan. We have assumed there may be as many as 200 such requests per second, and assume all will be answered within one second. Thus an A buffer is used to collect all such requests, and a B buffer is switched in every other second while the requests in the alternate buffer are processed.

Each request is prepared in a given format by the requester, and is executed during the scheduled time period each second. We have allowed for 200 requests per second and suggest allowing 250 microseconds for each transaction. The TCF is O(r) where r is the number of requests pr second. Since the action of this period is independent changes and additions are easily incorporated without affecting any other operation.