Inside Conflict Alert
Conflict prediction is probably the most complex task that the ATC computers at TRACONs and ARTCCs perform. If you're curious about how it works, here's a glimpse inside the CA patch. A special supplement to Mike Busch's article "Computer-Generated Safety Alerts."
In the TRACON version of CA, the process starts with a primary filter which runs through all possible pairs of tracked targets and eliminates those that are more than 40 seconds flying time from one another at their current groundspeeds. The track-pairs that make it through the primary filter are then analyzed by three different secondary filters, known as LINCON, PROCON,and MFMAMS.
LINCON is short for "Linear Conflict Projector". It projects the flight paths of each track-pair 40 seconds into the future, assuming that both aircraft will continue on their current track, groundspeed, and vertical speed (hence the term "linear"). LINCON calculates the projected lateral and vertical separation of the target-pair and compares them to threshold parameters in order to determine whether a potential problem exists. The threshold values are typically set at 1.2 nm lateral and 400 feet vertical for Type III (non-airport) airspace, and significantly less in Type II (approach) and Type I (airport) areas.
PROCON stands for "Proximity Conflict Detector", and looks for track-pairs that are already too close for comfort based on their current (rather than projected) positions and altitudes. PROCON is needed as a backup to LINCON, because LINCON can miss potential conflicts involving aircraft that haverecently changed course or vertical velocity. In most cases, PROCON generates an immediate alert anytime it finds two targets separated by less than 5 nm horizontally and 300 feet vertically.
MFMAMS means "Module for Maneuvers and Maneuver Sensitive Targets", and incorporates some complex algorithms for detecting and analyzing the geometry of tracks that appear to be maneuvering. It's purpose is to provide early warning of aircraft that are turning laterally toward each other or accelerating vertically toward each other.
The enroute version of CA is similar to the terminal version, except that the look-ahead time and alert thresholds are considerably greater, and there is no MFMAMS filter. There are also some timing differences: an ARTCC covers a huge amount of airspace and may be controlling many hundreds of aircraft at any given time. The center computer, which has to correlate radar data from many different radar antenna sites, is generally not fast enough to analyze all potentially-conflicting track-pairs in the time of a single radar sweep. Each complete track-pair analysis cycle can take a half-minute or more. Consequently, much longer look-ahead times are needed. At the Center, CA's primary filter performs a 2.5-minute look-ahead (vs. 40 seconds in the terminal version).