Some time ago I was asked to write about how the POLARIS software decides where an aircraft lands. At the time we knew of some limitations of the system and were planning to revise that part of the analysis, so it seemed sensible to hold on until that revision had been completed.
These limitations have now been addressed by a complete rewrite of the approach algorithm, so it now seems timely to dive into the subject. There is too much to put into a single blog post (I could write it, but you’d get bored) so I have split this into a number of bite-sized chunks.
In this first post I will describe the things that did not work in the earlier version, as knowing about problems is often more informative than just seeing a finished system working.
Offset ILS Approaches
For some airports, it is best not to approach the runway along the centreline. Here is an example from Bodo in Norway, where the high ground to the east of the runway means that the best approach is flown down the fjord and then the aircraft performs a left turn to land on the runway. From this plot you can see the pilot has flown the approach and in fact flies slightly to the right from the indicated aircraft position in order to line up on the runway centreline as early as possible.
The most dramatic offset ILS that I have come across is at Ronald Reagan Washington National Airport where the approach is down the Potomac River between the Pentagon and the White House. On the map below the aircraft is shown at the point where it first becomes established on the localizer.
On the original system, we did not have a record of which approaches were offset and so the system would indicate that the aircraft had diverged from the ILS and trigger an event. The analysts then manually marked these invalid, as they had access to approach plates and were familiar with the relatively low number of offset approaches.
This was fine for a small number of flights but as the workload increased this became impractical and so an automated solution was needed.
Another problem arose with approaches flown to one runway and then the aircraft changes approach and lands on a parallel runway. Here is a classic example, with the aircraft positioned just as it changes from runway 25L to 25C (I guess the Germans plan for a third runway rather better than the English!)
In these cases the approach is flown on the ILS for the first runway, and then when the pilot can see the airfield, he visually flies the step-across manoeuvre and lands on the other runway. Again, the old system would identify this as a divergence from the ILS approach path and trigger an event and the analyst would have to mark this invalid.
Dissimilar Receiver Operation
On aircraft with dual ILS receivers, it sometimes happened that one would start working a long time before the second (this may relate to when the receivers were tuned but we don’t always have the selected frequency values for both receivers). To illustrate the kind of problems that can arise, here are the two localizer signals for an aircraft on the approach to an airfield. The red No1 system works well, and No2 starts and finishes well, but in between goes off scale then appears to work for a period then diverges again.
There are many permutations of dissimilar receiver operation, and most of the corrections here come under the issue of parameter cleansing and merging which will happen before the main approach identification task. The analyst needs to be aware that these kinds of errors can arise.
In the next post I will look at how we build up a picture of the approach from the available information and later we will look at how we can cater for the offset ILS and runway change cases.