Optimise What?
Good to see you onboard with the discussion Highspeed. Although there are some points I ought to address from earlier posts I’ll just offer some thoughts on your area as my own experience in approach sequencing goes back about 15 years. The trail is littered with failures because some fundamental truths have not been recognised.
In the USA there was a proposal to eliminate holding stacks until some people pointed out that the problem is not deterministic but rather, that probability and queuing theory govern the results. Anyone still working on 4-D control?
NATS toyed with approach sequencing for CCF Stage 5 ‘Tunnels in the Sky’ until it became clear that the whole idea couldn’t work. The traditional rule ‘First Come First Served’ has no meaning if you extend your view outside a single sector or move away from the controller as the decision maker and final arbiter.
EATCHIP and eFDP missed much of the value they could have added by sprinkling the word ‘optimise’ everywhere in the requirements without saying what had to be optimised. Fuel? Time? Aircraft Aggregate Time? Aircraft Aggregate Fuel Burn? Passenger Hours? Landing Rate? Over what period?
You can only optimise against a SINGLE variable. If you want to optimise against more than one parameter then you have to relate them mathematically. As a minimum you will end up with partial differential equations. As a simple example, most aircraft use a cost index to optimise flight time and fuel burn. Time and fuel have to be given monetary values that can be traded off. You can’t have your cake and eat it.
Many years ago the London Ambulance Service tried to ‘optimise’ response times by sending the closest ambulance to an incident. Anyone who had encountered a Monte Carlo (Random Walk) Simulation would have known that this algorithm simply causes the ambulances to wander all over London and get lost. Some people believe that lives were lost when this happened in the real world.
So, please, if you want to optimise anything, write down exactly what it is and make sure you include ALL the dependent variables. And be aware that if you put this into a machine algorithm to influence the real world of ATC then I predict that the airlines will soon want to know exactly how it affects them.
Let’s suppose (for environmental reasons) that we have to minimise aggregate fuel burn. A simple algorithm would grade aircraft according to rates of fuel consumption and land the heaviest consumers as quickly as possible. So 747s would get close to straight in approaches while small business jets might have to hold until the tanks were nearly dry.
Having said all of that, the control of intermediate and final approaches by radar vectoring is one of the triumphs of the human mind. I agree that it may be almost impossible to design a machine to do as well by doing it the same way. What seems to be missing is the recognition that the airlines don’t want to do it this way. MLS is available and installed. The USA and many other countries are pushing towards GPS. I have read that EVERY new Boeing and Airbus aircraft has Multi-Mode Receiver (GPS/ILS/MLS) capability.
Airlines want curved approaches and continuous descents from cruising levels but these cannot be achieved using human beings, radar and VHF voice communications. The most impressive continuous descent approaches in the world are executed by the Space Shuttle. The night landing at Edwards AFB by Eileen Collins in Discovery was superb. I put it to you that the case for automation is proven (yes, proven) on all counts and that political factors are now the only reason for persisting with our current methods.