You definitely have captured my idea, and it is an interesting linear programming 'game.' Such a model would become slightly more reliant on price yield models, as you'd likely want to be able to predict certain things further in advance than others. Obviously the schedule for a Tuesday would be reliant on planes and crew ending up in the right cities on Monday evening. Historical models would help to get the basics right, but establishing a fairly fixed schedule seems to be a deadly faux-pas on the part of airlines. Sometimes they oversell (bad), and many times they undersell (worse.) The solution would, and should, definitely vary by the daily demand. Obviously the key elements are maximizing aircraft and crew utilization, as they are the most costly expense. Reality is that in many cases you will end up with a hub like model, as certain external constraints (e.g. gate and ground crew utilization) is something not easily controlled and a hub model does solve the problem. But the hub model does not support a point to point connection, even if it was the most profitable solution. Interesting thoughts... Matthew On Tuesday, May 20, 2003, at 09:27 AM, John Kurtzke wrote: > Matthew & List, > Now this is getting interesting, but it is getting into realms of > linear > programming about which I am ignorant. But, what the heck: Consider > some > of the variables: > > Passengers in city i who want to go to city j (do you accomodate all by > point to point service, or by huge -- ugh! -- hubs or by a mixture of > point to point and smaller hubs?). And by the way, I do not want to > have > to show up at the airport before 9 am (others may have getting to the > airport at 5 am as an ascetical practice, and God love them for it, > but I > believe 5 am is a time for sleeping) > > What aircraft you have in each city & what type; what will be coming > in & > going out? > > Where are the crews? What planes are the pilots certified on? > > Do you have your computers crank out this information at midnight and > call > folks at 1 am? Or, do you rely on historical data, and get what we have > now? > > But it is an interesting problem. I'd be curious to see someone run it > with somewhat realistic data. > > john > > On Mon, 19 May 2003, Matthew Montano wrote: > >> I've been building models that handle scenario's that take nearly 25 >> seconds for Excel to recompute,. Now-a-days, the traveling salesman >> problem, even over 100k pax a day, is a breeze for modern systems. I >> used a Blaze Business Rules Engine that was essentially built for it.. >> >> ..but I never tried it. >> >> In the example of making a connection, it becomes a constraint. The >> customer has to be in JFK by "X" time. Virtually everything in the >> model is a constraint, and you'd simply look for the lowest cost >> solution. >> >> Right now, airlines are still running with 60's models where the >> constraints of staff and crew (and scheduling) are such that they >> essentially drive the modern airline schedule. Then the revenue >> management guys go price the routes to try and fill the planes. >> >> After the plane flies, the airline then counts it's pennies and >> determines if it made money. More often than not ever since about >> 1970, >> they haven't. >> >> I was just thinking there had to be a better way. >> >> (Back to my day job.) >> >> Matthew > > -- > John F. Kurtzke, C.S.C. > Department of Mathematics > 278 Buckley Center > University of Portland > Portland, OR 97203 > 503-943-7377 > kurtzke@xxxxxx >