Specfcally, for a multclass M n /M/ queue wth controllable arrval charges, common demand curves, and lnear holdng costs, we research the problem of maxmzng the anticipated revenues mnus holdng costs by selectng a par of dynamc prcng and sequencng polces. Usng a determnstc and contnuous (flud mannequin) relaxaton of ths downside, whch could be justfed asymptotcally as the capacty and the potental demand develop giant, we present the followng: () greedy sequencng (.e., the c -rule) s optmal, () the optmal prcng and sequencng decsons decouple n fnte tme, after whch () the system evoluton and thus the optmal prces depend only on the entire workload. Buldng on () (), we propose a one-dmensonal workload relaxaton to the flud prcng downside that s smpler to analyze, and results in ntutve and mplementable prcng heurstcs. Numercal outcomes llustrate the near-optmal performance of the flud heurstcs and the benefts from dynamc prcng. Topic classfcatons: income management; yeld administration; dynamc prcng; queung; sequencng; flud models.
Space of revew: Manufacturng, Servce, and Provide Chan Operatons. Hstory: Receved November 3; revsons receved December 4, August 5; accepted September 5.. Introducton The final decade has been marked by a growng nterest n the adopton of dynamc prcng strateges n such dverse areas because the arlne, resort, and retal ndustres. In most of those circumstances, the frm controls a fxed capacty of sources (e.g., the number of seats n a flght) that have to be bought up to a deadlne (e.g., the flght departure tme). By dynamc prcng, we discuss with the tactcal optmzaton of the prce of a product or servce (e.g., of an arlne tcket) as a functon of the remanng capacty and tme-to-go to maxmze the expected revenues extracted from these fxed sources. Ths practce, sometimes called income administration, s supported by sophstcated nformaton techniques processng giant amounts of demand information, and reles on an mplct assumpton that the frm can apply such prce changes n a relatvely effcent method.
Extra recently, manufacturng frms have additionally started evaluatng the use of such tactcal economc optmzaton instruments, Sales wth one notable instance comng from the automotve ndustry n the context of ts effort to market and produce custom cars n a make-to-order fashon. Broadly speakng, automoble manufacturers try to dynamcally alter the prce, target lead tme, rebate, etc., for a brand new order as a functon of the exstng outstandng orders, and smultaneously choose the approprate producton schedule to optmze ther proftablty. Jont use of economc and operatonal controls allows the producer to be extra responsve to changes n the market condtons and fluctuatons n the operatng envronment on account of randomness of the demand and producton functons. Operatonally, ths rases several nterestng questons. For example, n what methods are the economc and operatonal decsons coupled? What are the benefts of dynamc prcng n a producton settng? And what are practcal and effcent prcng and sequencng heurstcs for such problems? Wth ths motvaton, ths paper studes the problem of jontly optmzng over the dynamc prcng and sequencng polces for the multproduct, sngle-server queung system.
Broadly speakng, ths problem les n the nterface of stochastc network concept and revenue management, and the approach taken n ths paper combnes modellng and analyss technques from these two areas. Its outcomes llustrate the potental benefts of jontly optmzng prcng and producton decsons, and offer some nsght on learn how to practcally ntegrate these two functons that are likely to operate individually n many organzatons. We consder a make-to-order frm that produces multple merchandise, that s modelled as a sngle-server multclass M n /M/ queue. The frm s assumed to operate n a market wth mperfect competton, and has energy to nfluence the demand for the varous products by varyng ts prce menu. Assumng a common demand curve, lnear holdng prices ncurred by the frm, and convex capacty costs, we research the problem of fndng the optmal state-dependent prcng and sequencng strategy in addition to a statc vector of producton charges to optmze the system s lengthy-run expected proft fee. 2 Operatons Analysis 54(5), pp , 6 INFORMS 95 queue wthn the framework of Markov decson processes (MDPs).
Whle MDPs provde detaled descrptons of the system dynamcs and the optmal control problem, they're wth the excepton of very restrcted examples not amenable to exact analyss. Ths paper studes an approxmate formulaton of the proft maxmzaton drawback of nterest, posed n the context of the assocated determnstc and contnuous flud mannequin approxmaton to the underlyng stochastc producton system. Ths might be rgorously justfed by a robust-law-of-giant-numbers type of scalng n settngs where the producton charge and potental demand develop proportonally massive. Such models have been used successfully n the lterature each n revenue management settngs that lack producton dynamcs, and n producton techniques that do not nclude the tactcal prcng decsons. The man fndngs of ths paper are the followng. Capacty choce. We present that the long-run common proft maxmzaton downside for the flud mannequin reduces to a statc downside of choosng a vector of goal demand feelingcutelol.com charges and the servce price vector that maxmze profts n the absence of holdng prices (Theorem ). The optmal servce fee vector makes the capacty constrant bndng (Corollary ). Ths problem determnes the optmal capacty, however s too coarse to specfy good prcng and sequencng polces. C ontent was gen erated by GSA Conte nt Generator D emoversion!