Blind Network Revenue Management
We consider a general class of network revenue management
problems where products being sold are linked by various resource
constraints. Mean demand rate at each point in time is dictated by the
demand function and the prevailing vector of prices, and the
objective is to dynamically adjust these prices so as to maximize
expected revenues over a finite sales horizon. If realized demand is
modeled as a controlled Poisson process whose intensity is given by the
demand function, then this problem can be solved, at least in
principle, using dynamic programming. The main point of this talk is to
consider the above dynamic optimization problem in a setting where the
decision maker is ``blind'' to the underlying demand function, and is
only able to observe realized demand.
We
introduce a family of pricing policies which are designed to balance
tradeoffs between exploration (demand learning) and exploitation
(pricing to optimize profits), and characterize the revenue loss that
stems from absence of a priori information on the demand function.