Adaptive Robust Revenue Management with Adjustable Regret

Robust revenue management has received much interest from both academics and practitioners alike. But the airlines have not embraced it for its current inadequacy to address the practical needs for both fine control of conservatism and the adaptability to changing environments. The single-leg capacity control problem is studied to address such needs with limited information on demand uncertainty. A robust optimization framework with the adjustable regret criterion is employed along competitive analysis, which only requires the lower and upper bounds on demand, and comes with a performance guarantee. The analysis finds one extreme scenario for each fare class to formulate a linear program of modest size, providing nested booking limit policies that are proved optimal among online policies. Closed-form solutions to continuous problems are found, and shown theoretically to be less conservative for a larger control parameter, and their effectiveness in adaptability and fine control of conservatism is demonstrated by a simulation study. A robust single-leg capacity control method is available to allow by fine-tune its level of conservatism forcontrolled response to real time key performance indicators or to a more optimistic outlook from textual data analysis or artificial intelligence. If the control parameter is well calibrated, the simulation study shows that the new method can achieve a less than 2% revenue gap from the EMSR policies, and an improvement of up to 3% in average revenue over the absolute regret robust policies