Title:
Online Learning and Pricing for Service Systems with Reusable Resources  

Abstract:
Revenue management with reusable resources finds many important applications in today’s economy, such as cloud computing services, car/bicycle rental services, ride-hailing services, and call center services. In this work, we consider a price-based revenue management problem with finite capacity of reusable resources over a finite time horizon. Customers arrive following a price-dependent Poisson process and each customer requests one unit of homogeneous reusable resources. If there is an available unit, the customer gets served within a price-dependent exponentially distributed service time; otherwise, the customer waits in a queue until the next available unit is released. We assume an incomplete information setting where the system operators do not know how the arrival and service rates depend on posted prices, and thus they make adaptive pricing decisions in each period based only on past observations to maximize the cumulative revenue. We formulate this pricing problem as an analogue to the multi-armed bandit problem and propose two online learning algorithms, termed Batch Upper Confidence Bound and Batch Thompson Sampling. By bounding the transient system performance upon price changes via a novel coupling argument, we prove that the regret upper bound of the proposed algorithms matches the lower bound with up to a logarithmic factor. We start with a single-type resource setting under discrete candidate prices, and then extend the algorithm and performance guarantee to a continuous price setting and service networks with multiple types of resources.  

Bio of Huiwen Jia:
Huiwen Jia is an Applied Scientist at Amazon Middle Mile Pricing, Research, and Optimization Sciences. She received her Ph.D. and M.S. degrees in Industrial and Operations Engineering from the University of Michigan, Ann Arbor in 2022 and 2018, respectively, and received her B.S. degree in Industrial Engineering from Tsinghua University, China, in 2017. Her research interests include robust and stochastic optimization and online learning algorithms, with applications in sustainable and resilient transportation design and revenue management. Huiwen’ work has been published in high impact journals including Operations Research, INFORMS Journal on Computing, and INFORMS Journal on Optimization, Her work has also been recognized in top computer science conferences, including ICML and NeurIPS.