Reinforcement Learning for Dynamic Pricing with resource constraints in a competitive context

Laurie Guenin, Dominique Barth, Christian Cadéré

Confirmation: I have read and agree with the workshop's policy on behalf of myself and my co-authors.

Tracks: Main Track

Keywords: Dynamic pricing, Reinforcement Learning, Car rental, Revenue Management.

TL;DR: This paper defines and evaluates a specific reinforcement learning approach for dynamic pricing in a resource-constrained competition context, applied to car rental

Abstract: Dynamic pricing has emerged as a critical strategy in industries with limited resources and fluctuating demand, such as airlines, hotels and car rentals. Traditional approaches rely on static price grids or manual interventions, which lack adaptability to real-time market changes. This study investigates the application of Reinforcement Learning (RL) to dynamic pricing in the car rental industry, incorporating resource constraints and competitive dynamics. A repeated game model simulates the competitive interactions of car rental companies seeking to optimize profits while managing limited fleet capacities. Based on real-world data, an experimental evaluation compares RL with a traditional resource-based pricing method and a mixed approach. Results indicate RL's superior performance in adapting to market variability, though it risks underpricing in capacity-constrained scenarios. A proposed mixed method balances competition and resource considerations, outperforming both RL and resource-based strategies by dynamically adjusting to market pressures and resource availability. The findings highlight the potential of RL and hybrid approaches in enhancing revenue management in competitive, resource-limited contexts. Future research will explore automated parameter tuning for dynamic scenario adjustments.

Submission Number: 28