2037. Reinforcement learning with delayed rewards for optimizing the parking permit prices
Invited abstract in session MC-49: Analytics and the link with stochastic dynamics 1, stream Analytics.
Monday, 12:30-14:00Room: Parkinson B10
Authors (first author is the speaker)
| 1. | Daiki Min
|
| School of Business, Ewha Womans University | |
| 2. | Jihee Seo
|
| Bigdata analytics program(Graudate School), Ewha Womans University |
Abstract
This study proposes a reinforcement learning model to address the dynamic pricing problem in O2O (Online to Offline) parking platforms. To manage uncertainty from the time gap between permit purchases and actual usage, we design a dynamic pricing model that incorporate lead time considerations. The proposed model leverages DQN with a backtracking mechanism, which redistributes delayed rewards to earlier decision points, allowing the model to adjust for penalties caused by overbooking and capacity constraints. Parking demand is modeled using probabilistic distributions, incorporating time variations and cross-elasticity between parking permits. Numerical tests show that the proposed reinforcement learning model outperforms conventional approaches by optimizing the reward function while reducing vehicle rejections. Comparisons with baseline strategies reveal that our approach optimally balances revenue and operational stability.
Keywords
- Analytics and Data Science
- Practice of OR
- Revenue Management and Pricing
Status: accepted
Back to the list of papers