2224. Reinforcement Learning-Based Real-Time Inventory Management for Quick Commerce Supply Chains
Invited abstract in session WB-47: Empirically Driven OR in Retail, stream Retail Operations.
Wednesday, 10:30-12:00Room: Parkinson B08
Authors (first author is the speaker)
| 1. | Rahul Chavhan
|
| Shailesh J. Mehta School of Management, Indian Institute of Technology, Bombay | |
| 2. | Angela Biswas
|
| 3. | Pankaj Dutta
|
| Shailesh J. Mehta School of Management, Indian Institute of Technology |
Abstract
The rise of Quick Commerce (Q-Commerce), fueled by post-COVID e-grocery expansion and the demand for ultra-fast delivery, has intensified pressure on retail logistics to ensure seamless product availability. This shift necessitates real-time inventory management, driven by unpredictable demand, stockouts, and perishable goods. Traditional replenishment models struggle to adapt, requiring data-driven solutions. This study collects data through web scraping from a leading Q-Commerce platform over a month, capturing daily fluctuations in inventory levels, pricing, and discount variations. These insights reveal demand patterns, stock volatility, and the impact of dynamic pricing, forming the foundation for robust forecasting and optimization techniques. Time-series models like LSTM and ARIMA enhance demand prediction, while Change Point Detection (CPD) identifies sudden demand shifts. A Soft Actor-Critic (SAC) reinforcement learning framework dynamically adjusts inventory replenishment, optimizing stock levels and reducing wastage. By deploying a real-time reinforcement learning agent across multiple dark stores, this approach outperforms traditional methods, improving fulfillment rates and minimizing holding costs. This research contributes to more sustainable and efficient retail logistics by enabling real-time decision-making, reducing waste, and enhancing operational resilience.
Keywords
- Inventory
- Analytics and Data Science
- E-Commerce
Status: accepted
Back to the list of papers