Optimizing Dynamic Systems: From Effective and Equitable Distribution to Coordinated Two-Stage Order Fulfillment
Optimizing Dynamic Systems: From Effective and Equitable Distribution to Coordinated Two-Stage Order Fulfillment
Gal Neria, PhD. student at the department of Industrial Engineering
Advisor: Prof. Michal Tzur
Abstract:
Dynamic decision-making under uncertainty is a central challenge in several domains such as logistics, warehousing, and production systems. This talk introduces hybrid operations research (OR) and machine learning (ML) methodologies designed to optimize stochastic dynamic combinatorial problems. The first part of the talk focuses on making real-time routing and resource allocation decisions, ensuring effective and equitable distribution in uncertain environments, such as food banks and disaster relief operations. It addresses resource allocation when uncertainty exists on the demand side alone or when both supply and demand are uncertain. The second part introduces the Dynamic Two-Stage Order Fulfillment Problem, a unifying framework for coordinating sequential operations such as meal preparation and delivery or order picking and dispatching. While traditional OR methods effectively manage complex decision spaces and constraints, they struggle with large state spaces. Conversely, ML techniques excel at handling large state spaces but lack the ability to model intricate decision constraints. By combining these approaches, we leverage their complementary strengths to develop more efficient and scalable solutions. Some key contributions include the development of several novel solution frameworks that are generalizable across dynamic systems, extending Benders Decomposition to MDPs, leveraging ML-driven feasibility predictions to enhance computational efficiency, and enhancing the solution process of stochastic dynamic programming by leveraging function interpolation. We validate our approaches using several realworld data sets, demonstrating significant improvements relative to benchmark solution methods for problems across food bank operations, meal-delivery, and various order fulfillment applications.
Bio:
Gal Neria, PhD candidate in the Department of Industrial Engineering at Tel Aviv University, advised by Prof. Michal Tzur. Her research focuses on combining operations research and machine learning to optimize dynamic stochastic combinatorial systems. She received multiple awards for her research contributions, including the Council for Higher Education in Israel (VATAT) Scholarship for Outstanding women doctoral students in the hi-tech field, and the Abraham Mehrez Prize for Best Paper by a Graduate Student in Operations Research. Two papers from her Ph.D. thesis were already published in Transportation Science
