A Robust Control Approach to Optimizing Production, Inventory and Transportation

Graduate Seminar byProfessor Thomas MarlinDepartment of Chemical EngineeringMcMaster University
Ontario, Canada AbstractRecently, advances in computing and optimization algorithms have lead to a renewed interest in analyzing logistics systems with the recognition that (1) substantial uncertainties exist in their dynamic behavior and (2) periodic re-optimization (rolling horizon optimization) affects the future behavior. The Model-Predictive Control (MPC) structure is ideal for modeling these closed-loop logistics systems. This talk will introduce the concept of robust model-predictive control of uncertain systems and how it can be implemented in real-time. Challenges in formulation and computation will be introduced, and proposals for a computationally tractable approach presented. Application to a simple (but real) industrial logistics problem will be presented. The problem has several manufacturing steps with intermediate inventory and transportation to regional distribution outlets. Uncertainty occurs in manufacturing times, transportation times, and customer demands. The goal is to reduce holding (inventory) costs while preventing backorders, where possible. The behavior of the system under various control approaches will be compared, and the advantages of a robust approach quantified.This work has been performed in conjunction with Adam Warren and Xiang Li at McMaster University.

Location: Olin Hall of Engineering (OHE) - 122

Audiences: Everyone Is Invited

Contact: Petra Pearce Sapir