Theses and Dissertations
Advisor
Marufuzzaman, Mohammad
Committee Member
Ma, Junfeng
Committee Member
Wang, Haifang
Date of Degree
5-15-2026
Original embargo terms
Immediate Worldwide Access
Document Type
Graduate Thesis - Open Access
Major
Industrial & Systems Engineering (Data Analytics)
Degree Name
Master of Science (M.S.)
College
James Worth Bagley College of Engineering
Department
Department of Industrial and Systems Engineering
Abstract
The timber supply chain connects landowners and mills to provide wood products but faces challenges from stochastic demand, seasonal variations, and disruptions such as hurricanes. Fur- thermore, sustainability concerns like transportation emissions create trade-offs in procurement. This study proposes a feasibility-aware Deep Reinforcement Learning framework for sustainable timber procurement and inventory control under joint demand–hurricane uncertainty. We develop a stochastic mathematical model capturing mill-landowner interactions, seasonal demand, hurricane- driven pricing, and carbon emissions. The problem is formulated as a constrained Markov decision process and solved using Proximal Policy Optimization with a feasibility-enforcing layer. A Mississippi-based case study with 2,100 landowners evaluates the approach against three baseline inventory policies. Results confirm the method optimizes ordering decisions, balancing resilience and sustainability while outperforming baselines. Experiments further test variations in annual demand and sustainability parameters.
Sponsorship (Optional)
2020-67019- 30772 and 2022-67022-37861
Recommended Citation
Wright, Jarod, "Feasibility-aware deep reinforcement learning for sustainable timber procurement under hurricane demand uncertainty" (2026). Theses and Dissertations. 7012.
https://scholarsjunction.msstate.edu/td/7012
