Date of Graduation


Document Type


Degree Name

Doctor of Philosophy in Engineering (PhD)

Degree Level



Industrial Engineering


John Austin White, Jr.

Committee Member

Kelly Sullivan

Second Committee Member

Shengfan Zhang

Third Committee Member

Sunderesh S. Heragu


Block stacking storage, Facility logistics, Markov Decision Process, Optimization, Unit load storage systems, Warehouse operations


The focus of the dissertation is developing the optimization problem of finding the minimum-cost operational plan of block stacking with relocation as well as devising a solution procedure to solve practical-sized instances of the problem. Assuming changeable row depth instead of permanent row depth, this research is distinguished from conventional block stacking studies.

The first contribution of the dissertation is the development of the optimization problem under the assumption of deterministic demand. The problem is modeled using integer programming as a variation of the unsplittable multi-commodity flow problem. To find a good feasible solution of practical-sized instances in reasonable time, we decompose the original problem into a series of generalized assignment problems. In addition, to establish a good lower bound on the optimal objective function value, we apply a relaxation based upon Lagrangean decomposition in which the relaxed problem separates into a set of shortest path problems and a set of binary knapsack problems.

The second contribution of the dissertation is the development of the optimization problem under the assumption of stochastic demand. The problem is formulated as a discrete time finite horizon Markov decision process model, incorporating the recursive daily situation of determining the assignment of product lots to storage areas for a day based on uncertain daily demand and observed system information. To tackle computational intractability in solving practical-sized instances, we develop a heuristic solution approach taking an on-line manner by instantly determining an action for a single observed state rather than an off-line manner by predetermining an action for every state.