Seminar on Tackling Stochastic Constrained Problems via Simulation: Feasibility Determination and Optimization

Title: Tackling Stochastic Constrained Problems via Simulation: Feasibility Determination and Optimization

Time: 11am

Palce: ECEC 202, NJIT (Warren St. and Summit St., Newark, NJ 07102)

Abstract: In this talk, we consider a class of global optimization problems with a set of stochastic constraints, where noise-corrupted observations of these constraints need to be evaluated via simulation. These problems are challenging in that the solution space often lacks rich structure to explore the optimal solution, and the feasibility of a solution is unknown due to the noisy measurements of the constraints. To this end, we study two sets of such problems: (1) to find all feasible solutions that satisfy the stochastic constraints from a finite set of alternatives (known as feasibility determination); and (2) to optimize a deterministic objective function in presence of these stochastic constraints (known as optimization via simulation). For the feasibility determination problem, a sequential selection algorithm is developed based on a rule that asymptotically maximizes the probability of correct selection. For the optimization via simulation problem, a partition-based random search algorithm is developed which efficiently detects solution feasibility and converges to the set of global optima with probability one. Effectiveness of both algorithms are shown via numerical examples.

Bio: Dr. Chen’s research interest lies in supply chain operations planning and scheduling, interface of supply chain operations and finance, and intermodal transportation. He also works on simulation and randomized global optimization methodologies, and uses business analytics in solving practical problems in smart grid and healthcare operations. His work has appeared in journals, books and patents, including Operations Research, Manufacturing & Service Operations Management, Interfaces, Automatica, IEEE Transactions on Automatic Control, IEEE Transactions on Smart Grid, etc. Dr. Chen received his Ph.D. degree in Industrial Engineering from the University of Wisconsin-Madison, and the M.S. and B.S. degree from Tsinghua University, Beijing, China. Prior to joining Rutgers Business School, he was a Scientist at GE Global Research, NY, solving various problems from GE businesses, collaborators and customers, such as GE Energy, GE Aviation, Lockheed Martin and electric utility companies.

Contact: Prof. Mengchu Zhou, zhou@njit.edu

Agenda:

11:00 Seminar

Location:
Newark, New Jersey
07102