Advancing Simulation and Optimization Methodologies
Integrated Simulation and Optimization
I have developed a novel integrated simulation and optimization methodology using stochastic programming and discrete-event simulation. The optimization model is first solved to make a decision under uncertainty, which is evaluated over several iterations of the simulation model. The output (e.g., performance measures) from the simulation model revises the optimization model (e.g. modify the objective function, constraint(s), parameter(s)). This process is repeated until a set of stopping criterion are met. The framework can be applied to challenging problems in many application areas.
Algorithms for Mean-risk Stochastic Integer Programs (SIP)
The mean-risk SIP models used in my research are also challenging to solve individually because of complicating constraints that prevent scenario separation. To address this problem, I contributed to the development of a Fenchel and disjunctive decomposition algorithm for mean-risk SIPs with fixed recourse for the absolute semideviation mean-risk measure. The work uses subgradient-based optimization to solve the LP relaxation and then generated Fenchel decomposition cuts based on a subset of scenarios. For the case of binary first-stage decision variables, the approach uses disjunctive programming to lift and translate the cuts so that they are valid for the rest of the scenarios.
Multi-method Simulation Modeling
I am currently interested in multi-method simulation modeling. This methodology combines systems dynamics, discrete-event, and agent-based simulation into one model. AnyLogic is the best known software for doing multi-method simulation modeling. I am using multi-method simulation modeling for my health policy research and plan to make further methodological extensions in this area as well.