World-wide Crude Transportation Logistics: a Decision Support System Based on Simulation and Optimization

World wide crude transportation is the central logistics operation that links the upstream and downstream functions and plays a crucial role in the global supply chain management in the oil industry. In this work, we develop a decision support system to investigate and improve the combined inventory and transportation system in a representative world-wide crude supply problem. The decision support system is based on the integration of discrete event simulation and stochastic optimal control of the inventory/transportation system. A unifying simulation framework that integrates the simulation model and controller is constructed to simulate the controlled inventory/transportation system. It provides the decision makers valuable insights into the behavior of the dynamic and stochastic system and also a powerful tool to evaluate strategies and policies for the design and operation of the system. We formulate the optimal design/control problem rigorously as a Markov decision process that incorporates such uncertainties as travel time and crude demand. Due to the overwhelming computational requirements of the rigorous methods, approximate methods based on dynamic programming principles are needed to determine the near-optimal control policies that minimize the expected total cost. We propose an approximation architecture that involves two stages: decomposition of the system into individual subsystems and use of parametric function approximators for the cost-togo functions. We also provide future directions on computationally practical approaches to solve large scale industrial problems.