Two Views of Simulation
Prior to this chapter, the focus of the book has been on building—literally writing computer code for—simulation models. This chapter sets up the remainder of the book which addresses issues related to experiment design and analysis. To do so, we present
- PDF / 293,248 Bytes
- 19 Pages / 439.36 x 666.15 pts Page_size
- 81 Downloads / 194 Views
Two Views of Simulation
Prior to this chapter, the focus of the book has been on building—literally writing computer code for—simulation models. This chapter sets up the remainder of the book which addresses issues related to experiment design and analysis. To do so, we present two different, but complementary ways of viewing computer simulation. Section 5.1 makes clear the roles of real systems, simulated systems, and the conceptual systems that the analyst wants to design. This framework highlights sources of error in a simulation study; it is abstract, but not mathematically formal. Section 5.2 views simulation output as a stochastic process to provide a framework for designing simulation experiments and analyzing the results; it is a mathematical treatment that provides the foundation for statistical analysis. Both of these perspectives are necessary to have a deep understanding of stochastic simulation.
5.1 A Framework for Simulation Modeling and Analysis Example 5.1 (The call center). A software company has a customer call center for inquiries or problems related to its products. Currently the agents are trained only for particular products, and therefore they only handle calls related to their product. The company would like to reduce staff, while still delivering the same level of service, by cross-training some agents to handle calls regarding more than one product. What staff level is required? This situation has characteristics that are common to many system-design problems for which we seek a simulation solution: There is a conceptual design for a new system (the call center with cross-trained agents). There is also an existing real system that is related to the conceptual one, but is not identical to it (the current call center with dedicated agents). Using the real system as a starting point, a simulated system will be constructed to evaluate how well the conceptual system might work. In this section we describe a framework for simulation problems such as the call center, a framework that will be informative as we think about simulation input modeling, output analysis and experiment design (the subjects of Chaps. 6–8). B.L. Nelson, Foundations and Methods of Stochastic Simulation: A First Course, International Series in Operations Research & Management Science 187, DOI 10.1007/978-1-4614-6160-9 5, © Springer Science+Business Media New York 2013
79
80
5 Two Views of Simulation
'
Inputs: FR Logic: LR &
$ Observable: Logic LR Inputs XR ∼ FR Outputs Y
-
Inputs: FSR Logic: LSR
%
Real System
Simulated Real System FSC ⊆ FSR
Conceptual System
'
Inputs: FC Logic: LC &
$
LC ≈ LSC FC ≈ FSC
Simulated Conceptual System
?
Inputs: FSC Logic: LSC
%
Fig. 5.1 Relationship between real, simulated and conceptual systems.
For the purpose of this framework, a system consists of inputs and logic. Inputs are the uncertain (stochastic) components of a system, while the logic may be thought of as a collection of rules or algorithms that govern how the system behaves as a function of the inputs. Let
Data Loading...