2.3 Combination of Simulation and Optimization Methods
35
encouraging, with implementations of specific or general algorithms. However,
for the success of simulation optimization in practice, software that (i) connects
data to models, (ii) simulates and optimizes models using effective solution algorithms, and (iii) provides the users with ways to assess the quality of the obtained
solutions need to be further developed.” The following section 2.3.3 gives a short
overview of the state of the art of software packages including both, simulation
and optimization.
2.3.3 Optimization Packages Interfaced With Simulation
Software
The combination of simulation and optimization software is required in situations,
when the modeled system is too complex to use simulation to evaluate the systems
performance for each set of possible input parameters. As described already in the
previous section, the technique used to optimize the systems performance needs
to be robust enough to locate an optimal set of parameters and it is desirable that
the used technique is reasonably efficient [An1998, p. 329].
Law has summarized the key requirements that an optimizing package for the
integration in a simulation software should meet [La2007, pp. 659–660]. While
the quality of the solution and the amount of execution time to get to the solution
are seen as the most important features, a dynamic display of information during
the execution, the consideration of linear and nonlinear constraints, stopping rules
and confidence intervals for expected values of the objective function, as well as
the possibility to make more replications for higher variance configurations in
order to precisely estimate the objective function should be part of the software.
There are several different commercial optimization packages available for
simulation software products, such as AutoStat, Evolutionary Optimizer (Extend),
RISKOptimizer, OptQuest, SimRunner2, and WITNESS Optimizer 17 . Nearly all
of them use metaheuristics as for problems with large solution spaces, neither
ranking and selection methods nor enumeration are the optimization technique
of choice. “Since it becomes difficult to use a variable number of replications,
as needed in ranking and selection, with metaheuristics, one usually uses a large,
but fixed, pre-determined number of replications (samples) in evaluating the function at any point in the solution space” [Go2015, p. 89]. Metaheuristics are based
17 Details on the commercial software packages as well as on supported simulation software products can be found on the vendor’s websites [Au2019a; Au2019b; Pa2019; Op2019;
Pr2019; La2019].
35
encouraging, with implementations of specific or general algorithms. However,
for the success of simulation optimization in practice, software that (i) connects
data to models, (ii) simulates and optimizes models using effective solution algorithms, and (iii) provides the users with ways to assess the quality of the obtained
solutions need to be further developed.” The following section 2.3.3 gives a short
overview of the state of the art of software packages including both, simulation
and optimization.
2.3.3 Optimization Packages Interfaced With Simulation
Software
The combination of simulation and optimization software is required in situations,
when the modeled system is too complex to use simulation to evaluate the systems
performance for each set of possible input parameters. As described already in the
previous section, the technique used to optimize the systems performance needs
to be robust enough to locate an optimal set of parameters and it is desirable that
the used technique is reasonably efficient [An1998, p. 329].
Law has summarized the key requirements that an optimizing package for the
integration in a simulation software should meet [La2007, pp. 659–660]. While
the quality of the solution and the amount of execution time to get to the solution
are seen as the most important features, a dynamic display of information during
the execution, the consideration of linear and nonlinear constraints, stopping rules
and confidence intervals for expected values of the objective function, as well as
the possibility to make more replications for higher variance configurations in
order to precisely estimate the objective function should be part of the software.
There are several different commercial optimization packages available for
simulation software products, such as AutoStat, Evolutionary Optimizer (Extend),
RISKOptimizer, OptQuest, SimRunner2, and WITNESS Optimizer 17 . Nearly all
of them use metaheuristics as for problems with large solution spaces, neither
ranking and selection methods nor enumeration are the optimization technique
of choice. “Since it becomes difficult to use a variable number of replications,
as needed in ranking and selection, with metaheuristics, one usually uses a large,
but fixed, pre-determined number of replications (samples) in evaluating the function at any point in the solution space” [Go2015, p. 89]. Metaheuristics are based
17 Details on the commercial software packages as well as on supported simulation software products can be found on the vendor’s websites [Au2019a; Au2019b; Pa2019; Op2019;
Pr2019; La2019].