28
2 Simulation-Based Optimization
problem-independent principles and schemes for the development and control of
heuristic processes which often reproduce natural processes, such as genetic algorithms (GA), simulated annealing (SA), or ant colonialization [SG2013, p. 960;
SM2009, p. 13; CT2018, p. 22]. To do so, the individual metaheuristics have
adjustable parameters, that allow the search process to be adapted to a particular class of problems, requiring a suitable calibration of the method [Ra2008,
p. 61]. Metaheuristics are split up in trajectory-based and population-based algorithms. While trajectory-based algorithms start with one possible solution and
improve this successively, population-based algorithms develop several solutions
simultaneously within each iteration and improve them in parallel [So2018, p. 58;
Go2015, p. 93]. Population-based algorithms are subdivided into evolution-based
and swarm-based algorithms. Evolution-based methods are inspired by the laws
of natural evolution, e.g., GAs emulate the Darwinian evolution theory. Swarmbased algorithms reflect the collective behavior manifested by animals, to mention
the Particle Swarm Optimization (PSO) Algorithm, depicting the behavior of birds
or the Ant Colonialization Algorithm, the Firefly Algorithm or the Bee Colony
Algorithm, imitating the social collective behavior of insects.
Figure 2.8 provides an overview of existing optimization methods according
to the classification previously described. Single algorithms will not be discussed
any further for reasons of space. The interested reader is referred to more extended
reference literature 15 for deeper professional involvement. The algorithms being
relevant for this work are discussed in more detail in section 2.3.3 in the context
of their use in different optimization software solutions.
2.3
Combination of Simulation and Optimization Methods
Following Fu, “the two most powerful operations research/management science
(OR/MS) techniques are simulation and optimization” [Fu2015, p. V]. The
incentive to couple simulation and optimization can be considered from two
15 While Cavazzuti [Ca2013, pp. 77–102] and Suhl and Mellouli [SM2009, pp. 33–134]
provide a good overview of the exact optimization methods, a detailed description is given by
Bhatti [Bh2000] and Marti, Pardalos and Resende [MPR2018]. Yang [Ya2018, pp. 125–
194], Domschke et al. [Do+2015] and Michalewicz and Fogel [MF2004, pp. 35–111]
present background information on the mentioned heuristics of OR. Chopard and Tomassini
give an accessible introduction to metaheuristics [CT2018], additionally, Shaheen et. al
provide an excellent overview on latest metaheuristics in [Sh+2018, pp. 215–231], and Siarry
conveys very extensive and deep explanations of the individual trajectory- and populationbased metaheuristics [Si2016].
2 Simulation-Based Optimization
problem-independent principles and schemes for the development and control of
heuristic processes which often reproduce natural processes, such as genetic algorithms (GA), simulated annealing (SA), or ant colonialization [SG2013, p. 960;
SM2009, p. 13; CT2018, p. 22]. To do so, the individual metaheuristics have
adjustable parameters, that allow the search process to be adapted to a particular class of problems, requiring a suitable calibration of the method [Ra2008,
p. 61]. Metaheuristics are split up in trajectory-based and population-based algorithms. While trajectory-based algorithms start with one possible solution and
improve this successively, population-based algorithms develop several solutions
simultaneously within each iteration and improve them in parallel [So2018, p. 58;
Go2015, p. 93]. Population-based algorithms are subdivided into evolution-based
and swarm-based algorithms. Evolution-based methods are inspired by the laws
of natural evolution, e.g., GAs emulate the Darwinian evolution theory. Swarmbased algorithms reflect the collective behavior manifested by animals, to mention
the Particle Swarm Optimization (PSO) Algorithm, depicting the behavior of birds
or the Ant Colonialization Algorithm, the Firefly Algorithm or the Bee Colony
Algorithm, imitating the social collective behavior of insects.
Figure 2.8 provides an overview of existing optimization methods according
to the classification previously described. Single algorithms will not be discussed
any further for reasons of space. The interested reader is referred to more extended
reference literature 15 for deeper professional involvement. The algorithms being
relevant for this work are discussed in more detail in section 2.3.3 in the context
of their use in different optimization software solutions.
2.3
Combination of Simulation and Optimization Methods
Following Fu, “the two most powerful operations research/management science
(OR/MS) techniques are simulation and optimization” [Fu2015, p. V]. The
incentive to couple simulation and optimization can be considered from two
15 While Cavazzuti [Ca2013, pp. 77–102] and Suhl and Mellouli [SM2009, pp. 33–134]
provide a good overview of the exact optimization methods, a detailed description is given by
Bhatti [Bh2000] and Marti, Pardalos and Resende [MPR2018]. Yang [Ya2018, pp. 125–
194], Domschke et al. [Do+2015] and Michalewicz and Fogel [MF2004, pp. 35–111]
present background information on the mentioned heuristics of OR. Chopard and Tomassini
give an accessible introduction to metaheuristics [CT2018], additionally, Shaheen et. al
provide an excellent overview on latest metaheuristics in [Sh+2018, pp. 215–231], and Siarry
conveys very extensive and deep explanations of the individual trajectory- and populationbased metaheuristics [Si2016].