20
2 Simulation-Based Optimization
The idea of combining different simulation paradigms is far from new as there
are references from the 1960’s and earlier. “Over the years ‘hybrid simulation’
has meant a number of things: models that are simultaneously implemented on
both analogue and digital computers, or models that contain both discrete and
continuous variables, or even models that combine simulation with an analytical
method such as optimization” [Br+2019, p. 2]. Examples for the combined use
of DES and SD can be found in Zeigler’s and Oeren’s “futuristic simulation
environments which support flexible adoption of multiple perspectives” [ZO1986,
p. 708], Pritsker’s approach to simulate the loading and unloading processes of
oil tankers [Pr1995, pp. 354 ff.], or Mosterman with his models, having continuous and discrete proportions that can be switched between to adjust the level
of detail [MB1997; Mo1999]. Even though DES and SD have been established
methods for more than 60 years, “[…] the combination of continuous-time simulation and DES is still a challenging area of research” [BFG2015, p. 139]. The
youngest of the three simulation paradigms, ABS, has become popular within the
simulation community in the last decade.
Several approaches combining ABS and SD 9 or ABS and DES 10 pairwise as
well as ABS, DES and SD 11 together in one approach can be found in literature.
ABS has proven to be a useful tool to model autonomous agents and their interactions in complex systems to illustrate the agent’s behavior in detail [Kh+2015,
p. 1491].
This thesis follows the definition of hybrid simulation according to
Brailsford, who states that “hybrid simulation is one single conceptual model
9 Examples on the combined use of ABS and SD can be found in Djanatliev et al., who
develop a combined approach to assess health care technologies to learn about impacts of new
products before product releasement [Da+2014], in Milling, who is building a SD model to
analyze diffusion patterns in combination with an ABS to model the occurrence of innovations
and the adaptation over generations [Mi2002], as well as in Schieritz and Grössler who
combine SD and ABS for reducing the a priori complexity of supply chain models [SG2003].
10 The combination of ABS and DES is used in Dubiel and Tsmihoni to model human
agents traveling freely through a DES environment [DT2005], in Fakhimi et al. for analyzing sustainable planning strategies for emergency medical service [Fa+2014], in Liraviasil
et al. to simulate efficient manufacturing processes under change [Li+2015], and in Nguyen
et al. for the strategic aircrew manpower planning [Ng+2017]. Khedri Liraviasl et al.
use ABS and DES to model assembly systems in production for depict interchangeable and
reconfigurable production layouts [Kh+2015].
11 Examples on the combined use of all three simulation techniques are found in Djanatliev
and German to predict the effects of medical products [DG2013] and in Konstantinos
and Angelopoulou who present a modeling framework for combining DES, SD and ABS
approaches [MA2019].
2 Simulation-Based Optimization
The idea of combining different simulation paradigms is far from new as there
are references from the 1960’s and earlier. “Over the years ‘hybrid simulation’
has meant a number of things: models that are simultaneously implemented on
both analogue and digital computers, or models that contain both discrete and
continuous variables, or even models that combine simulation with an analytical
method such as optimization” [Br+2019, p. 2]. Examples for the combined use
of DES and SD can be found in Zeigler’s and Oeren’s “futuristic simulation
environments which support flexible adoption of multiple perspectives” [ZO1986,
p. 708], Pritsker’s approach to simulate the loading and unloading processes of
oil tankers [Pr1995, pp. 354 ff.], or Mosterman with his models, having continuous and discrete proportions that can be switched between to adjust the level
of detail [MB1997; Mo1999]. Even though DES and SD have been established
methods for more than 60 years, “[…] the combination of continuous-time simulation and DES is still a challenging area of research” [BFG2015, p. 139]. The
youngest of the three simulation paradigms, ABS, has become popular within the
simulation community in the last decade.
Several approaches combining ABS and SD 9 or ABS and DES 10 pairwise as
well as ABS, DES and SD 11 together in one approach can be found in literature.
ABS has proven to be a useful tool to model autonomous agents and their interactions in complex systems to illustrate the agent’s behavior in detail [Kh+2015,
p. 1491].
This thesis follows the definition of hybrid simulation according to
Brailsford, who states that “hybrid simulation is one single conceptual model
9 Examples on the combined use of ABS and SD can be found in Djanatliev et al., who
develop a combined approach to assess health care technologies to learn about impacts of new
products before product releasement [Da+2014], in Milling, who is building a SD model to
analyze diffusion patterns in combination with an ABS to model the occurrence of innovations
and the adaptation over generations [Mi2002], as well as in Schieritz and Grössler who
combine SD and ABS for reducing the a priori complexity of supply chain models [SG2003].
10 The combination of ABS and DES is used in Dubiel and Tsmihoni to model human
agents traveling freely through a DES environment [DT2005], in Fakhimi et al. for analyzing sustainable planning strategies for emergency medical service [Fa+2014], in Liraviasil
et al. to simulate efficient manufacturing processes under change [Li+2015], and in Nguyen
et al. for the strategic aircrew manpower planning [Ng+2017]. Khedri Liraviasl et al.
use ABS and DES to model assembly systems in production for depict interchangeable and
reconfigurable production layouts [Kh+2015].
11 Examples on the combined use of all three simulation techniques are found in Djanatliev
and German to predict the effects of medical products [DG2013] and in Konstantinos
and Angelopoulou who present a modeling framework for combining DES, SD and ABS
approaches [MA2019].