Assistance for Simulation in Production and Logistics

Although discrete-event simulation has been useful for planning and operating production and logistics systems in industry for many years, there are still obstacles to its application today. These are primarily due to the high modeling effort, the lack of data or the insufficient data quality, the inadequate model validity for the intended purpose, and high experimental effort. In addition, the shortage of skilled workers means that there is a lack of qualified personnel to ensure consistent application. This is particularly important for the implementation of digital twins. For this reason, research has focused for years on reducing the effort involved in using simulation and making simulation studies more efficient.

This article analyzes, based on the literature, how the different phases of a simulation study can be supported, highlighting possible phase-specific assistance. The field of simulation-based assistance systems is not considered. The article concludes with a discussion of open research topics.

Brief definition of assistance and assistance systems

The term “assistance system” in the context of decision support is not new. As early as 2008, this term was the subject of an interdisciplinary discourse between computer science, logistics, and sociology as part of the Collaborative Research Centre SFB 559 “Modeling Large Logistics Networks”. Subsequently, in [1, p. 242], assistance systems were defined as “computer-based systems that support humans in decision-making and implementation”.

Furthermore, [1, p. 242] specifies that assistance systems for decision support are “characterized by the identification of a set of solutions, the selection and evaluation of alternatives, and autonomous action.” The authors add that the performance potential of assistance systems lies in their “ability to perform intelligent procedures that humans systematically fail to execute due to their limited cognitive abilities when faced with high levels of difficulty and (system) complexity” [1, p. 242].

Assistance systems are human-machine systems that have different levels of automation, depending on the extent of digital support. This results in different distributions of tasks and decisions between humans and computers. Assistance tools that support tasks involved in the conducting simulation studies in production and logistics are discussed in [2], among others. However, the focus here is on providing knowledge and assistance functions to reduce routine activities; decision support is only a secondary focus.

With the advancing use of artificial intelligence (AI) methods, decision-making processes are also increasingly being shifted to computers. However, in this article, the term “assistance” is deliberately defined more broadly than “assistance system” with regard to the supporting simulation studies. For the literature review, it encompasses both computer-based tools and research on effort reduction, such as procedure models and methodologies. This broadening of the term was chosen because computer-based implementations can be realized from initially non-computer-based assistance if the application is sufficiently accepted.

Assistance for discrete-event simulation

In order to investigate the range of assistance available for discrete-event simulation, a search is conducted in the relevant scientific simulation literature. For this purpose, the conference proceedings of the ASIM symposium “Simulation in Production and Logistics” and the “ASIM Symposium Simulation Technology SST” since 2010 are analyzed. Additionally, the production and logistics-related tracks of the Winter Simulation Conference since 2010, as well as all issues of the Journal of Simulation and Simulation Notes Europe since 2020 are included.

Based on the titles and abstracts, publications are identified whose results are classified in the above-defined areas of assistance or assistance systems. Further literature related to the current basic research projects “LOCsGEN – Generation of operating curves for order-picking systems under uncertainty” (DFG 544959882) and “KL4SiM – Automated generation of structural variants for simulation models of production and logistics using combinatorial logic” (DFG 511349842), which focus on improving the use of simulation, is also considered.

This approach initially identifies 69 scientific articles. Based on the aforementioned definition of assistance and the preliminary results of the initial research, the following search string is formulated for a systematic search in the Web of Science database:

(TI=(“discrete-event simulation”) OR TI=(“discrete event simulation”) OR AB=(“discrete-event simulation”) OR AB=(“discrete event simulation”)) AND (ALL=(“production”) OR ALL=(“manufacturing”) OR ALL=(“logistic“)) AND (ALL=(“standardization”) OR ALL=(“standardisation”) OR ALL=(“procedure model“) OR ALL=(“checklist“) OR ALL=(“guide“) OR ALL=(“knowledge base”) OR ALL=(“template“) OR ALL=(“automat“) OR ALL=(“ontolog“) OR ALL=(“machine learning”) OR ALL=(“LLM“) OR ALL=(“Large Language Model“) OR ALL=(“Large-Language Model“)).

The search period is limited to the years from 2010 onwards. This query initially returns 545 results, which are evaluated based on their titles and abstracts. Subsequently, 81 sources are selected from the database for full-text review.

In addition, there are 69 sources that were already identified in the initial search, so that a total of 150 scientific articles are evaluated on the basis of their full text. After reviewing the content, 73 sources dealing with assistance during the conduction of simulation studies remain. Where sources represent the same subject by the same team of authors, only the most recent source is included. These sources are listed in the concept matrix in Figure 1 and assigned to their corresponding simulation study phases. These phases are based on those of the procedure model for conducting simulation studies according to [3, p. 6], with the exception of the offer phase.

Additionally, system analysis, model formalization, and implementation fall under the umbrella of model building (MB). The two phases of data collection and data preparation are assigned to information and data acquisition (I/D). Verification and validation (V&V) of data and models does not constitute a separate phase, but runs in parallel with all other phases and is therefore also considered. The sources listed in Figure 1 are indexed in the following text with [ID 1]-[ID 73].

Assistance for model building

The number of publications on assistance in model generation (Fig. 1) indicates that efficiency gains in this particularly complex phase are constantly being researched, with a focus on reducing the effort involved in manual model building by using (semi-)automatic model generation (AMG).

The authors in [ID 8] define AMG as a group of approaches in which models are not generated entirely manually but rather in a data-driven manner using (simulation) external data sources. Computer-aided design (CAD) models are one example of such data source, on the basis of which simulation model layouts can be automatically generated or checked.

This article elaborates on the concept of externalized knowledge as the basis for AMG. For instance, [ID 25] presents an ontology-based software module for the automatic configuration and verification of model layouts. Meanwhile, [ID 30] presents an ontology-based approach to model versioning that enables the detection and recording of changes between different model versions, including a check for inconsistencies.

[ID 73] uses a semantic modeling approach for model generation and result visualization, while [ID 31] applies ontologies to promote interoperability between simulation models and other digital factory tools. [ID 16] and [ID 21] query ontologies to identify suitable resources for representation in a simulation model. This research demonstrates the maturity of semantic models in simulation as models of symbolic artificial intelligence (SAI).

Other approaches to assistance in the form of SAI use software synthesis based on combinatorial logic for AMG. An example of this can be found in [ID 42], where the authors use software synthesis to generate different machine configurations in model variants. They then use these simulation model variants to define joint system solutions for sales purposes. Other publications focus on automating the transformation of phase results during model generation. One example of this is [ID 55], in which the authors present an assistance system for AMG that derives an executable model from an existing (semi-formal) concept model (from the system analysis phase).

Building on this, recent research uses large language models (LLMs) to automatically generate model specifications and executable models based on natural language descriptions [ID 15, ID 33]. In particular, [ID 15] applies LLMs to create a “copilot” for model building. LLMs are machine learning (ML) models trained on data. Other data-driven approaches are taken to model building. For instance, [ID 22], [ID 41] and [ID 47] use process-mining algorithms to extract executable models or model parameters (semi-)automatically based on event data. [ID 50] extract model structures, including model topology, based on data from programmable logic controllers.

Another example of assistance during model building is the use of standardized interfaces, such as CMSD (Core Manufacturing Simulation Data), which were particularly popular during the 2010s. For example, [ID 7], [ID 8] and [ID 29] use CMSD-formatted data to reduce the effort involved in modeling elements such as resources, shift models, or orders. [ID 51] present standardized data structures, enabling users with limited knowledge of a specific simulation tool to build models.

The research presented suggests that assistance in the context of model generation already encompasses a range of artificial intelligence methods. Nevertheless, these do not yet constitute fully fledged assistance systems.

Assistance for information and data acquisition

In order to perform simulations successfully, relevant information and data must be obtained and processed. However, this often poses a considerable challenge due to missing data, limited access to data sources, and poor data quality [ID 38]. Approaches that support information and data acquisition (Fig. 1) therefore aim to reduce the effort associated with information and data acquisition, improve the quality of the input data, and consequently, enhance the quality of the simulation results.

Procedural models [ID 13, ID 18, ID 38] and software tools [ID 39] are already available to support information and data acquisition. [ID 38] expands the procedure model for conducting a simulation study [3] to include phases for information and data acquisition as well as consistent verification and validation. The individual steps range from defining objectives to identifying, collecting, and recording information to structuring and analyzing data and testing data usability.

Additionally, specialized software tools support data collection, preparation, and analysis across the various phases of the simulation study. For example, EDASim [ID 11] and AssistSim [ID 39] include auxiliary functions for selecting, validating, and preparing input data, and for analyzing output data, as well as for designing and executing experiments (see more below). These tools are now available on the market as a joint assistance system [4].

Another approach is to use an expert system that encompasses relevant simulation knowledge [ID 44]. The structured collection and management of expert knowledge according to a defined set of rules improves data collection and preparation and reduces the learning and training effort required for simulation. The expert system can be implemented in the form of a wiki, for example [ID 19]. Visual wikis, in particular, enable the collaborative development and maintenance of system descriptions and facilitate communication between the modeling and specialist departments [ID 19].

Furthermore, data repositories are used for data acquisition [ID 65], while value stream analysis is employed as an additional method of data capturing [ID 5]. The latter visualizes material and information flows, thus facilitating the identification of the required data and the categorization of existing information. This streamlines the data collection process and provides valuable insight into the system to be modeled.

In addition to assisting acquisition of information and data, other approaches focus on improving access to data and information. The authors of [ID 1] demonstrate how time parameters for simulation can be extracted from refined RFID data, thereby avoiding the effort associated with manual data collection. [ID 3] and [ID 8] present a tool for extracting simulation-relevant data from various operational information systems (e.g. ERP or ME systems), using statistical methods to extrapolate and thus prepare missing data.

The tool in [ID 59] offers similar functionality but it also formalizes data in CMSD format. This means that it can describe data independently of the simulation tool. [ID 34] presents a similar but less functional approach using a connector to transfer operational data to a simulation model.

[ID 62] and [ID 65] also present an approach to preprocessing data by applying statistical methods to investigate systems with a particularly wide range of product variants. Rather than changing the properties of a data point directly, their framework uses clustering to automatically select representative product families for simulation. Conclusions can then be drawn about the overall system based on these families. This reduces the effort involved in information and data acquisition, model building, and experiment execution.

Assistance for the “Experiments and analysis” phase

The “experiments and analysis” phase can be time-consuming, as it involves design and execution a large number of simulation experiments, followed by analysis. Assistance systems offer a variety of options to reduce the effort involved (Fig. 1). Tools such as the aforementioned AssistSim [ID 39] or commercially available tools like SimAssist [4] or SAFE [ID 49] aim to automate experiment design and execution to save time and reduce errors. The support covers everything from the design of the experiment plan with parameter configurations to the evaluation and visualization of results, which is particularly beneficial for inexperienced users. Further approaches to automating the design and execution of simulation experiments can be found, for example, in [ID 18] and [ID 68].

In addition, there are several approaches that support specific aspects of experiment design and execution. These include, for example, template-based simulation experiment specification [ID 56] and the automatic variation of model structures using combinatorial logic [ID 69, ID 70]. Unlike the approaches for model generation [ID 42], model structure variants are automatically generated based on a valid executable model, enabling automated structure variation. This enables a large number of model structure variants to be simulated and compared in terms of their quality of result.

Another important approach is statistical metamodeling [ID 32, ID 37, ID 71, ID 72]. Rather than carrying out a large number of complex simulation experiments, a simplified, approximate model – the metamodel – is built based on a few simulation runs and replications. This metamodel can then be used to investigate the behavior of the underlying simulation model in a broader parameter space.

Furthermore, distributed simulation [ID 43, ID 64] enables the modeling of large and complex production and logistics systems despite the associated computational load. Standards [ID 6, ID 64] are often used to promote the interoperability of distributed simulation models. In addition to web-based approaches for distributing simulation experiments, mobile devices can also be used [ID 43], enabling more flexible and accessible execution of simulation studies for inexperienced users.

In addition, the visualization of simulation experiments [ID 17] can facilitate analysis and communication of results, providing intuitive access for everyone involved. In this context, cloud-based assistance systems [ID 2, ID 66] also allow simulation experiments to be executed and accessed flexibly via mobile devices.

Assistance for verification and validation

V&V software functionalities are already offered in commercially available discrete-event simulation tools for checking the executable model (consistency checks) or the input and output data (functions for regression and correlation analysis or for calculating confidence intervals) [5]. Additionally, the tools directly support individual V&V techniques such as animation (for V&V techniques, see [6], among others).

However, V&V assistance is often application-specific and cannot be used consistently for all phases of the simulation process. For instance, [ID 23] presents a V&V assistance system for simulation models running in parallel with system operation in the semiconductor industry. [ID 47] offers a V&V framework based on operational data using process mining, but this can only be used during model building.

Assistance for re-use

Provided that the information and knowledge gained are documented accordingly, simulation studies provide valuable experience for the application of simulation in follow-up studies. For this purpose, [ID 61] presents an ontology-based methodology that supports the re-use of simulation knowledge and is based on a continuously expanding knowledge base.

Evaluation of the analysis and open research topics

The research presented here illustrates the diversity of assistance available for simulation studies, as well as the significant variation in the maturity levels and the lack of continuity. Tools are available to automate experiment design and execution, and to aggregate and prepare data in the information and data acquisition phase. In the modeling phases, various assistance approaches have emerged over the last 15 years, including the standardization of interfaces, the use of simulation tool-independent model formalisms, or statistical metamodeling. However, these approaches have not yet been consolidated.

Research on assistance for V&V and re-use is still in its infancy. Although procedure models and methodologies have been identified, the scope of the research is limited compared to the aforementioned phases. This offers an opportunity for further research, as consistent V&V is essential for carrying out reliable investigations and accelerating the systematic re-use of results.

Analysis of the scientific literature reveals that there are a variety of forms of assistance available for various tasks in the phases of a simulation study. These forms scan be implemented as checklists, procedure models, software functions, or even standalone tools. However, assistance systems for actual decision support (as defined at the beginning of this article) are still largely unavailable, despite the evident need for them and the associated research gaps in both science and application. There is a complete lack of continuity, cross-phase assistance systems. In order to close these gaps, the following fields should be prioritized for future assistance systems in the context of simulation:

1. Improving automated model generation
2. Ensuring that information gathered for simulation models is both quality-compliant and time-synchronous, based on real system data
3. Ensuring the validity of models and data through continuous, guided V&V assistance
4. Self-learning and error-detecting simulation models (see [7])
5. Natural language design, execution, and evaluation of experiments, as well as reducing experimental effort while improving the accuracy of results
6. Creating sustainable, reusable simulation knowledge, even after the project has ended
7. Rule-based enablement of interoperability between models and tools

Current research in the fields of statistical metamodeling, artificial intelligence, and theoretical models of computer science could be useful here (see [8] and [ID 15, ID 32, ID 70]). For instance, the use of LLMs offers potential for fields 1 and 3, while statistical metamodeling methods could be used in field 5.

As an additional starting point for field 5, theoretical models of computer science, such as those in [ID 70], offer expanded possibilities for automatically varying non-numerical parameters, such as structures or control strategies, in experiments. Methods of statistical data analysis and AI can be used for field 2, while machine learning methods can be used for field 4, and ontologies can be used for field 6.

The prevalence of combining simulation with other methods such as optimization, machine learning, or process mining, makes the interoperability of models and tools (field 7) increasingly important (for the implementation of digital twins, for example) and requires appropriate assistance, for example in the form of ontology-based “copilot” applications. When using AI methods, particularly machine learning methods such as LLMs, explainability (“Explainable Artificial Intelligence” or “XAI”) and, more broadly, trustworthiness (“Trustworthy Artificial Intelligence” [9]) are essential prerequisites for successful implementation.

Summary and outlook

Based on a literature analysis, this article identified various possibilities for assistance along the procedure model for discrete-event simulation in production and logistics in order to derive research gaps. The research shows that a variety of assistance options are already available to support simulation use. However, it also becomes clear that assistance systems for decision support in simulation have not yet been sufficiently researched in terms of their methodological foundations. Methods from the fields of AI and statistics that complement discrete-event simulation could open new avenues of decision support, thereby reducing the time and effort needed for simulation studies.

The original German version of this article can be accessed via DOI: 10.30844/I4SD.25.5.66

You might also be interested in

Digital Factory Planning for Startups A simulation-based production structure design Herwig Winkler , Tobias Isau

With the increasing complexity of production and logistics systems, traditional factory planning approaches are reaching their limits. In this context, digital factory planning offers a promising solution for enabling well-informed decisions, particularly during the early planning phases. For startups, the optimal planning of a production facility is challenging, as they often operate with limited financial and infrastructural resources. This paper presents a methodological approach to digital factory planning that utilizes VR simulation for the layout planning of a factory hall for a young company in the solar industry. The proposed approach demonstrates how simulations can support the design of flexible production structures, particularly in startup environments.

Industry 4.0 Science | Volume 42 | 2026 | Edition 3 | Pages 68-75

Site Assessment for Flexible Intralogistics in Brownfield Sites Innovative decision support in practice Jolanda Schierbaum, Carsten Feldmann, Lars Renhof

Due to its dynamic environment, space planning in intralogistics is not a one-time task but a recurring decision-making process subject to numerous constraints imposed by existing infrastructure. Decisions are often based on incomplete data, resulting in a high risk of poor planning decisions and inefficient use of space. This paper presents a practice-oriented process model for space evaluation in brownfield projects. The proposed approach improves the standardization and consistency of space evaluation and promotes best practices among all stakeholders. By supporting systematic decision-making, the process model contributes to optimized planning and resource allocation, thereby reducing risks and avoiding costly implementation errors.The process model is demonstrated through a case study conducted at a commercial vehicle manufacturer.

Industry 4.0 Science | Volume 42 | 2026 | Edition 3 | Pages 124-133

Experiencing Digital Twins in Production and Logistics The fischertechnik® Learning Factory 4.0 as a development platform for possible expansion stages Deike Gliem , Sigrid Wenzel , Jan Schickram, Tareq Albeesh

The fischertechnik® Learning Factory 4.0 has proven to be a suitable experimental environment for testing digital twins. Depending on the targeted maturity stage, the functions of a digital twin range from status monitoring and forecasting to the operational control of production and logistics systems. To systematically classify these functions, this article presents a maturity model that serves as a framework for the development of a digital twin. Building on this, selected use cases are implemented in a test and development environment based on a system architecture with multi-layered logic structure. These initial implementations serve to highlight application purposes, relevant methods, and typical challenges and potentials in the transfer to real factory environments.

Industry 4.0 Science | Volume 42 | Edition 2 | Pages 30-37 | DOI 10.30844/I4SE.26.2.30

Loam Construction and Wooden Shelving A contribution to sustainability in warehouse logistics Viviano De Giacomo , Nathalie Fritsch , Jakob Kennert , Dieter Uckelmann

This study examines the contribution of natural building materials, in particular loam and wood, to the sustainable development of logistics infrastructure, assessing ecological, economic, and technical dimensions across the entire life cycle. Potentials, restrictions, and supportive framework conditions are identified based on literature analyses and expert interviews. Wood proves to be technically mature and ecologically advantageous, especially in high rack construction, while loam offers high potential for energy- and resource-efficient construction. The study concludes with recommendations for research, policy, and practice to establish circular construction methods in logistics.

Industry 4.0 Science | Volume 41 | Edition 6 | Pages 82-89

Electric Trucks in Intermodal Terminal Pre- and Post-Carriage Impact on terminal processes in combined road-rail freight transport Ralf Elbert, Samira Ghaneian Sebdani

Electric trucks (e-trucks) play an important role in reducing CO₂ emissions especially on short distances in pre and post-carriage in combined road-rail freight transport (CT). Using the example of a CT terminal, this article highlights the logistical and energy challenges involved in using e-trucks to establish suitable charging infrastructures and ensuring a reliable power supply.

Industry 4.0 Science | Volume 41 | Edition 6 | Pages 70-77