When business processes meet complex events in logistics: A systematic mapping study

Computers in Industry 144 (2023) 103788 Available online 14 October 2022 0166-3615/© 2022 The Author(s). Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/bync-nd/4.0/). Contents lists available at ScienceDirect Computers in Industry journal homepage: www.sciencedirect.com/journal/computers-in-industry When business processes meet complex events in logistics: A systematic mapping study Belén Ramos Gutiérreza,Antonia M. Reina Quinteroa,∗,Luisa Parodyb, María Teresa Gómez Lópeza aDpto. Lenguajes y Sistemas Informáticos, Universidad de Sevilla, Spain bDpto. Métodos Cuantitativos, Universidad Loyola Andalucía, Spain ARTICLE INFO Keywords: Logistics Event-driven business process Complex event processing Business process Systematic mapping study Systematic literature review ABSTRACT Logistics processes are attracting growing attention because of the globalisation of the market. Its growing complexity and the need for reducing costs have provoked the seek of new solutions based on the processing of the complex events that the business processes produce. Event-Driven Business Process Management (EDBPM) is a discipline that studies the integration of business processes and complex events. The analysis of the maturity level of the approaches and gaps to point out future lines of research could help not only logistics organisations, but also academia. Logistics organisation could benefit from producing more environmentally friendly and optimal solutions in transport, and academia could benefit from revealing open problems. Thus, this study aims to identify current approaches, frameworks, and tools that integrate business processes and complex events in the logistics domain. To do so, we follow a systematic approach to do a mapping study that captures and synthesises the approaches, frameworks, and tools that integrate these two fields. As a result, 10,978 articles were gathered and 169 of them were selected for extraction. We have classified the selected studies according to several criteria, including the business process life cycle in which they are being applied, the business process modelling language, and the event process modelling language, among others. Our synthesis reveals the open challenges and the most relevant frameworks and tools. However, there is no mature enough framework or tool ready to be used in companies, and a promising research must provide solutions that cover all phases in the process life cycle. 1. Introduction Product exchange and globalisation of the market have caused an increase in logistics processes as a key part of supply chain management. However, the rise in shipping rates and the cost of cargo containers, the increase in fuel cost, the lack of wooden pallets, the clogging of containers in port yards, or the low availability of storage make the logistics problem a global concern that Forbes has classified as much bigger than the pandemic (Broadman,2021). To solve these problems, Information and Technology (I&T) is playing an important role in supply chain organisations, as stated in the Gartner report (Klappich et al., 2020). This report claims that ‘‘by 2023, 50% of global product-centric companies will invest in real-time transportation visibility platforms’’. In this globalisation context, logistics processes might be very complex since they can choreograph various entities, including resources and time restrictions. Furthermore, the collaboration of various entities, which may be geographically distributed, generates an enormous ∗Corresponding author. E-mail addresses: [email protected] (B. Ramos-Gutiérrez), [email protected] (A.M. Reina Quintero), [email protected] (L. Parody), [email protected] (M.T. Gómez-López). number of events that are complexly related. As a consequence, the extraction of knowledge from logistics processes implies the analysis of the processes that choreograph the organisations that collaborate, and the analysis of single events to derive other events defined at a higher level of abstraction, known as complex events. Traditionally, Business Process Management (BPM) and Complex Event Processing (CEP) are the two disciplines that face the problems related to business processes and complex events separately. However, recently, Event-Driven Business Process Management (EDBPM) (Ammon,2009a) has emerged as a new discipline to integrate both, in such a way that the events generated through the BPM systems can be paralelly analysed by a CEP taking the advantages of both disciplines. On the one hand, Business Process Management (BPM) (Weske,2012) is the most valuable corporate asset, and according to the Business Process Management Market Research Report (Future,2021), the ‘‘BPM market is expected to grow to approximately USD $16 billion by https://doi.org/10.1016/j.compind.2022.103788 Received 21 January 2022; Received in revised form 22 September 2022; Accepted 24 September 2022 Computers in Industry 144 (2023) 103788 2 B. Ramos-Gutiérrez et al. 2023’’. BPM represents an integration of technologies and methodologies that facilitates the modelling and deployment of process models by coordinating a set of activities choreographed to achieve the objectives of an organisation (Pérez-Álvarez et al.,2018). Complex Event Processing (CEP) (Luckham,2012) refers to a set of concepts and principles for processing events and the methods for implementing those concepts. The advantages of integrating both BPM and CEP are widely known (Eyers et al.,2016), and have been considered in logistics (Ammon et al.,2008;Schiefer et al.,2007;Emmersberger et al.,2009), due to the complexity and heterogeneity of the systems and the types of events that are handled in this context. However, there is no comprehensive overview of the state of research on Event-Driven Business Process Management in the logistics context. The aim of this paper is to analyse the state-of-the-art of EDBPM in the logistics context to understand how business processes and complex events are being integrated into logistics processes. To do so, we conducted a systematic mapping study that aims at giving an overview of the research area through classification and counting contributions in relation to the categories of that classification (Khan et al.,2019). The results of this mapping study are not only valuable for organisations because they can be aware of the most mature approaches, frameworks, and tools, but also for researchers because they reveal the gaps that can drive future research lines. Therefore, if an organisation with logistics management wants to leverage the level of digitalisation of its processes, our study can provide information on how many approaches, techniques, and tools exist to support the deployment of business processes through complex event processing in the logistics context. The rest of the paper is structured as follows: Section 2introduces the background needed to understand the analysis; Section 3presents the related work; Section 4details the method we used in our systematic mapping study; Section 5presents the data extracted following the method previously mentioned; in Section 6, the research questions are discussed; in Section 7the threads to validity are analysed; and finally, Section 8draws conclusions. 2. Background This section defines the concepts needed to fully understand this study according to the main areas of research that frame this work: logistics, business processes and complex events. 2.1. Logistics The importance of supply chain management has increased drastically during the last decades, especially with the necessity of coordinating various entities. While supply chain is related to the more general actions related to coordinate and manage items and persons in an organisation, the term logistics is used to refer to the set of activities that describe the flow of items in a company or its interchanging between various of them. In this context, a logistics system (Ghiani et al.,2013) describes the activities that determine the flow of the materials and information among the facilities of the organisations that can be in different places. It includes the infrastructure, equipment, means and resources (including humans) necessary for performing the activities. The capacity of modelling (Bassil et al.,2004) and reasoning over these systems (Chow et al.,2005) can improve performance in real scenarios. For this reason, the advantages that business process management can offer make them prone to logistics systems. 2.2. Business processes Business process management is a mechanism for modelling, observing and improving the activities developed in an organisation to achieve their goals according to a set of constraints that govern their behaviour (Gómez-López et al.,2015). Process modelling is the first step to understand how the activities performed by organisations, both manually and supported by information systems, are related. According to Weske (2012), a business process1is composed of a set of activities belonged from one or over one organisation. These activities are performed in a coordinated way to achieve a business goal. Thus, business process management (Dumas et al.,2013) gathers the techniques and methods to support the life cycle of business processes. Furthermore, according to Augusto et al. (2019), the business process life cycle is composed of different phases: identification, discovery, analysis, redesign, implementation, and monitoring and controlling. For modelling business processes in the phases of the life cycle different languages can be used, both imperative and declarative. Thus, abusiness process modelling language is a language oriented towards the description of a set of rules to govern how the elements involved in a process can be combined (Dumas et al.,2013). One of the most widely used business process modelling language is BPMN (OMG, 2013), although others can be used, such as Petri-nets (van Hee et al., 2013), YAWL (Ter Hofstede et al.,2009), Declare (Pesic and Van der Aalst,2006), BPEL (OASIS,2007), GSM (Guard Stage Milestone) (Hull et al.,2010) and UML Activity Diagram (Force,2001). 2.3. Complex events The complexity of logistics processes is derived from the multiple entities that can be involved, the distribution of the processes, and the different levels of abstraction of the generated events. This complexity makes necessary the analysis of the single events produced to infer more complex actions. These complex actions are represented by more than one event, known as complex events. In computer science, an event can be defined as ‘‘anything that happens, or is contemplated as happening’’ (Chandy et al.,2011). However, the definition of event varies depending on the context (BPM or CEP). From a business point of view, an event is an action or occurrence that affect the business, and that happens in a timestamp that may be handled and stored by a software system. Events can include attributes, such as time at which they happened, resources or location (Luckham,2012). Furthermore, in the context of business processes, event logs used to be represented by using the XES standard (eXtensible Event Stream).2Event logs are formed of a set of traces that contains a sequence of events. The term complex event processing (CEP) (Luckham,2005) gathers a set of technologies to discover relationships between single events to infer the existence of relevant information, such as timing, causality, membership, or existence of patterns of behaviour or correlation (Schiefer and McGregor,2004). To describe the type of events and the attributes they hold, we need an event model, and for processing them an event processing language is necessary. Thus, an event model is the mechanism for describing the events produced by the involved systems (Cugola et al.,2015), that includes a set of attributes 𝐴𝑡 and a set of domains 𝐷for each attribute, {A𝑡1;𝐷1,…, A𝑡𝑛;𝐷𝑛}. Whereas, an event processing language (EPL) is a high-level declarative language that permits defining functions to manage events within a data stream. 1There is a wide range of definitions of business process in literature (Davenport,1993;Hammer and Champy,1993;Johansson et al.,1993). We have chosen the definition proposed by Weske because it is the most recent and one of the most relevant taking into account the number of citations. 2https://www.xes-standard.org/. Computers in Industry 144 (2023) 103788 3 B. Ramos-Gutiérrez et al. The declarative capacity of the EPL facilitates the description of complex event conditions, event correlations, and the inclusion of possibly spanning time windows (Luckham,2008). The discipline that integrates business processes and complex events is named event-driven business process management. Event-Driven Business Process Management (EDBPM) (Ammon,2009b;Goetz, 2010) combines both Business Process Management (BPM) and Complex Event Processing (CEP) techniques to support the actions performed in a business process that originates complex events. In this context, the BPM engine generates events during the execution of the daily processes, that are analysed by the CEP in a parallel way, with the objective to detect the relevant information. These events can also influence the execution or the process and give as result other events. 3. Related work We consider as related work the secondary studies focused on the combination of complex events and business processes in the logistics context. Secondary studies are those studies that summarise and classify the published literature, in contrast to primary studies that present a primary work and are the focus of the analysis performed in secondary studies. Secondary studies in software engineering can be classified into surveys, systematic literature reviews, and systematic mapping studies. A survey is a kind of secondary study that is not conducted systematically nor with a systematic protocol (Khan et al.,2019). On the contrary, systematic literature reviews and systematic mapping studies follow a systematic approach by following a systematic protocol. One of the fundamental differences between systematic literature reviews and systematic mapping studies is their goals (Kitchenham et al.,2010, 2015). While the goal of systematic literature reviews is to address very specific questions, systematic mapping studies aim at giving an overview of a research area through classification and counting contributions in relation to the categories of that classification (Khan et al., 2019). None of the analysed secondary studies has been conducted as a mapping study and only a few analyse the integration of business processes and complex events in the area of logistics. To analyse these secondary studies, we have classified them into systematic literature reviews, surveys, and other secondary studies. The following subsections analyse in detail these groups of studies. 3.1. Systematic literature reviews The research questions of systematic literature reviews are more focused compared to the questions proposed in mapping studies. We have found various systematic literature reviews focused on some aspects of the relationship between complex events and business processes (Krumeich et al.,2014;Amjad et al.,2018;Augusto,2020). These studies explore how events drive the management of business processes (Krumeich et al.,2014), how business requirements can be modelled and validated with event-driven processes (Amjad et al., 2018), and the accuracy and efficiency of process mining in the context of event-driven processes (Augusto,2020). However, these studies do not give an overview of how, within the life cycle of processes, events can be applied or affected, and of course, how their study can be applied to the field of logistics. Falco (Jaekel,2019) conducts a systematic review of the existing literature in the area of cloud logistics. He selects 83 studies and classifies them according to the meaning of the term ‘‘cloud logistics’’. However, the logistics encompasses much more than what is covered by the term cloud logistics, so this study is insufficient for answering the research questions we proposed. 3.2. Surveys Surveys are another type of secondary study. In general, surveys are less rigorous than mappings and systematic literature reviews. We have found a set of surveys related to CEP, BPM, and logistics (Davidsson et al.,2005;Li,2005;Fulop et al.,2010;Akila et al.,2016;Li et al., 2017;Dayarathna and Perera,2018). As in some of the aforementioned studies, the surveys have not been conducted systematically, and as a result, neither the research questions nor the identified studies are explicit, so they are difficult to replicate. Davidsson et al. (2005) presents a consistent view of the research efforts made in freight transportation and some work of traffic and transport of people, but the more generic aspects of logistics, such as the organisational part, external factors (traffic, weather, ...), or even the part of supply chain management are not included. Furthermore, they do not consider the role of CEP and BPs in logistics. In Li (2005), Li surveys the recent progress of event-driven applications and investigates their potential implications at the system and middleware levels. On the one hand, this study focuses on the military and intelligence community and the logistics is not included. On the other hand, activities are not contextualised within a business process. Fulop et al. (2010) examine the complex event processing and the related field of predictive analytic. The survey includes the terminology, research achievements, existing solutions, and open issues related to both areas. In Akila et al. (2016), Akila et al. analyse the techniques, challenges, and future directions of complex event processing over uncertain events. Lastly, Dayarathna and Perera (2018) summarise the latest cutting-edge work in 2018 done on event processing system architectures, event processing use cases, and event processing open research topics. In these three surveys, the focus is on events without considering business processes or logistics. Finally, Li et al. (2017) presents a variety of data-driven techniques and applications with a focus on computing system management. 3.3. Other secondaries studies This group of studies includes secondary studies that cannot be formally classified as mapping studies, systematic literature reviews, or surveys. There are several points that differentiate the studies in this group from ours. Firstly, neither of the studies deals with the combination of CEP and BPM in the logistic context. Alias et al. (2016) focus on the combination of events with logistics, while Soffer et al. (2019) focus on the combination of events with processes. Secondly, the number of studies they analyse is considerably smaller than the number of studies we have considered in our mapping study. In addition, the rigour of such mappings precludes the exact replication of the studies. The mapping study by Alias et al. (2016) aims at determining the updated status in 2016 of complex event processing and predictive analytics in the transportation and logistics sector. One of the major problems with this study is that it does not follow a formal method for conducting the mapping. Among other things, they do not specify exactly what terms have been used to perform the searches, making the replication of the study difficult. They identified 58 studies that are categorised into the areas of transport, logistics, and supply chain management. However, they do not mention any relation with the business process technology. Soffer et al. (2019) focus on the challenges and opportunities for the combination of CEP and process mining. Although they illustrate and motivate the study through a logistics example, the focus of the study is not the logistics sector. Furthermore, in relation to replicability, they do not present the number of articles analysed, nor the sources used, nor what criteria they use to select the articles. Finally, as their research is focused on process mining, they do not cover the studies that are framed on the rest of phases of the business process life cycle. Computers in Industry 144 (2023) 103788 4 B. Ramos-Gutiérrez et al. Table 1 Research questions. Id Research question RQ1 What are the demographics of the published studies? RQ1.1 Which contributions were made over the years? RQ1.2 Which are the most influential researchers in the area and were are they from? RQ1.3 Which are the most influential studies? RQ1.4 Which are the top venues? RQ2 What are the current trends in the area? RQ2.1 What is the type of contribution made by the study? RQ2.2 Which are the application areas of the studies? RQ2.3 What type and which business process modelling languages are utilised? RQ2.4 What type and which event processing languages are utilised? RQ2.5 How is logistics covered in the studies? RQ2.6 Which are the event producers? RQ3 What are the potential gaps in the area? RQ3.1 Is event-driven business process treated in every phase of the process life cycle in a logistics context? RQ3.2 What are the challenges taken off in the field? 4. Method The main purpose of our study is to analyse the state-of-the-art of integrating business processes and complex events in the logistics context. Our goal is to identify the current approaches, frameworks, and tools, their maturity level, and the areas not explored in depth to point out future lines of research. Thus, the principal objectives of this mapping study are as follows: OBJ1: To synthesise the studies in the logistics domain with approaches related to event-driven business processes to provide insight into this area. OBJ2: To investigate the trends in event-driven business processes in the logistics domain. OBJ3: To reveal the gaps and foresee future directions of research. To conduct the mapping study, we have followed the guidelines proposed by Kitchenham and Charters (2007), Kitchenham et al. (2015) and Kuhrmann et al. (2017) for performing systematic literature reviews in software engineering. And the guidelines for the use of snowballing by Wohlin (2014). The review process consists of three main stages: planning, conducting, and documenting the review. Fig. 1 depicts these stages and the principal activities involved in each phase. The protocol is defined during the planning stage (phase 1), which includes assigning responsibilities and activities to the reviewers. The result of this stage is the protocol. The activities defined in the protocol are executed in the conducting stage (phase 2). These activities include identification, selection, assessment of the quality of primary studies, data extraction, and synthesis of the extracted data. Since phase 2 is the stage that comprises more activities, we have detailed it in Fig. 2 by means of a process specified with the BPMN notation (OMG,2013). Fig. 2 describes the activities performed in Phase 2, in which the different search engines are searched. Duplicated studies are removed from the set of studies to be analysed. The rest of the studies are submitted for first screening and, if they are related to the objectives of the study, for a later careful read. After this analysis, the researchers discuss the selection to arrive at a consensus on possible doubts. The selected studies are used to perform a backward and a forward snowballing, that is, the studies that are included in the references and the studies that cite a selected study are included in the set of studies to be analysed. After that, each selected study is used to extract the required data, and its quality is assessed. Eventually, the extracted data is synthesised to facilitate later knowledge extraction. Finally, in phase 3, a report is produced to distribute the results. Fig. 1. Main stages for conducting the SMS according to Kitchenham and Charters (2007). 4.1. Research questions The specification of the research questions is an essential part of the conduction of a systematic mapping study because they drive the entire review process (Kitchenham et al.,2015). As we have developed a mapping study, some research questions are broad and are concerned with classifying the literature. The questions are aligned with the three objectives mentioned previously. We have defined three broad fundamental questions (R1-R3) that are based on the categories of research questions identified by Khan et al. in Khan et al. (2019): demographics, current trends, and research gaps. Additionally, each question has been broken down into some secondary questions. Table 1 lists the research questions addressed in this study. Question RQ1 belongs to the demographics category and aims to characterise the studies published in the area. This question has been broken down into sub-questions RQ1.1-RQ1.4. These subquestions will help us identify the number and frequency of publications, the most influential researchers and their countries, the most influential studies, and the top venues. Question RQ2 belongs to the category of current trends and aims to identify the nature of the existing work in the area. We have broken this question down into sub-questions RQ2.1-RQ2.6 that help us to identify the type of contributions, the application areas, the business process modelling and event processing languages, the logistics coverage, and the event producers. Finally, RQ3 belongs to the research gaps category and aims at the identification of gaps in the area. This question has been broken down into sub-questions RQ3.1- RQ3.2. The rationale behind these sub-questions is the identification of the process life cycle phases in which the approaches are framed, and the identification of the challenges in the area. 4.2. Search process To collect the studies, we used a broad automated search followed by an automated backward snowballing and a later forward snowballing. Table 2 lists the digital libraries and search engines used as data Computers in Industry 144 (2023) 103788 5 B. Ramos-Gutiérrez et al. Fig. 2. Phase 2: Process definition for conducting the review. Fig. 3. Proportion of studies recovered in relation to the data source. Table 2 Digital libraries and search engines employed as data sources. Search engine URL ACM DL http://dl.acm.org Springer http://www.springer.com IEEE Xplore http://ieeexplore.ieee.org ScienceDirect http://www.sciencedirect.com Google Academic http://scholar.google.es Scopus http://www.scopus.com sources during the automated search process. Furthermore, to arrive at proper search strings, we have followed a ‘‘Trial-and-error Search’’, as recommended in Kuhrmann et al. (2017). The terms resulting from this approach are the following: •‘‘complex event’’ AND ‘‘business process’’ AND ‘‘logistics’’ •‘‘event-driven business process’’ AND ‘‘logistics’’ •‘‘complex event’’ AND ‘‘workflow’’ AND ‘‘logistics’’ As our goal is to have an overview of how event-driven business processes are used in the logistics domain, we have selected the following keywords: complex event and business process. Furthermore, since the term workflow is used as a synonym of business process in this context (von Ammon,2018), we have also included workflow as a search string. The queries to collect the studies from the digital databases and search engines were executed in December 2020, and later, in March 2022, a second round has been carried out to update the searches. It deserves to be mentioned that the searches in ACM DL, Springer, IEEE Xplore, Science Direct and Scopus have been performed through the Web interface, while searches in Google Academic have been performed using Publish or Perish.3This tool returns a maximum of 1000 results for this search engine. As search engines have different query syntaxes, the previous search strings must be rewritten for every search engine. Table 3 lists the query strings used in each search engine. Note that a Search ID identifies every search. This ID is automatically generated by the Start tool,4 which is the tool that we have used as a support and helps us trace the origin of the recovered study. Regarding snowballing, we have performed an automatic extraction of the references of the selected studies (the so-called backward snowballing) following a two-step approach: first, we have queried Crossref5 with a Python-based script to extract references6; second, if the study is not indexed in Crossref, we have used GROBID (2008–2021) to extract the references directly from pdf files. The forward snowballing has been performed by querying Opencitations7with a Python-based script,8and has been complemented by manually querying Scopus. As a result of the search process 10,978 studies have been found. Fig. 3 summarises the number of studies obtained from the data sources, as well as the number of studies obtained with the backward and forward snowballing process. Note that more than the 61% of the incorporated studies have been obtained thanks to the snowballing process. 4.3. Study selection Once the studies have been recovered from the data sources, they are loaded into the Start tool, which helps us detect some duplicates (see the first activity in Fig. 2). As the tool only detects duplicates by checking the exact match of titles, some duplicates are marked manually. After that, an initial screening is performed based on title, abstract, and keywords. The initial screening of the 3450 studies obtained as a result of the automatic searches in digital databases is undertaken by all researchers. The screening of the 7528 obtained with backward and forward snowballing was also undertaken by all researchers.9We kept a full record of the researchers who conducted the initial screening of the study. Table 4 lists the inclusion and exclusion criteria used in the review. Note that some of these criteria are de facto standards, as noted in Kuhrmann et al. (2017). As a result of the screening, 621 studies have been selected. These studies have been thoroughly reviewed by all researchers. Note that the researcher who performed the screening of one study is not in charge 3https://harzing.com/resources/publish-or-perish. 4http://lapes.dc.ufscar.br/tools/start_tool. 5https://www.crossref.org/. 6https://github.com/reinaqu/bibreferences. 7https://opencitations.net/. 8https://github.com/reinaqu/snowballing-opencitations. 9See Section 7to see who did what. Computers in Industry 144 (2023) 103788 6 B. Ramos-Gutiérrez et al. Table 3 Search queries employed in the different search engines. Source SearchIDs Search string ACM DL SEARCH0 [Full Text: ‘‘complex event’’] AND [Full Text: ‘‘business process’’] AND [Full Text: logistics] SEARCH6 [Full Text: ‘‘event-driven business process’’] AND [Full Text: logistics] SEARCH7 [Full Text: ‘‘complex event’’] AND [Full Text: workflow] AND [Full Text: logistics] Google Scholar SEARCH5 ‘‘complex event’’+‘‘business process’’+ logistics SEARCH16 ‘‘event-driven business process’’+ logistics SEARCH17 ‘‘complex event’’+‘‘workflow’’+ logistics IEEE Xplore SEARCH3 (((‘‘Full Text & Metadata’’:‘‘complex event’’) AND ‘‘Full Text & Metadata’’:‘‘business process’’) AND ‘‘Full Text & Metadata’’:logistics) SEARCH10 ((‘‘Full Text & Metadata’’:‘‘event-driven business process’’) AND ‘‘Full Text & Metadata’’:logistics)) SEARCH11 (((‘‘Full Text & Metadata’’:‘‘complex event’’) AND ‘‘Full Text & Metadata’’:workflow) AND ‘‘Full Text & Metadata’’:logistics) Science Direct SEARCH2 ‘‘complex event’’+‘‘business process’’+ logistics SEARCH12 ‘‘event-driven business process’’+ logistics SEARCH13 ‘‘complex event’’+‘‘workflow’’+ logistics Springer SEARCH1 ‘‘complex event’’+‘‘business process’’+ logistics SEARCH8 ‘‘event-driven business process’’+ logistics SEARCH9 ‘‘complex event’’+‘‘workflow’’+ logistics Scopus SEARCH4 TITLE-ABS-KEY (‘‘complex event’’ AND ‘‘business process’’ AND logistics) SEARCH14 TITLE-ABS-KEY (‘‘event-driven business process’’ AND logistics) SEARCH15 TITLE-ABS-KEY (‘‘complex event’’ + ‘‘workflow’’ + AND logistics) Table 4 Exclusion and inclusion criteria. Criteria Description Exclusion The study is not written in English. It is a proceeding book.a It is an index or a table of contents, not a book chapter. It is an editorial. It is not a paper, but the slides of a presentation. It is a call for paper, not a paper. It is an encyclopaedia entry, not a paper. It is an extended abstract. The study occurs multiple times in the result set. The study is not accessible electronically. It is the result of an importation error. Title, keyword list, and abstract make explicit that the paper is NOT related to logistics, business processes, or complex events. Inclusion Title, keyword list, and abstract make explicit that the paper is related to logistics, business processes and complex events. aThis criteria helps to avoid duplicates due to the behaviour of some search engines, which returns two results when searching for a study: one that refers to the single study, and another one that refers to the whole book of the proceedings in which the paper was published. of reading it carefully and extracting the information of that same study. Furthermore, if one reviewer has doubts about the acceptance or rejection of a study after a thorough review, another researcher reviews that study, and the final decision on the acceptance of that study is taken in a voting workshop in which all researchers are involved. That is, we have followed an approach similar to the alternative approach proposed in Kuhrmann et al. (2017). Once a study is selected for extraction, a backward snowballing of the references included in the study and a forward snowballing of the papers that cite the study is made. As Start does not work properly with the references obtained through snowballing, we have set up a procedure to add the references to the tool and to maintain the traceability of the snowballing process. On the one hand, the references/citations obtained automatically from a study are included in a BibTex/RIS file that is loaded in a new search session in Start. The session has as a keyword the 𝑖𝑑 of the study that includes the references. On the other hand, we have used a spreadsheet to annotate the 𝑖𝑑 of the search session, the 𝑖𝑑 of the study, and the number of references recovered. As a result of the selection, 169 studies are collected. Subsequently, their quality is assessed as explained in Section 4.4, and the relevant information is extracted for our mapping study (see Section 4.5). Fig. 4 Table 5 Assessment of Intrinsic IQ. Category Intrinsic IQ output J1 ∨C1 HIGH J2 ∨C2 MEDIUM J3 ∨C3 ∨O1 ∨O2 ∨O3 LOW shows a diagram and a Start screenshot that summarises the results obtained during the three principal activities carried out in the review stage: identification, selection, and extraction. 4.4. Quality assessment The quality assessment of the studies has been carried out in parallel with the data extraction activity, because, as in Varela-Vaca and Reina Quintero (2021), we have used an Information Quality framework (Wang and Strong,1996) to assess the studies. The quality assessment depends on the type of publication and the quality of the venues and the amount and completeness of the extracted data. The framework (Wang and Strong,1996) defines four dimensions of Information Quality, namely: Intrinsic, Contextual, Representational, and Accessibility. Information Quality (IQ) can be measured regarding one or several dimensions depending on the extracted attributes. As the format of the data and the access to information are irrelevant in our context, we measure only Intrinsic and Contextual IQ. Intrinsic IQ (Wang and Strong,1996) measures the accuracy, objectivity, believability, and reputation of the data for the task at hand. Thus, to measure this dimension, we take into consideration the type of publication, the venue, and the ranking of the venue, and as a result, a value is obtained based on a three-point Likert scale with HIGH, MEDIUM, and LOW values. On this scale, HIGH is the best result, and LOW is the worst result. To calculate the value, first, the study is classified into one of the categories shown in Fig. 5, and then the value is obtained according to the category (Journal, Conference/Workshop or Others), as shown in Table 5. Afterwards, to obtain the value of the Intrinsic IQ, we have to look after the category in Table 5, and return the associated value. For example, if the category of the paper is J1 or C1, then the value is HIGH. Contextual IQ (Wang and Strong,1996) measures the amount of data and the completeness of the extracted data. To obtain this measure, we have defined a questionnaire composed of Yes/No questions (see Table 6). In relation to the number of Yes and Noes, we have defined the metric depicted in Eq. (1) that returns a value on a scale of [0-1]. The contextual IQ is defined based on a three-point Likert Computers in Industry 144 (2023) 103788 7 B. Ramos-Gutiérrez et al. Fig. 4. Summary of the results. Fig. 5. Classification of a study based on the type and quality of the venue. scale with HIGH, MEDIUM and LOW values, in which HIGH is the best result and LOW is the worst result. We calculate the Contextual IQ value according to the criterion defined in Table 7. As stated in Even et al. (2010), data quality assessment cannot be defined in a general way because it is context dependent and, as a consequence, must be defined by domain experts. As our context differs from Varela-Vaca and Reina Quintero (2021), we maintain the same level of granularity as them, but we use different thresholds. Note that they present a multivocal mapping study that includes information from the grey literature, which requires a higher level of extracted data to classify a study with a high level of contextual IQ. In our case, a study with fewer than four extracted data has a low Contextual IQ, a study with seven or more extracted data has a high Contextual IQ; otherwise, the study has a medium Contextual IQ. 𝑚𝐶𝑜𝑚𝑝𝑙𝑒𝑡𝑒𝑛𝑒𝑠𝑠 =|𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑎𝑛𝑠𝑤𝑒𝑟𝑒𝑑 𝑞𝑢𝑒𝑠𝑡𝑖𝑜𝑛𝑠| |𝑇 𝑜𝑡𝑎𝑙 𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑞𝑢𝑒𝑠𝑡𝑖𝑜𝑛𝑠|(1) Finally, the quality of a study is determined based on the level of quality obtained in the two dimensions, that is, intrinsic and contextual. We consider the study acceptable when the HIGH or MEDIUM quality level is obtained in both dimensions. The study is rejected when both quality dimensions have a LOW value. 4.5. Data extraction and synthesis The data extraction process has two stages: first, the data are extracted; and second, the extracted data are reviewed. All researchers are involved in the data extraction process. The researcher who selects a study for extraction is responsible for extracting the data from that study. If the researcher in charge of the extraction has doubts about the extraction of certain labels, we discussed them at a later meeting in Table 6 Yes/No questions included in the quality questionnaire. ID Yes/No question QA01 Could the type of contribution be extracted? QA02 Could the application area be extracted? QA03 Could the logistic coverage be extracted? QA04a Could the type of business process modelling language be extracted? QA04b Could the business process modelling language be extracted? QA05a Could the type of event processing language be extracted? QA05b Could the event processing language be extracted? QA06 Could the event producer be extracted? QA06 Could the process life cycle phase be extracted? QA07 Could the challenge be extracted? Table 7 Assessment of Contextual IQ. Metric Values Contextual IQ 𝑚𝐶𝑜𝑚𝑝𝑙𝑒𝑡𝑒𝑛𝑒𝑠𝑠 >0.66 HIGH 𝑚𝐶𝑜𝑚𝑝𝑙𝑒𝑡𝑒𝑛𝑒𝑠𝑠 (0.33,0.66] MEDIUM 𝑚𝐶𝑜𝑚𝑝𝑙𝑒𝑡𝑒𝑛𝑒𝑠𝑠 0.33≤LOW which all researchers were involved. Data revision is made in parallel with data synthesis, because of inconsistencies, and errors are better detected at this point. Data extraction is performed using three different tools: a Publication Form, a form that provides the Start tool with common information about publications, such as title, authors, and so on; an Extraction Form, a customised form created in the Start tool with data about the specific domain of the literature review; and an Excel Datasheet, to store information that has been obtained automatically, such as the Computers in Industry 144 (2023) 103788 8 B. Ramos-Gutiérrez et al. Table 8 Type of data extraction form and alignment of extraction fields with research questions. Tool Extracted data RQ Publication Publication title RQ1.1, RQ1.2, RQ1.3 form Publication authors RQ1.1, RQ1.2, RQ1.3 Publication venue RQ1.4 Publication year RQ1.1 Excel Citations of the publication RQ1.3 datasheet Author’s countries RQ1.2 Extraction Type of contribution RQ2.1 form Application area RQ2.2 Type of business process modelling language RQ2.3 Business process modelling language RQ2.3 Type of event processing language RQ2.4 Event processing language RQ2.4 Logistic coverage RQ2.5 Event producer RQ2.6 Process lyfecycle phase RQ3.1 Challenge RQ3.2 citations of a publication or the author’s country. Table 8 lists the extracted data, the tool in which the data is stored, and the research questions that can be answered using the extracted data. Regarding the synthesis, we employ an approach based on both qualitative and quantitative methods to analyse the data. We use a qualitative approach when we are interested in questions about ‘‘what’’ and ‘‘how’’. To complement this qualitative analysis, we used descriptive statistics to discuss frequency and distribution. 4.6. Replication package To strengthen the replicability of our review, we have published a bundle with all the artefacts and the final results of our study in url.10 This bundle includes a Jupyter notebook with all generated graphics and some additional information, such as the raw numbers behind the graphics or the concrete studies under the different classifications made in the paper; three GitHub repositories, one with the Python code behind the graphics generation, and two with the Python-based scripts we have used to query Crossref11 and Opencitations12; and the datasets used in the study (Ramos-Gutiérrez et al.,2021). Furthermore, the selected studies can be accessed online by following the information reported in a separate Selected Studies section in the Appendix included as supplementary material. 5. Data extraction results The review process has been conducted from December 2020 to March 2022. During this process, we developed the protocol, identified and selected primary studies, performed data extraction and synthesis, and reported the results. All researchers participated in the entire process, as explained in Section 4. 5.1. Primary studies Our set of primary studies is composed of 169 studies. For space reasons, the selected studies are listed in the Appendix as a table which is ordered by year and type of venue. The table shows the reference, the title, and the venue. Fig. 6 shows the percentage of studies by type of publication. As can be observed, almost half of the studies (46.2%) have been published in conferences; more than a quarter of the studies (26%) have been published in journals, and only 18.3% of the studies have been published in workshops. This demonstrates that the field is quite mature. 10 http://www.idea.us.es/when-bp-meet-ce-in-logistics-a-sms/. 11 https://www.crossref.org/. 12 https://opencitations.net/. Fig. 6. Percentage of publications per type. Fig. 7. Studies quality assessment. 5.2. Study quality assessment The quality of each study has been determined by calculating the Intrinsic and Contextual IQ measurements (see Section 4.4). Fig. 7 shows a summary of the quality measurements. The figure depicts two dimensions, Contextual IQ and Intrinsic IQ, in a bubble chart. Each dimension can have the values LOW, MEDIUM, and HIGH. A bubble represents the number of studies that have the values depicted on the X and Y axes, respectively. Therefore, there are 41 studies that have HIGH Intrinsic and Contextual IQ; 43 studies that have a Medium Intrinsic IQ and a High Contextual IQ; and 67 studies that have a Low Intrinsic IQ and a High Contextual IQ. In summary, 151 out of 169 studies have a HIGH Contextual IQ, which means that in more than 89% of the studies the data can be extracted. Finally, it also deserves to be noticed that there is no study with Low Intrinsic and Contextual IQ, so no study has been rejected for quality reasons. 6. Discussion of research questions In this section, we discuss the research questions by synthesising the results obtained from the collected studies. RQ1. What are the demographics of the published studies? This research question helps identify the number and frequency of publications, the top researchers in the area, the top countries, the Computers in Industry 144 (2023) 103788 9 B. Ramos-Gutiérrez et al. Fig. 8. Number of publications per year. Fig. 9. (a) Proportion of type of contribution, and (b) Number of studies per contribution type and year. most influential studies, and the top venues. For space reasons, this section is mainly focused on the discussion of RQ1.1. However, details on questions RQ1.2-RQ1.4 (authors, leading countries, most influential studies, and leading venues) are provided in the Appendix. The goal of RQ1.1 is to provide information on the number and frequency of publications over time. The rationale behind this question is to analyse trends, such as the maturity of the field. Fig. 8 shows the number of relevant studies extracted in our review per year. As can be observed, the year with the highest number of publications is 2014, with a total of 23. Since 2014, the number of publications has been decreasing until 2018. This implies that event-driven business processes in logistics were an emerging research topic in 2007–2014. Afterwards, the number of publications has decreased until 2018, when it began to increase. Since then, it has started to decrease again in the years of pandemics (2020 and 2021). RQ2. What are the current trends in the area? This research question helps identify the nature of existing work in the area. The following subsections discuss the questions RQ2.1-RQ2.4, because they are the ones that allow us to obtain more interesting conclusions. The reader interested in RQ2.5 and RQ2.6 can consult the Appendix. RQ2.1. What is the type of contribution made by the study? This question aims to identify the nature of the work in the area. As in Petersen et al. (2008), the categories were determined by carefully reading the studies. It should be noted that there are studies that produce more than one type of contribution. For example, the study (Linden et al.,2013) proposes an architecture and a framework. Fig. 9(a) shows the proportion of types of contributions, and Fig. 9(b) shows the distribution of contributions per type and year. Architecture is the most frequent type of contribution (59 out of 169 studies contribute with an architecture), although the number of studies that contribute with a framework (58 out of 169) is also important. Unfortunately, most of the proposals that contribute with frameworks, tools and architectures do not offer a definitive version of their solution. In fact, many of them propose solutions that are still under development or are purely conceptual (24.5%), proofs of concept (10.7%) and prototypes (19.8%) (see the Appendix). Furthermore, most of the approaches that provide more details on the implementation of their solutions do not make their software publicly available (19.1%) and only 9.9% of the studies propose the integration of well-known commercial tools (see Fig. 10). Finally, all these proposals have one thing in common: they focus on the design or development of prototypes and software tools for academia; none of them constitutes a solution close to being a commercial solution applicable in industry. RQ2.2. Which are the application areas of the approaches? The aim of this research question is to identify the areas within the logistics sector in which the approaches are applied. The application areas have been obtained from Alias et al. (2016). They include maritime transport, air transport, multimodal transport, general transport, manufacturing, foodstuffs, transshipment, chemical and pharmaceuticals, retail, and others. After carefully reading the studies, we have classified the use cases, running examples, or descriptions they present according to the areas of application previously mentioned. Note that a study can be associated with more than one area of application. As Computers in Industry 144 (2023) 103788 16 B. Ramos-Gutiérrez et al. Hull, R., Damaggio, E., Fournier, F., Gupta, M., Heath, F.T., Hobson, S., Linehan, M., Maradugu, S., Nigam, A., Sukaviriya, P., et al., 2010. Introducing the guardstage-milestone approach for specifying business entity lifecycles. In: International Workshop on Web Services and Formal Methods. Springer, pp. 1–24. Jaekel, F., 2019. Systematic review of cloud logistics knowledge. In: Cloud Logistics: Reference Architecture Design. Springer Fachmedien Wiesbaden, Wiesbaden, pp. 191–256. http://dx.doi.org/10.1007/978-3-658-22837-8_5. Johansson, H.J., McHugh, P., Pendlebury, A.J., Wheeler, W.A., 1993. Business Process Reengineering: Breakpoint Strategies for Market Dominance. John Wiley & Sons. Khan, M.U., Sherin, S., Iqbal, M.Z., Zahid, R., 2019. Landscaping systematic mapping studies in software engineering: A tertiary study. J. Syst. Softw. 149, 396–436. http://dx.doi.org/10.1016/j.jss.2018.12.018. Kitchenham, B.A., Budgen, D., Brereton, O.P., 2010. The value of mapping studies - A participant-observer case study. In: EASE. In: Workshops in Computing, BCS. Kitchenham, B.A., Budgen, D., Brereton, P., 2015. Evidence-Based Software Engineering and Systematic Reviews, first ed. CRC Press, An optional note. Kitchenham, B., Charters, S., 2007. Guidelines for performing Systematic Literature Reviews in Software Engineering. Tech. Rep. EBSE-2007-01, School of Computer Science and Mathematics, Keele University. Klappich, D., Muynck, B.D., Aimi, G., Titze, C., Stevens, A., 2020. Cpredicts 2021: Supply chain technology. Krumeich, J., Weis, B., Werth, D., Loos, P., 2014. Event-driven business process management: where are we now?: A comprehensive synthesis and analysis of literature. Bus. Process. Manage. J. 20 (4), 615–633. http://dx.doi.org/10.1108/ BPMJ-07-2013-0092. Kuhrmann, M., Fernández, D.M., Daneva, M., 2017. On the pragmatic design of literature studies in software engineering: an experience-based guideline. Empir. Softw. Eng. 22 (6), 2852–2891. http://dx.doi.org/10.1007/s10664-016-9492-y. Leitner, M., Rinderle-Ma, S., 2014. A systematic review on security in Process-Aware Information Systems - Constitution, challenges, and future directions. Inf. Softw. Technol. 56 (3), 273–293. http://dx.doi.org/10.1016/j.infsof.2013.12.004. Li, C.-S., 2005. Real-time event driven architecture for activity monitoring and early warning. In: Conference, Emerging Information Technology 2005. p. 4 pp.. http: //dx.doi.org/10.1109/EITC.2005.1544382. Li, T., Zeng, C., Jiang, Y., Zhou, W., Tang, L., Liu, Z., Huang, Y., 2017. Datadriven techniques in computing system management. ACM Comput. Surv. 50 (3), 45:1–45:43. http://dx.doi.org/10.1145/3092697. Linden, I., Derbali, M., Schwanen, G., Jacquet, J., Ramdoyal, R., Ponsard, C., 2013. Supporting business process exception management by dynamically building processes using the BEM framework. In: EWG-DSS. In: Lecture Notes in Business Information Processing, vol. 184, Springer, pp. 67–78. http://dx.doi.org/10.1007/978-3-319- 11364-7_7. Luckham, D.C., 2005. The Power of Events - An Introduction to Complex Event Processing in Distributed Enterprise Systems. ACM. Luckham, D., 2008. The power of events: An introduction to complex event processing in distributed enterprise systems. In: RuleML. In: Lecture Notes in Computer Science, vol. 5321, Springer, p. 3. http://dx.doi.org/10.1007/978-3-540-88808- 6_2. Luckham, D., 2012. Event Processing for Business. In: Organizing the Real-Time Enterprise, Wiley. Mousheimish, R., Taher, Y., Zeitouni, K., 2016. The butterfly: An intelligent framework for violation prediction within business processes. In: IDEAS. ACM, pp. 302–307. http://dx.doi.org/10.1145/2938503.2938541. OASIS, 2007. Web Services Business Process Execution Language Version 2.0. Tech. Rep., OASIS. OMG, 2013. Business Process Model and Notation (BPMN), Version 2.0.2. Tech. Rep., Object Management Group. Pérez-Álvarez, J.M., Maté, A., Gómez-López, M.T., Trujillo, J., 2018. Tactical businessprocess-decision support based on KPIs monitoring and validation. Comput. Ind. 102, 23–39. http://dx.doi.org/10.1016/j.compind.2018.08.001. Pesic, M., Van der Aalst, W.M., 2006. A declarative approach for flexible business processes management. In: International Conference on Business Process Management. Springer, pp. 169–180. Petersen, K., Feldt, R., Mujtaba, S., Mattsson, M., 2008. Systematic mapping studies in software engineering. In: EASE. In: Workshops in Computing, BCS. Petersen, K., Vakkalanka, S., Kuzniarz, L., 2015. Guidelines for conducting systematic mapping studies in software engineering: An update. Inf. Softw. Technol. 64, 1–18. http://dx.doi.org/10.1016/j.infsof.2015.03.007. Ramos-Gutiérrez, B., Reina Quintero, A., Parody Núñez, M., Gómez López, M., 2021. When business processes meet complex events in logistics: A systematic mapping study’’, mendeley data. http://dx.doi.org/10.17632/xwy9djwnh4.1. Schiefer, J., McGregor, C., 2004. Correlating events for monitoring business processes. In: ICEIS (1). pp. 320–327. Schiefer, J., Roth, H., Suntinger, M., Schatten, A., 2007. Simulating business process scenarios for event-based systems. In: ECIS. University of St. Gallen, pp. 1729–1740. Shanahan, M., 1999. The event calculus explained. In: Artificial Intelligence Today. Springer, pp. 409–430. Soffer, P., Hinze, A., Koschmider, A., Ziekow, H., Ciccio, C.D., Koldehofe, B., Kopp, O., Jacobsen, H., Sürmeli, J., Song, W., 2019. From event streams to process models and back: Challenges and opportunities. Inf. Syst. 81, 181–200. http://dx.doi.org/ 10.1016/j.is.2017.11.002. Ter Hofstede, A.H., Van der Aalst, W.M., Adams, M., Russell, N., 2009. Modern Business Process Automation: YAWL and its Support Environment. Springer Science & Business Media. Varela-Vaca, Á.J., Reina Quintero, A.M., 2021. Smart contract languages: A multivocal mapping study. ACM Comput. Surv. 54 (1), 3:1–3:38. http://dx.doi.org/10.1145/ 3423166. W3C, 2013. SPARQL 1.1 Query Language. Tech. Rep., W3C. Wang, R.Y., Strong, D.M., 1996. Beyond accuracy: What data quality means to data consumers. J. Manage. Inf. Syst. 12 (4), 5–33. http://dx.doi.org/10.1080/ 07421222.1996.11518099. Weske, M., 2012. Business Process Management - Concepts, Languages, Architectures, second ed. Springer. Wohlin, C., 2014. Guidelines for snowballing in systematic literature studies and a replication in software engineering. In: EASE. ACM, pp. 38:1–38:10. http://dx.doi. org/10.1145/2601248.2601268. Wohlin, C., Runeson, P., Höst, M., Ohlsson, M.C., Regnell, B., 2012. Experimentation in Software Engineering. Springer, http://dx.doi.org/10.1007/978-3-642-29044-2. Belén Ramos Gutiérrez, (Phd. Student) Universidad de Sevilla, Dpto. Lenguajes y sistemas informáticos – Spain. Belén Ramos-Gutiérrez is currently a Software Engineering and Technology Ph.D. Student at the University of Seville. Her research interests include process mining, data extraction, and optimisation for process mining techniques; and troubleshooting and decision support systems in industrial environments. She is also working as a predoctoral researcher with the University of Seville and collaborates on projects involving industrial aspects related to logisticsport and aeronautical environments. Her research goal is to improve and automate the extraction and processing of heterogeneous data from industrial environments for exploitation with process mining techniques. Antonia M. Reina Quintero, (Assistant Professor) Universidad de Sevilla, Dpto. Lenguajes y sistemas informáticos – Spain. Antonia M. Reina Quintero obtained her Ph.D. with honours in Computer Engineering from the University of Seville (2012). She has worked as a full-time lecturer at the Computer Languages and Systems Department from the University of Seville since 2000, although she also has worked as a computer engineer for a leading company in traffic control systems. Her current research is focused on advanced separation of concerns, Model-Driven Engineering applied to business processes, and systematic literature reviews in software engineering. She has participated in public research projects and prestigious conferences. She has published several high-impact papers. Luisa Parody, (Associate Professor), Universidad Loyola Andalucía, Sevilla, Spain. Luisa Parody studied Computer Engineering (including a minor in Systems Engineering) at the Universidad de Sevilla (Spain) and graduated with honours in July 2009. She then earned an M.Sc. Degree in Software Engineering and Technology(2010) and obtained her international Ph.D. with honours at the Universidad Sevilla (2014). Since 2018, she has been working as an associate professor at Dpto. Métodos Cuantitativos at the Universidad Loyola Andalucía. She belongs to the IDEA Research Group and has participated in several private and public research projects. She was nominated as a member of Program Committees, such as ISD and SIMPDA. She has participated in prestigious conferences and has published several high-impact papers. María Teresa Gómez-López is a Full Professor at the University of Seville and the head of the IDEA Research Group (www.idea.us.es). Her research areas include Business Processes and Data management, and how to improve the business process models including better decisions and enriching the model with Data Perspectives. She has led several private and public competitive research projects and has published more than 30 articles in high-quality conferences and journals. She was nominated as a member of Program Committees, such as ER, BPM, EDOC, ISD or CAISE Doctoral Consortium, and she has given various keynotes or was invited speaker in international forums.