1 Introduction

In the last decade, the possibility to manufacture micro-controllers and CPUs of small physical dimensions at reduced cost has made it feasible to embed these devices alongside various everyday physical objects at a massive scale – e.g. home appliances and industrial equipment, just (to name a few). Combining this increased computing capacity with sensors, actuators, and communication modules allows such objects to interface with the physical world and operate autonomously, thus becoming smart objects. In this context, Internet-of-Things (IoT) broadly refers to the network of connected smart objects able to collect and exchange data over the Internet, thus providing advanced functionalities such as real-time monitoring, data sharing, and remote control (Kopetz and Steiner 2022).

Typical early adopters of smart objects and IoT can be found in industrial manufacturing, logistics and healthcare domains. For these areas, the tools and practices of Business Process Management (BPM) play a relevant role (Leotta et al. 2019). BPM is a research area concerned with the identification, modeling, analysis, redesign, implementation, monitoring and mining of business processes (BPs) (Dumas et al. 2018). A BP is a collection of related, structured activities, events and decisions that accomplish a specific organizational goal. According to the BPM approach, BPs are formalized through explicit BP models expressed with a suitable graphical notation, such as the standard BPMN.Footnote 1 A specific type of information system, called Process Management System (PMS), is employed to support the execution and monitoring of BPs and analyze historical traces of BP execution (Weske 2019).

IoT’s presence in BPM activities continuously increasing since IoT data allows for finer-grained automated monitoring, analysis, and control during BP execution. Consequently, organizations’ BPs must be aligned to take advantage of IoT capabilities. This has resulted in the rise of relevant studies focused on IoT-aware BPM (Cheng et al. 2018, 2019; Grefen et al. 2019). The Industry 4.0 revolution was one of the significant accelerators for integrating IoT and BPM. In addition to the increased usage of smart objects, Industry 4.0 prescribes a factor of high degree of individual customizability for manufactured products, an aspect that IoT-aware BPM can effectively capture (Grefen et al. 2018). In this direction, Janiesch et al. in their manifesto about “The Internet of Things Meets Business Process Management” (IoT-Meets-BPM Manifesto for short in the following) identified many open challenges to address in order to achieve a more comprehensive and targeted integration of the main concepts of the two areas (Janiesch et al. 2020).

This study aims to gain profound insights into the nature and maturity of existing approaches to IoT-aware BPM. In particular, we present results from a Systematic Literature Review (SLR) conducted to systematically evaluate the literature on IoT-aware BPM concerning the phases of the BP life cycle covered by the different approaches, the specific topics addressed, the application domains involved, and the possibility to tackle the research challenges identified by Janiesch et al. (2020).

Our SLR, which follows the scientific and reproducible approach proposed by Keele (2007), consists of 3 phases: (1) the planning phase identifies the needs of the SLR, defines the research questions and develops a protocol for the review; (2) the conducting phase selects the studies, extracts and synthesizes relevant data; (3) the reporting phase analyzes the collected data and presents the results based on the prior research analysis.

With the support of the SLR, we identify 93 primary studies that are thoroughly analyzed. Based on this evaluation, we examine (i) which phases of the BPM life cycle were investigated by each study, (ii) which specific topics relating to IoT integration in BPM were primarily discussed by each study, and (iii) which are the main application domains addressed so far. Based on this analysis, we compare the studies according to which of the research challenges identified in the IoT-Meets-BPM Manifesto (Janiesch et al. 2020) are addressed. It emerges that the field of IoT-aware BPM is still in the development stage. Therefore, we consider our analysis beneficial for researchers and practitioners addressing such a novel and interesting field.

The paper is structured as follows. Section 2 shows the methodology we applied, including the research questions and the search protocol we adopted. In contrast, Sect. 3 describes the application of the protocol, and Sect. 4 gives an overview of the results achieved. Finally, to conclude our paper, Sect. 5 comprises a discussion of our results, including a presentation of the validity of the SLR, and Sect. 6 provides some concluding remarks.

2 Planning the Systematic Literature Review

In this section, we describe the planning phase of the SLR. Planning includes the following activities:

  1. 1.

    Related work analysis to motivate the need of the SLR;

  2. 2.

    Formulation of research questions to narrow the search;

  3. 3.

    Definition of the search strings;

  4. 4.

    Selection of data sources;

  5. 5.

    Definition of inclusion and exclusion criteria.

2.1 Related Work

The increased research community’s interest in the interplay between IoT and BPM has led to the publication of relevant literature reviews and surveys on the topic. In Del Giudice (2016), a literature review is conducted to analyze the impact and the role of IoT in fostering innovation within organizations, and which implications this phenomenon may have on BPM and the competitiveness of firms.

In Chang et al. (2016), the authors analyze existing PMSs for IoT and identify the limitations and their drawbacks based on a mobile cloud computing perspective.

In Stoiber and Schönig (2021), a survey on event-driven BPM is conducted focusing on its capabilities to enable the scale-up of IoT applications. Research on incorporating IoT into BPM has shown that the majority of SLRs focus on the IoT-aware BP modeling phase. In Torres et al. (2020), a SLR is performed to collect and describe ways to model IoT-aware BPs. Compagnucci et al. (2023, 2020) for example, investigated modeling views related to different types of IoT-aware BPs and the IoT requirements supported by the various modeling notations. Abouzid et al. (2022) conducted a SLR on BPMN extensions for IoT domains and their integration in supply chain BPs. Compared to the other papers, this SLR not only addressed the modeling of IoT-aware processes, but also the other phases of BPM lifecycle. Finally, the work Fattouch et al. (2020) presents a survey of different approaches that integrate IoT technologies within BPs. The authors provide a brief overview of each approach and compare them based on specific criteria by identifying some open challenges. This survey shares our objective and an intersection of 8 out of 93 primary studies, but it was conducted in a non-systematic way, thus not permitting the results’ replicability. Therefore another contribution of our approach is that it includes this replicability.

Whereas the above studies focus on analyzing specific aspects of the integration of IoT and BPM, in this SLR we provide a holistic view of such an integration by exploring its impact on the different BP life cycle phases and BPM-related application domains. As opposed to our approach, none of the works employ a generative statistical model to discover the most discussed topics in the collection of studies and automatically classify individual studies based on their relevance to these topics essential for achieving the integration of IoT and BPM.

Moreover, we discuss which challenges of the ones identified by the IoT-Meets-BPM Manifesto (Janiesch et al. 2020) have been (to date) adequately addressed or are still far from being tackled.

2.2 The Research Questions

Our main goal is to assess the current state of research on the IoT integration in the BPM ecosystem, consisting of methods, approaches and tools to manage the life cycle of a BP. To achieve this, we formulated the following research questions from collaborative brainstorming sessions among the authors:

  1. RQ1:

    Which phases of the BP life cycle was mainly impacted by the integration of IoT?

  2. RQ2:

    Which topics relating to IoT integration in BPM are discussed in the literature?

  3. RQ3:

    Which application domains are used to study IoT and BPM integration?

  4. RQ4:

    What are the research challenges addressed so far to support the management of IoT-aware BP?

2.3 Data Sources

We applied the search string on three search engines: Google Scholar (GS) and two more specialized in information systems and computer science, namely the IEEE Xplore Digital Library and ACM Digital Library. This approach has enabled us to overcome limitations associated with relying solely on a single search engine, specifically: (i) accessing a broader spectrum of pertinent sources beyond GS, (ii) mitigating the risk of missing relevant studies, (iii) reducing the inclusion of low-quality sources, and (iv) overcoming potential biases towards popular and recent publications.

As a further data source, we also reviewed the literature cited by the studies, performing a backward reference search. Additionally, we carried out a conference and journal search to retrieve other studies without any time constraints. The relevant studies for this SLR include publications from the most relevant conferences, seminar proceedings, and journals within the BPM and IoT fields and those focusing on practical solutions and applications. These sources have been selected through a well-defined and consistent process and align with the review’s topic and objective. The journals and conferences were initially selected based on the authors’ research experience. Subsequently, we expanded the search and defined criteria for adding new sources. Specifically, these were chosen because (i) they are subject to a peer-review process and (ii) they are registered in the libraries used to conduct this SLR. As a result, we included nine conferences, a dedicated workshop and eight journals, namely:

  • Conference on: Business Information Systems (BIS), Business Process Management (BPM), Advanced Information Systems Engineering (CAiSE), Enterprise Distributed Object Computing (EDOC), Conceptual Modeling (ER), Process Mining (ICPM), Computational Intelligence and Computing Research (ICCIC), Service Science (ICSS), Workshop on Business Processes Meet Internet-of-Things (BP-Meet-IoT), Business Process Modeling, Development and Support (BPMDS).

  • ACM Transactions on Intelligent Systems and Technology (ACM TIST), Business and Information Systems Engineering (BISE), Business Process Management Journal (BPMJ), Computers in Industry, Information Systems, IEEE Transactions on Knowledge and Data Engineering (TKDE), Software and Systems Modeling (SoSyM), Journal on Data Semantics (JoDS).

2.4 The Search String

The search terms for our SLR were constructed using the authors’ knowledge of the subject matter. We started with the two most relevant acronyms, BPM and IoT. Starting from these, a first survey of the databases was carried out to capture more information on the topic, thus identifying the advantages and disadvantages of the possible choices. However, the acronym BPM produced many studies that did not relate to BPM but referred to the concept of beats per minute. For this reason, we expanded it to include the terms process and workflow. On the other hand, we included the term Internet of Things in its longer version to avoid the appearance of improper studies. From this initial set, we extended the search by adding synonyms to increase the possible configurations and create a final string that avoided excluding relevant studies. For example, we included cyber physical, industry 4.0 and smart because IoT technologies were primarly observed in specific domains (Leotta et al. 2019). The search terms were linked using boolean AND, while alternative spellings and synonyms were combined using boolean OR.Footnote 2 The resulting search string is given below:

(process OR workflow OR processes OR workflows OR BPM) AND (IoT OR internet of things OR cyber physical OR industry 4.0 OR smart)

2.5 Inclusion and Exclusion Criteria

The SLR also includes identifying specific criteria that allow us to guarantee that only relevant studies are selected. Consequently, we defined the following inclusion and exclusion criteria. We did not use all filters simultaneously, but each criterion was applied within a specific process step.

2.5.1 Inclusion Criteria

We define one criterion for study inclusion. Criterion I.1 includes the studies proposing a case study, an approach, a technique or a conceptual framework that deals with the concrete use of IoT to tackle issues from the BPM community. Consequently, studies using BPM techniques for IoT community problems or merely mentioning IoT and BPM without being the primary focus of the research do not meet the inclusion criteria.

2.5.2 Exclusion Criteria

Five criteria for study exclusion were also been defined.

Criterion E.1 excludes the studies not written in English.

Criterion E.2 excludes the studies that are not available. This mainly concerns studies that are not electronically accessible, require payment for access, or are only partially available (e.g., abstracts only).

Criterion E.3 excludes the studies that are not peer-reviewed.

Criterion E.4 excludes the studies that do not satisfy any inclusion criteria.

Criterion E.5 excludes studies with fragmented results, where findings are reported across multiple publications. In such cases, only the most comprehensive publication is considered for inclusion.

A study was considered for inclusion in the SLR if it met at least one inclusion criterion and did not meet any exclusion criteria.

3 Conducting the Systematic Literature Review

This section describes the search protocol developed to identify and select studies.

3.1 Study Selection

To carry out the selection, we follow a three-phase search process as suggested by Keele (2007):

  1. 1.

    Phase 1 – Digital library search: During this phase, 5,315 studies were identified after entering the search string on GS,Footnote 3 IEEE and ACM. Then duplicate studies were removed, obtaining as a result a total of 4,869 potential studies. We then applied the inclusion/exclusion criteria, resulting in 309 potentially relevant studies. These studies were assessed based on their titles, abstracts, conclusions, and keywords to determine their relevance to the SLR.

  2. 2.

    Phase 2 – Conference and journal search: In the second phase, we selected the leading conferences and journals of the domain of interest and carried out a new search to add 233 studies. Then, duplicate studies were removed, obtaining, as a result, a total of 195 potential studies. We then applied the inclusion/exclusion criteria that allowed us to add 38 potential relevant studies to the studies included in the previous phase.

  3. 3.

    Phase 3 – Backward snowballing search: In the third phase, we performed a backward snowballing search to analyze references and citations of the studies from the previous two phases and from both searches to identify other relevant studies. Thus, we obtained further 8 studies.

In summary, at the end of the three search phases, 355 studies were obtained, which were further processed into 93 final studies after applying the inclusion/exclusion criteria. The search was formally completed in June 2023, and we decided to start from 1999 when the term IoT was coined. Data extraction involved collecting such data in an Excel spreadsheet. The Excel file can be found at http://tinyurl.com/m7j4x9mm. Figure 1 shows all the steps of the selection process by the guideline for reporting systematic reviews (PRISMA) proposed by Page et al. (2021). Table 1 shows all the primary studies.

Fig. 1
figure 1

Flow diagram summarizing the selection of the studies along the three search phases

Table 1 List of selected studies

3.2 Data Extraction

The data extraction and synthesis phase aims to design an appropriate form to record and collect the relevant information from reading the selected studies. The data collection form includes a common set of general fields such as title, author, year of publication and type of publication. We have also included multiple choice fields for each research question RQ1, RQ2, RQ3 and RQ4 (see Table 2). A section summarizing the research carried out in the study also includes its strengths and weaknesses in the form. The forms of the selected studies are available at this link: http://tinyurl.com/2udm342b.

Table 2 Attributes for data extraction

4 Results of the Systematic Literature Review

In this section, we discuss the SLR results based on the research questions listed in Sect. 2. We cover: (i) the BPM life cycle phases affected by the IoT, (ii) the main topic of the study on BPM and IoT, (iii) the application domain on the usage of BPM and IoT solutions, and (iv) the addressed research challenge(s).

4.1 Which Phases of the BP Life Cycle Were Mainly Impacted by the Integration of IoT?

The research question RQ1 seeks to group the studies dealing with the BPM life cycle phases. We found that all the selected studies (93) refer to, or in some cases explicitly mention, one or more phases of the BPM’s life cycle to contextualize their work (cf. Table 3). In detail, phases are considered by the identified studies with regard to the following aspects (cf. Figure 2): Identification (3.8%), Modeling (35.4%), Analysis (7%), Redesign (5.7%), Implementation (31.6%) and Monitoring and Mining (16.5%).

Fig. 2
figure 2

Distribution of selected studies according to the aspect of integration of IoT into the BP life cycle phases

  • Identification Hu et al. (2014) propose a dynamic integration mechanism of BPs to adapt to the dynamic characteristics of IoT, redirect and restructure the process logic using rules that can be edited at runtime and design a dynamic integration algorithm to implement this mechanism. Schief et al. (2011) propose a centralized framework that extends the process design and execution phases of BPM considering events generated by smart objects.

  • Modeling Based on the analysis, we have categorized the studies according to several key areas: Integration of IoT concepts in BPMN 2.0Gao et al. (2011) propose a BPMN 2.0 extension with sensor and smart device business functions to enhance the integration between the physical world and BPs. Meyer et al. (2011) extend the BPMN 2.0 to explicitly consider IoT basic concepts, enabling the modeling and development of IoT Services in terms of BPs. Meyer et al. (2013) integrate IoT devices as BP resources in the form of dedicated lanes and extend the BPMN 2.0 specification to include such IoT devices. Meyer et al. (2015) explore the representation of the IoT domain component “Thing” in a process model, introducing potential BPMN extension for IoT concepts. Verification and validation of IoT-BPs Graja et al. (2019) extend BPMN to sustain the various cyber-physical system concepts and a verification approach at design time to detect the mistakes in specifying the CPS process. Senderovich et al. (2016) derive events from sensor data, aligning them with process activities. Hornsteiner and Schönig (2023) extend BPMN to map security aspects and thus obtain security- and IIoT-aware BPs. Context management and IoT adaptation Schönig et al. (2018) propose a set of concepts for an IoT-enhanced process model re-engineering, addressing model changes and adaptation tasks. Suri et al. (2018) propose configuration concepts for handling IoT resource variability in the Configurable Process Model. IoT-aware BP decentralization and distribution Domingos et al. (2015) present an automatic approach to decentralize IoT-aware BPs defined using the BPMN 2.0. Jain and Tata (2017) have designed a distributed IoT application with annotated components and location information. IoT event integration and management Chadli et al. (2022) define the role of the IoT concept of smart business processes modeling SBPM, using an ad hoc approach with active help to detect data flow anomalies. Li et al. (2021) propose a business user-oriented BP modeling method supporting IoT event stream integration, with a framework that separates CEP and BP execution. Wang et al. (2022) monitor and model IoT sensing data in real-time, transforming it into IoT services for BP systems. Kirikkayis et al. (2023b) model IoT-driven events using BPMN 2.0 extension and DMN concepts. Kirikkayis et al. (2022b) introduce a BPMN 2.0 extension with IoT artifacts, enabling data acquisition and actuator control. Other aspects related to IoT and BPMZanfack et al. (2015) present a GSMFootnote 4-IoT framework which aims to represent the physical workflows in logistics, extending them by adding physical workflow concepts and considering particularities of IoT smart entities and their mutual interactions. Mottola et al. (2019) propose a BPMN extension for wireless sensor network context modeling. Bocciarelli et al. (2017) model and manage resource management during runtime. Hasić et al. (2020) compare standard BPMN and BPMN + DMN for IoT process modeling. Grefen et al. (2019) introduce a time and space specification for IoT-aware collaborative BPs. Hou et al. (2016) suggest fragmenting IoT-aware BP for performance improvement. Grambow et al. (2021) propose context-awareness integration in IIoT BPs, visual AR support, and enhanced BPMN modeling.

  • Analysis Antonius and Dachyar (2020) model and simulate the cardiac remote patient monitoring process, identifying weaknesses needing improvement. Vitali and Pernici (2016) present an approach based on the temporal events analysis for building a graph of dependencies between the whole set of events, unveiling interconnections between processes of cooperating organizations. Seiger et al. (2023) develop an interactive method for analyzing low-level IoT data to detect higher-level process activity executions based on sensor-actuator-activity patterns. Diamantini et al. (2023) discuss the role of semantic models when integrating IIoT data streams into an Industry 4.0 context. They propose a process-aware knowledge graph that enriches sensor data using ontological descriptions of IIoT sensors, processes, and KPIs. Gómez-Valiente et al. (2023) present the DOMINIoT architecture to address scheduling, resource allocation, and state management challenges.

  • Redesign The redesign phase is often applied across other BPM life cycle stages. In Ruiz-Fernández et al. (2017), a redesign of a given clinical process is presented, while in Ismaili-Alaoui et al. (2018), interdisciplinary healthcare processes are redesigned to ensure direct and real-time interaction between the patient and the medical staff. Song et al. (2022) mainly address this phase by proposing an IoT-enabled Context-aware BPM (IoT-CaBPM) framework, which facilitates the evolution of BPs in combination with IoT advances.

  • Implementation Chen et al. (2012) propose to extend WS-BPELFootnote 5 with IoT characteristics for a new process definition language. Friedow et al. (2018) implement and write adapters using the Bosch IoT Things service to connect IoT devices and BPs via event influence. Kikuchi et al. (2018) develop a cloud-IoT orchestration framework with IoT device. Kunz et al. (2011) provide novel components and architectural concepts for improved IoT communication. These include a scalable discovery service and a decentralized control structure, cloud EPCIS,Footnote 6 which facilitates access to RFIDFootnote 7 data, and middleware that enables flexible adaption of object-specific event processing rules. Marrella and Mecella (2017) automate process adaptation in cyber-physical domains during the occurence of exceptions and exogenous events. Mass et al. (2016) propose a system architecture enabling continuous, delay-tolerant BP execution on mobile nodes. Muhsin et al. (2016) present a mobile workflow management system for seamless and accurate data exchange between mobile devices and remote hubs. Seiger et al. (2018) propose an IoT workflow management system with dynamic service selection, complex event processing, human interactions and self-adaptation. Varga et al. (2018) propose a supporting system for service chaining and workflow execution using a Petri Net-based method, similar to a recent article that provide further insights into this area (Kozma et al. 2019). Wieland et al. (2008) define and implement two smart workflows and provide an architecture for transforming sensor data into business-level information. Zhu et al. (2014) design and implement a dynamic adaptation framework using an adaptive algorithm, allowing the system to infer the process when the environment changes. Di Martino et al. (2022) provide a methodology for the smart management of irrigation systems using ontologies, BPMN semantic annotation, and logical inference. Park et al. (2018) propose modeling and implementation methods for IoT execution design. Pastor et al. (2022) integrate IoT sensors for improved user comfort in shared spaces, emphasizing semantic representation and process mining (PM) techniques. Valderas et al. (2023) apply the SoC principle for the interdisciplinary development of IoT-enhanced BPs using BPMN, ontology-based technologies, and a microservices architecture.

  • Monitoring and Mining Seiger et al. (2020) propose a framework correlating IoT sensor streams with process events and activities using Complex Event Processing. Shamsuzzoha et al. (2014) discuss collaborative BP monitoring across organizations using IoT and cloud-based data repositories. Wombacher (2011) monitors workflow-state correlation with sensor data for online conformance checking. de Leoni and Pellattiero (2021) propose a technique to discover readable human-habit models for optimizing human experience with IoT systems. Janssen et al. (2020) present a technique that combines clustering and PM to discover activities and process models from motion sensors. Koschmider et al. (2020) present a framework to discover activities and process models from event location sensor data. van Eck et al. (2016) map sensor measurements to human activities, group them into process instances and convert them into an event log as input for any PM technique. Sora et al. (2017) transform raw movement measurements into actions and adapt sensor logs to apply BPM techniques. Bertrand et al. (2021) propose a model bridging the gap between IoT and PM using IoT ontologies and BP context models. Elali et al. (2022) enhance process models derived from PM with contextual information from sensor data. Elkodssi et al. (2022) present a PM approach for discovering and analyzing IoT-aware BPs and transforming sensor logs into structured event logs enriched with IoT concepts using a BPMN Ontology.

  • More phases A BPMN 2.0 IoT-enabled extension for modeling, executing and monitoring IoT-aware BPs is proposed by Gallik et al. (2022), while Kirikkayis et al. (2023a) also present a framework for IoT-driven business rules. Kirikkayis et al. (2023c) present a BPMN 2.0 extension for IoT, enabling real-time interaction with devices which aids decisions through collected data but lacks high-level aggregation consideration. Kirikkayis et al. (2022a) present a web-based framework integrating IoT data into BPs, enhancing real-time decision-making and monitoring.

Table 3 Studies distribution by BPM life cycle (RQ1)

4.2 Which Topics Relating to IoT Integration in BPM are Discussed in the Literature?

To answer the question RQ2, we analyzed the identified studies based on the Latent Dirichlet allocation (LDA) technique (Blei et al. 2003), which is one of the most effective ones for latent topics distribution within a corpus (Chauhan and Shah 2021). LDA is a generative probabilistic model of a corpus where the documents are represented as a random mixture of latent topics each characterized by a mix of words. The application of the LDA to identify the topics in the 93 studies is detailed in a specific appendix (available online via http://link.springer.com). The resulting four topics are listed in Table 4, where the bolded words represent the most relevant terms in terms of estimated frequency within the selected topic. Once we trained the model and derived the four topics, we observed the keywords associated with each topic using a web-based interactive visualization (see the appendix) to understand what they refer to from the semantic point of view. Then, we tried to contextualize them within a semantic context. Table 5 shows the 93 studies and their topics. Note that the topics associated with each study are not exclusive, as a study can be related to one or more topics. We then discuss the four topics concerning specific studies.

Table 4 The four topics mostly discussed in the 93 selected studies and terms related tp them
Table 5 Associations between papers and their topics (the study topic number is in bold)
  • Topic 1 – Event mining from sensors Knowledge extraction from sensor data has gained significant attention in the BPM community. Process mining techniques can greatly enhance BPM management when applied to analyze and transform raw data into usable information. The information provided by the sensors can be expressed through a functional business view, as demonstrated in Gao et al. (2011), representing the classification, storage, and distribution procedure of floral products within a flower company. This model in BPMN is connected to a functional model to import an ontology of the sensors and related instance data and then integrate the information modeled using the Semantic Sensor Network ontology with BPs. Similarly, various frameworks address the correlation between workflow states, sensor data, and events generated by IoT devices that initiate instances of BPs (Seiger et al. 2020; Wombacher 2011; Ismaili-Alaoui et al. 2018; Ruiz-Fernández et al. 2017; Ruppen and Meyer 2013; Chadli et al. 2022; Diamantini et al. 2023; Pastor et al. 2022). Koschmider et al. (2020) detect high-level events from raw event data and discover BPs from derived instances. Zhu et al. (2014) present a multi-agent framework that chooses the best step to move in the BP when an event caused by agents occurs. In addition, Sora et al. (2017) present an approach to discover process models from activities in the field of smart spaces where sensors must turn the sensor log into an event log consisting of human actions. Another interesting challenge when applying PM techniques in the context of sensor data is mapping sensor measurements to human activities and grouping activities into process instances (van Eck et al. 2016; de Leoni and Pellattiero 2021). The problem of mapping sensor data to event logs based on process knowledge is solved by Senderovich et al. (2016), which maps location-based events to activities by recognizing interactions between various agents and transforming historical sensor data registers into standardized event logs. Another contribution to translating sensor data into higher-level ones comes from Janssen et al. (2020). They discretize sensor data into activities using unsupervised learning through clustering, while Seiger et al. (2023) analyze sensor data to detect higher-level process activities.

  • Topic 2 – Physical resource management This topic tackles the specification and management of the resources associated with the BPs supporting Cyber-Physical Systems (CPSs). Graja et al. (2019) propose a verification framework to support the various CPS concepts and properties, which enables the designer to handle CPS process features. To integrate and utilize smart devices as BP resources to support the modeling, a service-oriented BPM system architecture was developed by Wehlitz et al. (2017). The resource management during runtime of a system to model a CPS-aware resource associated with a BP activity is proposed by Bocciarelli et al. (2017), while the DOMINIoT architecture is designed by Gómez-Valiente et al. (2023). In scenarios where physical resources are exchanged, knowing how a resource owned by a party is managed on anothers party’s premises is impossible. Thus possible misalignments can be detected too late. In Montali and Plebani (2017), the authors investigated an approach for compliance checking that mixes commitments and smart devices. Suri et al. (2017) developed a framework that describes IoT resources (e.g. extending the BP models with energy cost parameters to enable the energy-aware management of IoT resources in BPs).

  • Topic 3 – Context-aware execution A context-aware execution environment involves processes continuously observed and adapted according to the model specified when required by context changes. Contextual information enhances process execution, incorporating situation-aware insights for improved effectiveness. Such information is often gathered from sensors within intelligent environments, enabling pervasive execution known as intelligent workflows (Wieland et al. 2008). Frameworks that enhance process awareness and self-adaptation are proposed (Seiger et al. 2019; Pryss et al. 2015; Friedow et al. 2018; Malburg et al. 2020; Engels et al. 2018). To support context variables and sensors as well as communication paradigms for IoT, Domingos et al. (2014) extend the WS-BPEL workflow language, dynamically selecting IoT services based on availability, functionality, and context (Seiger et al. 2018; Elali et al. 2022). Other studies focused on the adaptability of BPs during the execution phase. A dynamic integration mechanism for coping with changes in BPs is proposed by Hu et al. (2014); Domingos et al. (2010); Tôn and Lê (2019). An application to the predictive process monitoring field using IoT events as a process context and developing a predictive model to predict the next activity is provided by Elhami et al. (2020). A new trend in running the IoT-aware BP is running on fog computing.Footnote 8 In this context, Cheng et al. (2018) dealt with IoT-aware BPs at the execution level, where they introduced a new intermediate layer consisting of a set of distributed fog nodes to perform certain parts of the process.

  • Topic 4 – IoT-aware process modeling To create IoT-aware BPs and exploit the full potential of IoT and BPM, this must be integrated into process models. Modeling strategies for IoT-aware BPs fall into two categories (Brouns et al. 2018): those utilizing known modeling languages and those proposing new domain-specific languages. The former includes integrating BPMN with ontologies (Valderas et al. 2022; Di Martino et al. 2022; Valderas et al. 2023; Bertrand et al. 2021; Elkodssi et al. 2022) and extending BPMN with IoT devices (Chiu and Wang 2015; Kirikkayis et al. 2022b; Hornsteiner and Schönig 2023), as well as using Petri Nets (Varga et al. 2018), or DMNFootnote 9 (Hasić et al. 2020; Kirikkayis et al. 2023b). In the latter, Chen et al. (2012) provide a new process definition language for IoT-enabled BPs, encapsulating physical devices as SOA services. IoT device interaction is critical. To this end, the application of mixed reality as a new interaction paradigm to facilitate the modeling and configuration of processes among IoT devices was elaborated on Seiger et al. (2021). From an access IoT data perspective, Cherrier and Deshpande (2017) present a gateway for transferring IoT events to BPs, managing device heterogeneity. Meyer et al. (2013) map IoT concepts to process models, and Martins and Domingos (2017) use BPMN to model IoT device behaviours. Other studies focus on collaborative structures to improve physical and digital collaboration between multiple actors to achieve a common goal (Maamar et al. 2020; Shamsuzzoha et al. 2014; Li et al. 2021; Wang et al. 2022). For example, Grefen et al. (2019) discuss synchronizing physical objects for successful IoT-aware digital processes.

4.3 Which Application Domains are Used to Study IoT and BPM Integration?

Question number three was answered based on data from research conducted by HSPIFootnote 10 Accessed 6 January 2022. Thus, from the HSPI process mining application database, we collected and analyzed the studies to map the applications provided through different industries and markets and understand the adoption of BPM and IoT. The application domains which the found studies refer to are (cf. Figure 3): Healthcare (17.6%); Industrial sector (41.2%); Public (7.1%); Consumer Services (5.9%); Consumer Goods (3.5%) and Services (24.7%). Only eight studies (11%) do not refer to any application domain.

Fig. 3
figure 3

Distribution of selected studies by application domains

  • Healthcare Most healthcare applications are related to remote patient monitoring systems, within hospitals or at home, and constantly enriched with IoT devices (Antonius and Dachyar 2020; Cheng et al. 2018; Jain and Tata 2017; Pryss et al. 2015; Seiger et al. 2018; Vitali and Pernici 2016; Ruiz-Fernández et al. 2017; Ruppen and Meyer 2013; Tôn and Lê 2019; Kirikkayis et al. 2022b; Gómez-Valiente et al. 2023). Maamar et al. (2018) use the example of a hospital that is on high alert being close to a severe car accident, where departments equipped with environmental sensors or smart wrists, enable the real-time transmission of patients’ vital signs to appropriate recipients, demonstrating how IoT can facilitate operations and improve efficiency. Senderovich et al. (2016) present an example of a process of treatment in a day hospital for cancer patients through Real-Time Locating System (RTLS) receivers that monitor all corporate entities involved in the process (e.g., patients, doctors, nurses) as well as some of the medical devices that record the data emitted and use it for real-time monitoring of process entities and equipment. Another frequent scenario is elderly care. Integrating the real-time event streams from bed pressure sensors into the BP field can reflect the health status of the elderly in real-time and thus monitor them through immediate warnings (Li et al. 2021).

  • Industrial sector In the domain of transportation, IoT might be used to realize monitoring of goods, such as dangerous or perishable goods (Domingos et al. 2010, 2014, 2015; Valderas et al. 2022; Suri et al. 2017; Valderas et al. 2023; Gallik et al. 2022; Seiger et al. 2023; Diamantini et al. 2023). These systems react to events that occur during transport using sensors and location technologies to improve the monitoring of transport conditions, such as temperature and pressure (Mass et al. 2016). Sensors and smart devices have great potential to provide further automation in various domains including logistics (Meroni et al. 2018; Gao et al. 2011; Grefen et al. 2019; Zanfack et al. 2015; Wombacher 2011; Schief et al. 2011; Song et al. 2022). Take, for example, the maritime transport scenario of heavy fog where disasters can be prevented in sea-land transport (Wang et al. 2022). Montali and Plebani (2017) use an example of a seafood company that organizes fish delivery and aims to verify if all actors operate correctly because of the complexity of the delivery process (deviations to the plan may occur). Several studies focus on smart factories as representatives of IoT environments (Grambow et al. 2021; Bocciarelli et al. 2017; Malburg et al. 2020; Wieland et al. 2008; Seiger et al. 2020; Kirikkayis et al. 2023c, a, b, 2022a). In the manufacturing industry, BPM and IoT support could improve both management by closely linking digital production and machine data (Schönig et al. 2018, 2020; Shamsuzzoha et al. 2014; Keates 2019) and cybersecurity by integrating security requirements into modeling using a ”security by design” paradigm (Hornsteiner and Schönig 2023).

  • Public Cheng et al. (2019) develop an actual sensor-aware BP application and validate it by using the designed system to protect a large area of forest in North China. Graja et al. (2019) evaluate their study through an example of disaster recovery systems. Maamar et al. (2020) present a city that runs many systems like transportation for traffic control and environment for air pollution monitoring; Marrella and Mecella (2017) rely on a case study previously conducted by the same authors in (Humayoun et al. 2009b, a; Marrella et al. 2011) to evaluate their study in an emergency management domain, in which teams of first responders act in disaster locations with the primary purpose of assisting potential victims and stabilizing the situation. Chadli et al. (2022) propose a classroom case study involving the deployment of hardware device and management software to measure the environmental parameters and control ventilation, lighting, and air conditioning. Pastor et al. (2022) employ a smart camera with active sensors for vehicle identification in road traffic.

  • Consumer Services Kahl et al. (2015) test their study for common BPs in the retail domain within the innovative retail lab of a large European supermarket chain; Meyer et al. (2011) evaluate their study using a use case from the domain of retail and show how sensors monitor perishable goods in a store; Meyer et al. (2013) test their study using a dynamic pricing process in the retail domain and show how the IoT device temperature sensor monitors the perishable good orchid in a store; Suri et al. (2018) evaluate their study using a process family from the Retail/Supply Chain Management domain; Ugljanin et al. (2018) test their study using a Smart City Tourism Organization (SCTO) scenario to automate its BPs related to visitor mobility by communicating with them using social networks and collecting their feedback.

  • Consumer Goods Friedow et al. (2018) use a simple coffee machine billing system to automate the process of counting the coffee amount for each user; Kunz et al. (2011) use the lifecycle of a fish fillet within the different domains of manufacturer, logistics service provider, wholesaler distribution hub, wholesaler store, and the customer; van Eck et al. (2016) use a case study performed at Philips where a smart baby bottle has been developed.

  • Services Scenarios that use smart device detection and implementation capabilities in smart home environments have been extensively addressed (Janssen et al. 2020; Loke et al. 2007; Wehlitz et al. 2017; Xing et al. 2012; Seiger et al. 2019; Hu et al. 2014; Chiu and Wang 2015; Elali et al. 2022). Examples of instances where service applications play a crucial role include the intruder detection system in a security scenario (Kikuchi et al. 2018); office supply processes (Cherrier and Deshpande 2017), temperature control processes (Seiger et al. 2021), warehouse management system monitoring processes (Hou et al. 2016), automatic irrigation control processes (Martins and Domingos 2017; Martins et al. 2020; Di Martino et al. 2022), processes for adaptive ventilation in a student dormitory (Mottola et al. 2019) or on student thesis defence (Zhu et al. 2014), processes for repair and maintenance of computers and peripherals (Muhsin et al. 2016); while Park et al. (2018) implement smart toilets for companion animals by simply attaching sensors and introducing BPM technology.

Table 6 Studies distribution by application scenario (RQ3)

4.4 What are the Research Challenges Addressed So Far to Support the Management of IoT-Aware BP?

In this section, we present the results related to research question RQ4. Enriching BPs with IoT technologies brings several advantages and many challenges (Janiesch et al. 2020); therefore, we provide the distribution of the studies according to the IoT-Meets-BPM Manifesto (cf. Figure 4 and Table 7). Not all studies face one of these challenges, actually, only 38% of them do.

Fig. 4
figure 4

Distribution of the studies by research challenges addressed in the IoT-Meets-BPM Manifesto

  • C1 – Placing sensors in a process-aware way The research challenge C1 is considered by 25% (i.e., 12) of the studies. The link between the position of the sensors and that of the corresponding actuator is discussed in Mottola et al. (2019). Using a physical simulation model, Malburg et al. (2020) conducted research in BPM and Industry 4.0, aiming to program a smart factory for BPs by selecting hardware components to detect process-relevant events. To place IoT sensors in a process-aware way and link them to the running process, a BPMS must be aware of the current values of IoT objects, and based on an established mapping from IoT variables to process models IoT data is sent to a BPMS (Schönig et al. 2020). Process model discovery from sensors or wearable devices depends on their location and amount (Koschmider et al. 2020; Senderovich et al. 2016; van Eck et al. 2016). Other interesting studies on how to integrate sensors with other services and into existing BPs have been conducted (Ruppen and Meyer 2013; Gallik et al. 2022; Song et al. 2022; Di Martino et al. 2022). Regarding sensors, the presented ontological model covers domain configuration, including IoT resources’ placement and gathered observations over time (Diamantini et al. 2023; Valderas et al. 2023).

  • C3 – Connection of analytical processes with IoT The research challenge C3 is considered by 8.3% (i.e. 4) of the studies. Elhami et al. (2020) use contextual events in the decision point to control the process execution and incorporate contextual changes with process rules and exaction logic at the runtime. With the proposed extension Kirikkayis et al. (2023b) can specify event-driven behaviour at various points in a BP but also demarcate IoT-based data sources from other process inputs for monitoring and error handling. BP execution necessitates updated information availability. Li et al. (2021) proposed a framework separating event processing and BP execution, reducing BPM engine load.

  • C4 – Integrating the IoT with process correctness checks The research challenge C4 is considered by 4.2% (i.e. 2) of the studies. By connecting individual tasks and workflow executions to the respective effects in the physical and virtual world with the help of additional sensor data, you can verify the correct execution and behaviour of the IoT entities involved (Seiger et al. 2018; Malburg et al. 2020).

  • C5 – Dealing with unstructured environments The research challenge C5 is considered by 6.3% (i.e. 3) of the studies. In the IoT world, which is much more ad hoc and situative, Domingos et al. (2010) introduce mechanisms to perform ad-hoc changes in IoT-aware BPs by identifying change primitives that support the change operations needed to modify parts of the BP.

  • C6 – Managing the link between micro-processes The research challenge C6 is considered by 6.3% (i.e. 3) of the studies. An example is to achieve an efficient and flexible production line by investigating the micro-processes at the individual machines and stations and their interconnections to achieve a more flexible composition of smaller processes (Malburg et al. 2020). By extending BPMN 2.0, Gao et al. (2011) links process models with business functions, connecting them to external open data, including sensor and instance ontology data; another extension suggest using low-level IoT data for decision-making, not including high-level data aggregation (Kirikkayis et al. 2023c).

  • C7 – Breaking down end-to-end processes The research challenge C7 is considered by 4.2% (i.e. 2) of the studies. Malburg et al. (2020) suggest relaxing and detailing the static and coarse-grained “hardwired” processes to achieve a more flexible composition of smaller processes. Kahl et al. (2015) propose a system that provides an agent-assisted realization and adaptation of BPs with semantic service selection and facilitates an event-driven selection and controlled execution of relevant BPs in intelligent environments.

  • C9 – Specifying the autonomy level of things The research challenge C9 is considered by 2.1% (i.e. 1) of the studies. Ensuring appropriate autonomy for resource constrained IoT device is investigated by Malburg et al. (2020).

  • C12 – Dealing with new situations The research challenge C12 is considered by 2.1% (i.e. 1) of the studies. Malburg et al. (2020) use past successful process executions to automatically learn possible adaptations of process instances to deal with new or similar situations.

  • C13 – Bridging the gap between event-based and process-based systems The research challenge C13 is considered by 25% (i.e. 12) of the studies. The complex IoT system is event-driven due to the large number of sensors and process-based (Malburg et al. 2020). While in the past, process models have been detected by documents or interviews with domain experts, today the challenge of automatically discovering the process model from the network of sensors or wearable devices also depends on their location and amount (Koschmider et al. 2020; Senderovich et al. 2016; van Eck et al. 2016). Therefore, considering the physical context and revealing the correlation between IoT data and process events is crucial (Bertrand et al. 2021).

  • C14 – Improving online conformance checking The research challenge C14 is considered by 6.3% (i.e. 3) of the studies. Seiger et al. (2020) mainly focus on generating event logs from streams of IoT sensor data with smart factories as an application domain.

  • C15 – Improving resource utilization optimization The research challenge C15 is considered by 10.4% (i.e. 5) of the studies. Suri et al. (2017) present a framework that formalizes IoT properties and rules to optimize resource management in BPs, defining resource constraints to be mapped to a task during the design phase.

Table 7 Definitions of the research challenges of the IoT-Meets-BPM Manifesto

5 Discussion

Table 8 Trends for research studies to support IoT-aware BPs life cycle phases

In this section, we discuss the relationships between the topics extracted from the studies, the application domains, the life cycle phases of the BPM and the research challenges introduced by Janiesch et al. (2020). Finally, the main issues which seem to need further investigation are discussed.

\(\blacksquare \):

BPM life cycle phases Our results suggest that many studies include more than one phase, and most (43%, 40 studies) combine the modeling and implementation phases. The impact of IoT technology on the modeling phase has been widely studied, giving rise to the so-called IoT-aware BP modeling that has attracted the attention of researchers. Introducing IoT entities into the modeling phase has encountered difficulties with the entities’ heterogeneity. Analysis of the studies showed that the approaches used to model IoT entities involve: the expression of entities through the use of standard elements of process modeling languages, although increasingly rarely, or more commonly, the extension of languages through the introduction of specific IoT elements. However, the existing BPM solutions have not yet precisely defined the behaviour of IoT entities, and this absence proves a significant obstacle to running BPs of traditional BPM systems on IoT devices automatically. The analysis showed that an integrated workflow engine approach can rapidly change the model and its execution, avoiding the situation where the process model changes. The system must re-convert the model into an executable source code and distribute the code to IoT devices. A second common approach involves the implementation of a middleware that makes BP independent of the device’s technology. However, the proposal of an approach that covers most of the phases of an IoT-aware BP is the most critical challenge to date. Based on the findings, we have defined a set of trends (or directions for research studies) for the life cycle phases of IoT-aware BPs (cf. Table 8).

  • Topic Modeling One of the biggest challenges we have encountered in topic analysis, which is to extract knowledge from sensor data, is to bridge the gap between sensor data clouds and event logs. To be more precise, the goal is to identify events from raw event data, discover their activities and correlate them with the process instance. Another interesting topic that emerged from the analysis concerns the modeling approaches of IoT-aware BPs. For example, based on the review conducted in this article, we have seen several approaches to modeling IoT devices. Most authors modeling IoT-aware BPs propose methods that rely on a BPMN extension to integrate IoT devices within the process model. Another interesting approach uses a linked data mechanism to create links between the BP model and external open data, including the ontology of sensors and instance data, to achieve better integration between the physical world and BPs. Many authors see IoT as the external data source for BPM, most commonly in terms of process context. Physical resource management was the last topic of interest in the BPM and IoT field. We have seen that it is essential to specify the resources used to perform tasks because IoT resources are active. However, the challenge remains to formalize the relationship between IoT resources to ensure efficient management.

  • Application domains and case studies Our results show that there are application domains that benefit more than others from the integration of IoT concepts into BPM application scenarios. Among these, the industrial categories (41.2%, 35 studies) and services (24.7%, 21 studies) are the most used as application scenarios in BPM and IoT. Within the industrial category, we have further classified studies into several sectors, and we have seen that many studies have introduced BPM for IoT-based logistics. It emerged that filtering and processing events are important but are only the first steps to obtaining transparent BPs. It is necessary to consider the quality of the information because the data in the logistics area are often incomplete and inconsistent. Data quality is a topic that needs to be addressed if you want to build approaches that integrate IoT into the BPM community. Existing solutions presuppose persistent and relatively static data sets contrary to the physical world’s needs. Notice that among these only a few studies utilize real scenarios, such as a temperature controlling process (Chiu and Wang 2015), a small scale physical smart factory model (Kirikkayis et al. 2022a), a smart hospital (Jain and Tata 2017) or a clinical process of the hypertension (Ruiz-Fernández et al. 2017). Furthermore, two studies propose that their future work will be dedicated to devising a comprehensive evaluation of the real-case approach (Maamar et al. 2018; Diamantini et al. 2023).

  • Research challenges Another essential aspect that we have observed in some of the studies analysed (38%, 35 studies) is the discussion of the challenges presented in the BPM-IoT Manifesto (Janiesch et al. 2020). Complex IoT environments consist of many sensors, actuators, and control units. In addition to IoT components, a further software stack is also needed to elevate programming and research to the level of abstraction of BPs and thus exploit the potential of BPM’s integration with IoT. This type of system can often be driven either by events, due to a large number of sensors, or by the process (C13 - Sect. 4.4). One of the research activities associated with this challenge concerns the functionality analysis of the available sensors and actuators and their grouping and abstraction at the BPM-oriented level. Identifying process events from IoT data, perfecting them, and generating associated events is often insufficient. What may remain are ambiguities and uncertainties, typical of the nature of IoT environments, as part of the event log and this must be considered in the following analysis steps (e.g. in the compliance check). The analysis of the research challenges also showed that it is necessary to place IoT sensors in a process-aware way and that these must be connected to running processes (C1 - Sect. 4.4). To date, this challenge mainly involves BPMS which must be aware of the current values of IoT objects. In addition, it is also necessary to investigate how to acquire current values from different data variables and obtain a mapping between IoT variables and process models so that IoT data can be sent to a BPMS. This process is fundamental as, on the one hand, based on the current values of some variables, tasks are activated or canceled and decisions are made; on the other hand, this allows IoT data variables to be more configurable and traceable.

5.1 Threats to Validity

The SLR’s validity should be rigorously assessed to ensure the level of scientific value in its conduct and findings. In this section, we discuss the factors (threats to validity) that dispute the results of an SLR or decrease the meaningfulness of the results. Our review adopted the guidelines specified by Zhou et al. (2016) in assessing the validity of Software Engineering. The authors classify validity categories and associate identified threats with them. One of the main threats to the validity of this SLR is incompleteness. In our case, what risk this threat poses depends heavily on the limitations of available search engines. A multidisciplinary search engine such as GS and two more specific ones, ACM and IEEE Digital Libraries, should significantly decrease such constraints. Another critical issue is bias in the selection of studies which could lead to inaccuracies in the data. To solve this, we employed various strategies. The reviewer thoroughly read the documents during the selection process to minimize misinterpretations from relying solely on titles and abstracts. Additionally, explicit reasons for exclusion decisions were documented during the screening phase. This approach aided in mitigating the threat of overlooking relevant studies. To maximize study coverage, we employed snowballing as a complementary search to reduce the possibility of missing relevant studies. Another action taken, which could be traced back to a wrong search method (Zhou et al. 2016), was the combination of automated and manual search methods.

6 Concluding Remarks

This article has provided an in-depth overview of the interplay between BPM and IoT. Unlike the existing surveys that deal with specific aspects of the interaction, we aimed to provide a holistic view in this study. The systematic analysis of reviewed studies provides answers to several research questions, enabling us to (i) identify the hottest topics studied within the literature (Mining from event sensors, Physical resource management, Context-aware execution, IoT-aware process modeling), (ii) classify the integration of IoT within the BP life cycle phases (Identification, Modeling, Analysis, Re-design, Implementation, and Monitoring and mining), (iii) identify which are the application domains where integration between IoT and BPM have the most significant expansion (Consumer goods, Consumer services, Healthcare, Industrials, Services and Public), and (vi) highlight the research challenges provided by the IoT-Meets-BPM Manifesto faced so far to support the management of IoT-aware BPs.