=Paper=
{{Paper
|id=Vol-3413/paper8
|storemode=property
|title=ProAmbitIon: Online Process Conformance Checking with Ambiguities Driven by the Internet of Things
|pdfUrl=https://ceur-ws.org/Vol-3413/paper8.pdf
|volume=Vol-3413
|authors=Marco Franceschetti,Ronny Seiger,Mauricio Jacobo González González,Enrique Garcia-Ceja,Luis Armando Rodríguez Flores,Luciano García-Bañuelos,Barbara Weber
|dblpUrl=https://dblp.org/rec/conf/caise/FranceschettiSG23
}}
==ProAmbitIon: Online Process Conformance Checking with Ambiguities Driven by the Internet of Things==
https://ceur-ws.org/Vol-3413/paper8.pdf
ProAmbitIon: Online Process Conformance Checking
with Ambiguities Driven by the Internet of Things
Marco Franceschetti1,*,† , Ronny Seiger1,† , Mauricio Jacobo González González2 ,
Enrique Garcia-Ceja2 , Luis Armando Rodríguez Flores2 , Luciano García-Bañuelos2
and Barbara Weber1
1
Institute of Computer Science, University of St.Gallen, Switzerland
2
Department of Computer Science, Tecnológico de Monterrey, Mexico
Abstract
The ongoing digitization of processes in everyday life shows great potential for process automation,
analysis, and optimization. However, digital traces of processes in the physical world, especially those
involving human interactions, are often incomplete. This limits the possibilities for an automated
process monitoring and analysis. ProAmbitIon proposes to use the Internet of Things (IoT) to bridge
the gap between physical world process executions and their digital traces. In this project we leverage
software-controlled sensors and actuators to enable a fine-grained monitoring and contextualization
of process activities. Digital traces of executed processes can be created from and enriched with IoT
data, and used for conformance checking to detect deviations–even at runtime and without relying on a
Business Process Management System (BPMS). In developing new approaches for IoT-driven process
conformance checking, we also address the issue of potential ambiguities originating from 1) informal
process descriptions and 2) the lack of process-related data in IoT data. The project is conducted using
real-world scenarios from smart healthcare and smart manufacturing.
Keywords
Process Mining, Conformance Checking, Explainability, Internet of Things, Ambiguous Process Models
1. Introduction
Processes and process-like descriptions are being increasingly adopted in every domain and
aspect of everyday lives. They are used to instruct humans, computers, machines, robots, and
all kinds of other resources how to interact and execute specific activities to solve a certain task.
These real-world processes have become increasingly digitized [1]. Contextually, in recent years
process mining became an important and mature discipline with broad adoption in industry.
RPE@CAiSE’23: Research Projects Exhibition at the International Conference on Advanced Information Systems Engi-
neering, June 12–16, 2023, Zaragoza, Spain
*
Corresponding author.
†
These authors contributed equally.
$ marco.franceschetti@unisg.ch (M. Franceschetti); ronny.seiger@unisg.ch (R. Seiger); A00814078@tec.mx
(M. J. G. González); enrique.gc@tec.mx (E. Garcia-Ceja); lrodriguez@tec.mx (L. A. R. Flores); luciano.garcia@tec.mx
(L. García-Bañuelos); barbara.weber@unisg.ch (B. Weber)
0000-0001-7030-282X (M. Franceschetti); 0000-0003-1675-2592 (R. Seiger); 0000-0002-4001-5026
(M. J. G. González); 0000-0001-6864-8557 (E. Garcia-Ceja); 0000-0001-9076-903X (L. García-Bañuelos);
0000-0002-6004-4860 (B. Weber)
© 2023 Copyright for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
CEUR
Workshop
Proceedings
http://ceur-ws.org
ISSN 1613-0073
CEUR Workshop Proceedings (CEUR-WS.org)
Process mining bridges process science and data science to extract event data about process
executions stored in event logs for an automated analysis of processes [2].
Conformance checking is a process mining task conducted to automatically detect deviations of
process executions (cases) from the prescribed underlying formal description (process model) [3].
Conformance checking assumes the existence of an event log describing execution traces. This
assumption is realistic when the process is executed using a Business Process Management
System (BPMS) or some other kind of process-aware control/monitoring system. However, in
settings where process executions are not fully orchestrated or at least monitored by a single
BPMS, the execution of certain steps may be outside the control of an IT system. Hence it
might be challenging to obtain complete event logs. For instance, the digital traces of processes
with a high degree of human involvement (e.g., in healthcare) are often on an abstract and
coarse-grained level, containing human activities as black-boxes, incomplete activity sequences,
and even erroneous data [4].
The gap between the physical execution of a process and its digital representation limits
the possibilities of conformance checking for a detailed automated analysis with respect to
deviations and exceptions [5]. As a result, these processes cannot be fully planned, predicted,
optimized, or adapted in case of non-conformance. Additionally, most of the existing approaches
and techniques related to conformance checking focus on the offline (a-posteriori) analysis
of event logs [3]. This further limits the applicability of conformance checking for a timely
analysis of processes executed in the physical world. Moreover, process descriptions for humans
are often not provided as formal process models but in more informal ways such as guidelines,
checklists or policies [6]. These informal descriptions may be interpreted in multiple valid ways,
resulting in ambiguities of the process models and executions [7]. These ambiguities, as well as
ambiguities from event logs (cf. [8]), must be considered for conformance checking.
We propose to use the Internet of Things (IoT) as an enabler for improved monitoring and
conformance checking of real-world processes that may exhibit ambiguity. With the adoption of
software-controlled IoT sensors and actuators, we are able to sense physical process executions
and generate corresponding digital representations (traces), enabling conformance checking for
those aforementioned settings where event logs are not available or incomplete.
The goals of the ProAmbitIon1 project are to:
1. Investigate how to enable domain experts to enrich process descriptions with IoT-related
execution aspects and use these for the monitoring of process elements (e.g., activities).
2. Investigate how conformance checking can be used in the presence of ambiguities to check
the correctness of process executions and provide interpretable feedback to end-users.
Towards these goals, four research questions drive the project development:
• RQ1: How can domain experts be enabled to enrich informal process descriptions with
IoT-related execution criteria for event abstraction and correlation?
• RQ2: How can stream analysis techniques be used to derive process event logs and event
streams from possibly ambiguous IoT data?
• RQ3: How can conformance checking be used in both offline and online settings on
IoT-based process event logs and streams that possibly contain ambiguities?
1
https://data.snf.ch/grants/grant/208497
• RQ4: How can end-users be provided with understandable feedback about conformance
regarding process execution and with means for resolving remaining ambiguities?
This research project is funded by the Swiss National Science Foundation (SNSF) within the
SPIRIT framework. The Principal Investigators (PIs) are Barbara Weber and Luciano García-
Bañuelos. The researchers of the PIs’ teams are supported by an international network of
highly renowned researchers, namely Andrea Burattin, Claudia Lizette Garay-Rondero, Jasmin
Niess, Manfred Reichert, Stefanie Rinderle-Ma and Matthias Schlegel. The project duration is
three years (01.11.2022–31.10.2025).
2. Project Objectives and Tangible Outputs
The project proposes to use the IoT as an enabler for an improved monitoring and conformance
checking of ambiguous processes. The overarching objective is to develop a comprehensive
framework that allows: 1) annotating informal process descriptions (e.g., activities) with moni-
toring points linking to IoT devices; 2) online monitoring and conformance checking of process
executions through the associated IoT devices while managing ambiguity; and 3) providing
interpretable and interactive feedback on the conformance.
The project will be conducted following design science research principles, and is expected
to produce a number of artifacts:
• A framework for IoT-driven process event log and event stream generation based on
annotated process descriptions.
• A domain-specific modeling language and tool to annotate and translate informal process
descriptions with monitoring points following a low-code approach.
• A catalog of event patterns relating process activities with IoT data to support the anno-
tation of process descriptions.
• A framework for the online analysis of IoT event streams to abstract to process events
based on Complex Event Processing (CEP) [9].
• An approach for online conformance checking considering ambiguities and providing
interpretable user feedback.
• Alignment techniques for process models and process event streams generated from IoT
event streams in the presence of ambiguities.
• Mechanisms for providing interactive feedback about process conformance to show and
resolve ambiguities.
3. Relevance of the Project
Several challenges at the intersection of the BPM and IoT research communities have been
identified in the BPM-IoT manifesto [1]. With this project, we address multiple open challenges
related to using IoT data to analyse the execution of business processes. We expect to advance
the state-of-the-art of information systems engineering as follows.
Considering IoT devices as novel data sources for process activity detection (cf. [10]), we
will be able to monitor and analyze processes executed even in the absence of a BPMS or other
Process-Aware Information System (PAIS) acting as the execution orchestrator or monitor.
Additionally, the developed monitoring and analysis methods will be useful to complement
BPMSs or PAISs, advancing the capabilities of existing systems and enriching process event
logs with IoT data [11]. This will help closing the gap between the physical and digital world
representations of processes, and bridging event-based and process-based systems [1].
To abstract IoT events to process events (cf. [9]), we will elaborate on the annotation of process
descriptions with monitoring points associated with IoT devices and IoT event patterns [12].
The artifacts designed for defining monitoring points and the respective event abstraction
procedures will simplify the generation of process event streams from IoT data. The IoT-based
event patterns, in analogy to [13], will result in a patterns catalog that will be useful for system
analysis in terms of patterns support and for the systematic development of event-based systems.
Considering processes with ambiguity, we will target the healthcare domain, which is charac-
terized by high human involvement and expertise, and semi-informal process descriptions. We
will develop procedures for checking process conformance that take ambiguity into account and
provide interpretable feedback to the end-users. The project will relax the usual assumption that
conformance checking is done in an offline setting and develop online conformance checking
procedures to detect deviations also in the case of partial traces. This will require extending the
notion of conformance (e.g., to partial [14]). Thus, with the project we will advance the state of
the art of conformance checking with respect to concepts, procedures, and results presentation.
4. Current Project Status
The project explores novel approaches for IoT-driven conformance checking of ambiguous
processes [10]. Two application domains are considered for the development and validation of
these approaches: smart manufacturing and healthcare. Processes in these domains present
significant differences, e.g., in terms of repetitiveness (resp. variability), structuredness, or
degree of automation. Considering these rather diverse domains allows us to demonstrate
the general nature of the solutions we will develop with this project. The first milestone for
ProAmbitIon concerns the development of a suitable scenario for the smart manufacturing and
for the smart healthcare domains. The achievement of this milestone is presented here.
4.1. Process in the Smart Manufacturing Domain
The domain of smart manufacturing is becoming increasingly penetrated with IoT technologies
to achieve high flexibility and high throughput production processes. Production machines
are equipped with a multitude of sensors and actuators producing high volumes of IoT data
at various levels of granularity that can be accessed via open interfaces [15]. This data may
range from low-level sensor readings to machine states to process-related information emitted
from the PLCs (Programmable Logic Controllers) and MES (Manufacturing Execution Systems)
controlling one or multiple production machines [16]. However, data quality and frequency
may vary in production environments consisting of different types of machines and other IoT
devices that are not controlled by one single MES. Legacy machines may only be equipped
with a small number of sensors. Standalone devices such as robots or environment sensors
may have their own PLCs or other forms of controllers that are not integrated with the rest
of the production control. Moreover, despite the high degree of automation, manual steps
and human-machine/human-robot collaborations may be a necessary part of a production
process that can only monitored by sensors to a limited degree. Thus, deriving a consistent and
homogeneous event log suitable for conformance checking from a typical smart manufacturing
environment without having a central process orchestration–or at least process monitoring
component–in place becomes a very challenging task [17, 18].
As a scenario we take an order-to-product process simulated using our smart factory model [19,
15]. A typical production process starts with storing new raw materials in a warehouse, and
then receiving orders for a specific type of product identified by its color. The corresponding raw
materials are retrieved and produced following a sequence of steps: Unloading a raw workpiece
from a warehouse – Check the quality of the raw workpiece – Transport the workpiece to an oven –
Bake the workpiece – Transport the workpiece to a milling machine – Mill the workpiece – Sort
the product – Check the product quality manually – Transport the product to pickup. As this
process follows a strict sequence of activities without a high degree of variability, we will adopt
a BPMN-based approach to represent the process [15]. From this rather simplified simulation
environment we will gradually move the process analysis to more sophisticated and realistic
production environments provided by our associated project partners [20].
4.2. Process in the Smart Healthcare Domain
The healthcare sector is one of the most promising industries for BPM, with numerous un-
dergoing endeavors as healthcare organizations strive for improved process effectiveness and
efficiency [21]. Healthcare processes are typically challenging as they are knowledge intensive,
loosely structured, mostly manual, and executed without the support of a PAIS [22]. Addition-
ally, they require high flexibility to adapt to unforeseen emergency situations of individual
patients. It is not realistic to orchestrate healthcare processes with a PAIS, and monitoring and
conformance checking of processes in such a domain is challenging (e.g., due to privacy issues).
To identify a suitable scenario from the healthcare domain, we held a workshop with domain
experts from the Division of Infectious Diseases & Hospital Epidemiology of the Cantonal
Hospital of St. Gallen. The Division is particularly concerned with the adherence to hand
hygiene guidelines, since health organizations have been calling for increased attention to hand
hygiene as an effective means to prevent infections [23]. The scenario developed after the
workshop was validated in a second workshop involving the same group of domain experts.
The scenario is based on a blood donation process, for which guidelines from the World Health
Organization are available [24]. Focusing on the adherence to the hand hygiene guidelines
from [23] in the context of the blood donation process, we consider a simplified version of
the process composed of 9 steps in sequence: Perform hand hygiene before touching the donor
– Perform preliminary operations – Perform hand hygiene before aseptic procedure – Perform
venipuncture – Monitor donor – Remove needle – Perform hand hygiene after potential exposure to
body fluids – Perform final operations – Perform hand hygiene after touching patient’s surroundings.
Here, the guidelines from [23] state indications for hand hygiene steps to be executed before or
after the execution of some blood donation process steps: for instance, hand hygiene must be
performed before touching a donor different from the one just touched (e.g., at the beginning of
the process), or after a potential exposure to body fluids.
Following the guidelines, the process may look purely sequential; however, in reality the
inherent characteristics of the healthcare environment demand for frequent and unpredictable
disruptions to the prescribed flow. Such disruptions stem from unforeseen emergency situations
(e.g., a donor fainting). They require the healthcare worker to deviate from the sequential
execution, interleave the treatments of different donors (i.e., process instances), or perform
additional hand hygiene steps before being able to proceed with the prescribed flow of operations.
Detecting in particular unforeseen hand hygiene indications is a challenge we take on here.
Additionally, often multiple instances of the process coexist in space and time, which makes the
IoT-driven monitoring and conformance checking challenging, since certain IoT data–activity–
process instance associations might not be obvious. For instance, from just the sensed position
of a healthcare worker it might not be clear, which donor in the proximity is being treated.
4.3. Current Work
Currently, we are investigating the detection of process activities from IoT data using our smart
factory model. A first approach following an interactive method that relies on domain expert
knowledge has been published in [25] and extended in [26]. The method is based on the manual
annotation of an IoT event stream with markings for the start and end of activity executions.
These annotations define activity signatures, i.e., patterns over a multivariate time series that
represent the activity execution in the IoT data [26].
Based on these first results, we are moving towards automating the activity detection ex-
ploring different approaches in parallel. On the one hand, given a non-annotated IoT data
set, we are investigating how to automatically infer patterns for activity signatures based on
unsupervised learning methods by finding the start and end times of activities as they occur.
This is accomplished by dividing the activities into ‘micro-activities’; information about the
frequencies of micro-activities is then used to infer the entire activities. On the other hand, given
an activity signature, we are studying how to automatically generate CEP-based applications
for the detection of similar activities. Changes in the time-series of the signature are determined
and translated into CEP queries in a CEP-based language [10]. The resulting CEP apps are
deployed to a CEP engine for online activity detection where the CEP app matches the incoming
IoT event streams with the sequence of changes encoded in the queries.
Additionally, we currently work on a formal conceptualization of IoT-driven process mon-
itoring in the absence of a BPMS/PAIS. Here we define requirements for IoT-based events to
support process monitoring and formalize a monitoring meta-model. These contributions will
serve as a formal foundation for the framework for IoT-driven process event log and event
stream generation based on annotated process descriptions (cf. Sect. 2).
We are also developing a laboratory environment to simulate the execution of the blood
donation process in a controlled setting. We will deploy a set of IoT devices (e.g., presence and
proximity sensors, touch sensors) to monitor the execution of the simulated process. We will
also design a formal model for the process as the reference for conformance checking. To this
end, we plan to adopt DCR graphs as a declarative event-based language [27]. This type of
language is particularly suitable for processes requiring high degrees of flexibility. DCR graphs
follow an alternative paradigm to the imperative one used for the smart manufacturing process,
which is in line with our intention to demonstrate the general nature of the developed solutions.
5. Conclusion
Emerging IoT technologies promise to enhance the capabilities of Business Process Management
Systems, in particular for real-world settings where digital traces are incomplete or unavail-
able. With ProAmbitIon, we are exploring novel approaches for IoT-driven online conformance
checking of processes with ambiguities at different levels (e.g., in the process descriptions
or executions logs). Here we expect several contributions to advance the state-of-the-art in
information systems engineering. In the project, we consider scenarios from smart manufactur-
ing and healthcare as two application domains that could benefit from the integration of IoT
technologies with process execution and vice versa. The diverse nature of these domains will
demonstrate the general applicability of the solutions developed in the course of the project.
First achievements include the definition and validation of two scenarios from smart health-
care and manufacturing. In addition, an interactive method to detect activities from IoT data
has been developed. Research on how to automate the activity detection method is currently
ongoing, along with a formal conceptualization for IoT-driven process monitoring. Further
ongoing work concerns the setup of a simulation environment for the healthcare process.
Acknowledgments
This work has received funding from the Swiss National Science Foundation under Grant
No. IZSTZ0_208497 (ProAmbitIon project).
References
[1] C. Janiesch, A. Koschmider, M. Mecella, B. Weber, A. Burattin, C. Di Ciccio, G. Fortino,
A. Gal, U. Kannengiesser, F. Leotta, et al., The internet of things meets business process
management: a manifesto, IEEE Systems, Man, and Cybernetics Magazine 6 (2020) 34–44.
[2] W. Van Der Aalst, Process mining: data science in action, volume 2, Springer, 2016.
[3] J. Carmona, B. van Dongen, A. Solti, M. Weidlich, Conformance checking, Springer (2018).
[4] S. Remy, L. Pufahl, J. P. Sachs, E. Böttinger, M. Weske, Event log generation in a health
system: a case study, in: Business Process Management: 18th International Conference,
Seville, Spain, September 13–18, 2020, Proceedings, Springer, 2020, pp. 505–522.
[5] M. Reichert, B. Weber, Enabling flexibility in process-aware information systems: chal-
lenges, methods, technologies, volume 54, Springer, 2012.
[6] J. Becker, K. Bergener, O. Mueller, F. Mueller-Wienbergen, Documentation of flexible
business processes-a healthcare case study, AMCIS 2009 Proceedings (2009).
[7] J. Sànchez-Ferreres, A. Burattin, J. Carmona, M. Montali, L. Padró, Formal reasoning
on natural language descriptions of processes, in: Business Process Management: 17th
International Conference, Vienna, Austria, Springer, 2019, pp. 86–101.
[8] R. J. C. Bose, R. S. Mans, W. M. van der Aalst, Wanna improve process mining results?, in:
Symp. on computational intelligence and data mining (CIDM), IEEE, 2013, pp. 127–134.
[9] S. J. van Zelst, F. Mannhardt, M. de Leoni, A. Koschmider, Event abstraction in process
mining: literature review and taxonomy, Granular Computing 6 (2021) 719–736.
[10] R. Seiger, F. Zerbato, A. Burattin, L. García-Bañuelos, B. Weber, Towards iot-driven process
event log generation for conformance checking in smart factories, in: 2020 IEEE 24th Int.
Enterprise Distributed Object Computing Workshop (EDOCW), 2020, pp. 20–26.
[11] J. Mangler, J. Grüger, L. Malburg, M. Ehrendorfer, Y. Bertrand, J.-V. Benzin, S. Rinderle-Ma,
E. Serral Asensio, R. Bergmann, Datastream xes extension: Embedding iot sensor data into
extensible event stream logs, Future Internet 15 (2023).
[12] N. Herzberg, Integrating Events Into Non-automated Business Process Environments:
Enabling Transparency, Traceability, and Optimization for Business Processes Driven by
Humans, Ph.D. thesis, Universität Potsdam, 2018.
[13] A. Lanz, M. Reichert, B. Weber, Process time patterns: A formal foundation, Information
Systems 57 (2016) 38–68.
[14] A. Burattin, G. Guizzardi, F. M. Maggi, M. Montali, Fifty shades of green: How informative
is a compliant process trace?, in: Advanced Information Systems Engineering: 31st Int.
Conf., CAiSE, Rome, Italy, June 3–7, 2019, Proceedings 31, Springer, 2019, pp. 611–626.
[15] R. Seiger, L. Malburg, B. Weber, R. Bergmann, Integrating process management and
event processing in smart factories: A systems architecture and use cases, Journal of
Manufacturing Systems 63 (2022) 575–592.
[16] K. Traganos, P. Grefen, I. Vanderfeesten, J. Erasmus, G. Boultadakis, P. Bouklis, The
horse framework: A reference architecture for cyber-physical systems in hybrid smart
manufacturing, Journal of Manufacturing Systems 61 (2021) 461–494.
[17] I. Beerepoot, C. Di Ciccio, H. A. Reijers, S. Rinderle-Ma, W. Bandara, et al., The biggest
business process management problems to solve before we die, Computers in Industry
146 (2023) 103837.
[18] F. Zerbato, R. Seiger, G. Di Federico, A. Burattin, B. Weber, Granularity in process mining:
Can we fix it?, in: CEUR Workshop Proceedings, volume 2938, 2021, pp. 40–44.
[19] L. Malburg, R. Seiger, R. Bergmann, B. Weber, Using physical factory simulation models
for business process management research, in: Business Process Management Workshops,
Springer, 2020, pp. 95–107.
[20] S. Rinderle-Ma, J. Mangler, Process automation and process mining in manufacturing, in:
Business Process Management, Springer International Publishing, Cham, 2021, pp. 3–14.
[21] L. Pufahl, F. Zerbato, B. Weber, I. Weber, Bpmn in healthcare: Challenges and best practices,
Information Systems 107 (2022) 102013.
[22] C. Combi, B. Oliboni, G. Pozzi, F. Zerbato, et al., Models and architectures for the enactment
of healthcare processes, in: Process Modeling and Management for Healthcare, CRC Press,
2017, pp. 3–33.
[23] W. P. Safety, W. H. Organization, et al., WHO guidelines on hand hygiene in health care,
Technical Report, World Health Organization, 2009.
[24] W. H. Organization, Who guidelines on drawing blood: best practices in phlebotomy, 2010.
[25] F. Weyers, R. Seiger, B. Weber, Method to identify process activities by visualizing sensor
events, in: Business Process Management Workshops, Springer, 2022.
[26] R. Seiger, M. Franceschetti, B. Weber, An interactive method for detection of process
activity executions from iot data, Future Internet 15 (2023).
[27] T. T. Hildebrandt, R. R. Mukkamala, Declarative event-based workflow as distributed
dynamic condition response graphs, arXiv preprint arXiv:1110.4161 (2011).