Modeling, Executing and Monitoring IoT-Driven Business Rules in BPMN and DMN: Current Support and Challenges [original]

Modeling, Executing and Monitoring IoT-Driven

Business Rules with BPMN and DMN: Current

Support and Challenges

Yusuf Kirikkayis, Florian Gallik, and Manfred Reichert

Institute of Databases and Information Systems, Ulm University, Germany

{yusuf.kirikkayis,florian-1.gallik, manfred.reichert}@uni-ulm.de

Abstract. The involvement of the Internet of Things (IoT) in Business

Process Management (BPM) solutions is continuously increasing. While

BPM enables the modeling, implementation, execution, monitoring, and

analysis of business processes, IoT fosters the collection and exchange

of data over the Internet. By enriching BPM solutions with real-world

IoT data both process automation and process monitoring can be im-

proved. Furthermore, IoT data can be utilized during process execution

to realize IoT-driven business rules that consider the state of the phys-

ical environment. The aggregation of low-level IoT data into process-

relevant, high-level IoT data is a paramount step towards IoT-driven

business processes and business rules respectively. In this context, Busi-

ness Process Modeling and Notation (BPMN) and Decision Model and

Notation (DMN) provide support to model, execute, and monitor IoT-

driven business rules, but some challenges remain. This paper derives

the challenges that emerge when modeling, executing, and monitoring

IoT-driven business rules using BPMN 2.0 and DMN standards.

Keywords: IoT ·BPM ·BPMN ·DMN ·Business Rules ·Challenges

1 Introduction

As electronic components have become smaller, less expensive, and more pow-

erful, the Internet of Things (IoT) has received an upswing [3]. Many embedded

components are equipped with software, sensors, actuators, and network connec-

tivity that enable the collection and exchange of data (sensors) as well as physical

responses to events (actuators) [2]. Such physical objects can be embedded in

everyday devices such as smartphones, wearable devices, washing machines, or

refrigerators. They can be further found in large systems such as, smart cities,

logistics or healthcare [4]. In general IoT refers to a network of physical objects

populated by sensors and actuators that communicate and exchange data over

the Internet [5]. While sensors are used to collect data about the real-world

(e.g., temperature sensor, humidity sensor, heart rate sensor, or camera sensor),

actuators are used control the physical world (e.g., watering systems, security

systems, or air conditioner) [6]. Such interconnected IoT devices enable capturing

the dynamic context of the physical world into the digital world.

2 Y. Kirikkayis et al.

While IoT enables exchanging and collecting data about the physical world

over the Internet, BPM enables modeling, implementing, executing, monitoring,

and analyzing business processes [7]. By enhancing business processes with IoT

capabilities, process execution and monitoring as well as decision making can be

enhanced. Furthermore, a more comprehensive view becomes possible for such

IoT-aware business processes. Besides sensing the physical world, physical tasks

such as moving a robot, as well as digital tasks, such as notifying a system, can be

automated based on IoT devices [1]. By integrating the physical world as a key

perspective in business processes, contextual information that was previously in-

visibly embedded in various environments can be continuously and automatically

captured by IoT devices. IoT-aware business processes understand the dynamic

context of the physical world, which makes them context-aware as well [8].

IoT has the ability to continuously and automatically support IoT-aware

business processes with real-world IoT sensor data in real-time. IoT-driven de-

cisions in business processes expose a need for context aggregation, context-

awareness, and up-to-date (i.e. real-time) data, which are the key data source

for dynamic decision making [9] [8]. To address this need, IoT sensor data collec-

tion should proceed as follows (I) sensing low-level data from the real-world (e.g.,

temperature, switch state, humidity, brightness), (II) combining low-level data

and aggregating them into high-level information, and (III) enabling decision

making based on the obtained information [10]. This means, low-level data

are captured in the physical world and need to be aggregated and combined to

process-relevant high-level data [11] [10]. Data from traditional repositories

such as databases and data warehouses are not sufficient for IoT-aware decision

making [1]. Decisions in IoT-aware business processes require up-to-date data

about the physical environment [10]. For example, when using IoT devices such

as temperature sensor, humidity sensor, and brightness sensor, the condition of

the goods in a truck can be checked. By aggregating these low-level IoT data

and combining them, decisions can be made in the course of a business process.

Related to the example, the temperature and humidity value can be combined. If

the maximum temperature and humidity are exceeded, the decision start cooling

system can be made. We refer to this type of conditions business rules.

The integration of IoT in BPM has gained significant attention in litera-

ture, in particular several BPMN extensions and notations [19,33–35] have been

proposed to integrate IoT in business processes in terms of resources. Conse-

quently, IoT data is directly used without aggregating and combining it with

other contextual process data. As a result, the possibility of generating high-level

information is not exploited, which impairs the potential capability. In addition,

decisions are traditionally hard-coded into business processes, which affects the

ability to make dynamic decisions [8]. Current approaches mostly focus only on

the integration of IoT into business processes to visually represent IoT involve-

ment. The modeling, execution, and monitoring of IoT-driven business rules is

neglected. Moreover, the integration of IoT and BPM is constrained due to the

lack of a methodological framework for connecting the IoT infrastructure with

the BPM system [8]. For modeling business rules, in turn, the Decision Model

Modeling, Executing, and Monitoring IoT-driven business rules 3

and Notation (DMN) [12] standard can be used in combination with BPMN.

By using DMN, the decision logic can be separated from the process logic. Fur-

thermore, DMN enables the aggregation of low-level data into high-level one.

However, DMN does not provide official support for modeling IoT-driven busi-

ness rules, which creates new challenges. In this paper, we derive and highlight

research challenges that need to be tackled in order to properly model, execute,

and monitor IoT-driven business rules.

This paper is structured as follows: Section 2 illustrates the support for mod-

eling, executing, and monitoring IoT-driven business rules based on either BPMN

or BPMN plus DMN. In Section 3 we derive challenges that need to be tackled

when modeling, executing, and monitoring IoT-driven business rules. Finally, in

Section 4 we summarize and discuss the results.

2 Current support for IoT-driven business rules in

BPMN and BPMN + DMN

2.1 IoT-driven business rules in BPMN

BPMN 2.0 is a standardized graphical process modeling language that provides

elements for modeling business processes and workflows [13]. However, BPMN

2.0 does not provide official support for modeling IoT involvement and capabili-

ties, but provides different possibilities that can be used for representing IoT such

as (i) tasks, (ii) events, and (iii) resources [14]. For IoT-driven business rules,

different gateways may be used in the current BPMN 2.0 standard, these can be

divided into the following categories (I) exclusive, (II) inclusive, (III) paral-

lel, (IV) complex, and (V) event-based [13]. Example 1 describes a business

process with IoT-driven business rules. Note that the IoT-driven business rules

are modeled exclusively with standard BPMN 2.0 elements.

Example 1: Consider a medical system that monitors the health status of a

patient who has been diagnosed with Chronic Obstructive Pulmonary Disease

(COPD). COPD is a disease in which the lungs are permanently damaged and

the airways (bronchi) are restricted. At anytime, the patient may experience un-

pleasant complications such as shortness of breath on exertion, coughing, sounds

when breathing, fast heart rate, hyperactive muscle use, increased blood pressure,

and a cold skin. Several studies [15] [16] have shown that the IoT-driven monitor-

ing of sensor-equipped patients can improve their quality of life by identifying the

severity of COPD disease and responding accordingly. In order to detect COPD,

all required sensors are polled (cf. Figure 1). Based on the values provided by

the IoT sensors and the defined IoT-driven business rule, either no treatment,

treatment with an oxygen mask, or treatment with an inhaler is administered.

4 Y. Kirikkayis et al.

EMG = Electromyography

(Muscle activity)

respiration

Query EMG

sensor

Query skin

temperature

sensor

Query

respiratory

sensor

Query blood

pressure sensor

Skin temp.

fast

normal

Administer

oxygen mask

Administer

inhaler

Diastolic

blood

pressure

irregular

≥120 fast

Muscle

activity

hyper

normal

<120 normal

normal

Heart

rhythm

fast

Heart

rhythm

Administer

oxygen mask

Administer

inhaler

Muscle

activity

hyper

Heart

rhythm

cold

normal

fast

Heart

rhythm Muscle

activity

fast normal

normal

hyper

Fig. 1. Example of an IoT-driven business rules expressed in terms of BPMN 2.0

2.2 IoT-driven business rules in BPMN + DMN

The combined use of BPMN [13] and DMN [12] has already been studied in [17]

and [18]. The interplay between process and decision logic plays a crucial role

for business processes, as business rules are evaluated during process execution

and may affect process outcomes [25]. DMN is a decision modeling standard

that consists of two levels: The first one represents the decision requirements,

where the dependencies between the elements involved in the decision model

are captured [8]. The decision requirements are represented by DRDs (Decision

Requirements Diagrams) and form the dependencies between the data and sub-

decisions. The input data for DRDs may be static or dynamic. The second level

is the decision logic, which is usually modeled in terms of decision tables [8] [5].

To construct a DMN model, low-level data needs to be aggregated to higher-level

one and enables consequently to aggregate contextual data [5] [12]. Example 2

describes a decision-aware COPD process (cf. Figure 2). Note that the business

process is modeled in terms of BPMN 2.0 and the business rules in terms of

DMN using the elements provided by the two standards.

Example 2: To identify the severity of COPD, the patient is equipped with sev-

eral sensors. The severity of COPD is determined based on the defined business

rules (cf. Figure 2 [6]), the data values provided by sensors in real-time (cf. Fig-

ure 2 [2]), and data from a database (cf. Figure 2 [3]) in DMN. The decision

in DMN becomes evaluated when activating the Check COPD severeness busi-

ness rule task in BPMN 2.0 (cf. Figure 2 [1]). After deciding whether treatment

with oxygen mask or inhaler, or no treatment becomes necessary, the heart status

is checked based on real-time sensor data (cf. Figure 2 [5]). Depending on the

result, the patient is either not treated or treated with a defibrillator.

Modeling, Executing, and Monitoring IoT-driven business rules 5

severeness

Administer

oxygen mask

Administer

inhaler

severe

attach

mild

attack

none Check heart

status

heart

status

Treat patient

with defibrillator

irregular

regular

Equip patient

with sensors

Check COPD

severeness

Skin temperature

Sensor

[2]

Business Process Model and

Notation (BPMN)

COPD

severeness

Heart status

Respiration Muscle activity

Respiration dataSkin sensor EMG data Blood pressure

sensor

Final result

Data

EMG sensor

Intermediate

result

Decision Model and Notation

(DMN)

[1]

[3] Data[3] Sensor

[2] Sensor

[2]

[4]

[5]

Input 1 Output

Blood_Pressure

Input 2

EMG

Heart status

< 80

"normal"

"regular

85 - 89

"midline"

"regular"

85 - 89

"myotendon"

"regular"

90 - 99

"midline"

"regular"

90 - 99

"innervation"

"irregular"

Decision table

[6]

Fig. 2. Relationship between BPMN 2.0 and DMN

3 Challenges

Although the BPMN 2.0 and DMN allow expressing certain aspects of IoT-driven

business rules, several challenges remain [5] [14]. We studied literature and IoT-

driven business rules from different domains modeled in terms of either BPMN or

BPMN + DMN. Moreover, we were able to identify additional challenges in the

context of the two IoT-related projects BPMN Extension for IoT (BPMNE4IoT)

[2] and IoT Decision Making for BPMN (IoTDM4BPMN) [10] we are involved in.

We discuss the derived challenges and divide them along the modeling, execution,

and monitoring of IoT-driven business rules. The following structure is applied

for each challenge; we briefly describe the challenge, provide an example, and

reference relevant literature that tries to address the research gaps described in

the challenge. A summary of the challenges can be found in Table 1.

Table 1. Challenges for IoT-driven business rules in BPMN and BPMN + DMN.

Modeling Challenges

C1 - Modeling IoT-driven business rules in BPMN 2.0

C2 - Modeling IoT-driven business rules with BPMN 2.0 + DMN

C3 - Reducing the complexity of IoT-driven business rules

Execution Challenges

C4 - Extending process log with IoT data

C5 - Treatment of IoT data outliers

C6 - Treatment of defective IoT devices

Monitoring Challenges

C7 - Traceability of IoT-driven business rules

C8 - Fault monitoring in IoT-driven business rules

C9 - Real-time monitoring of IoT-driven business rules

6 Y. Kirikkayis et al.

3.1 Modeling Challenges

C1 - Modeling IoT-driven business rules in BPMN 2.0

Description: In Section 2.1, we discussed the current support of the BPMN 2.0

standard for incorporating IoT devices and the modeling of IoT-driven business

rules. In order to express IoT involvement within business rules, it should be

possible to model the involved IoT devices. Note that the returned data values

of the IoT devices are used as basis for evaluating the IoT-driven business rule.

Therefore, it is crucial to be able to properly capture IoT involvement.

Example: To describe the problem, Figure 3 illustrates the treatment of COPD.

When modeling the business process and the corresponding IoT-driven business

rules with the standard BPMN 2.0 elements, it remains unclear which tasks

are IoT-related and which are not. Furthermore, it is unclear which sensors are

actually used in the context of IoT-driven business rules (cf. Figure 3). To de-

cide whether no treatment, a treatment with oxygen mask, or a treatment with

inhaler is required, skin temperature (Task 2), respiration (Task 3), and EMG

value (Task 4) are needed, whereas the ECG value (Task 5) is not needed for the

treatment but for the alarm. Note that it is unclear which sensor is important

for which business rule. The entire process model must be carefully read and un-

derstood in order to determine this. Furthermore, it is impossible to distinguish

between sensors (Tasks 2-5) and actuators (6). In order to distinguish between

sensors and actuators, the labels should reflect the involvement of IoT and the

modeler needs to be familiar with IoT devices and their behavior. As another

drawback, no visual difference between IoT-aware service tasks (Tasks 2-6) and

BPMN service task (Tasks 1,9, 10) exists. Moreover, the complexity and thus,

the comprehensibility of the business rules increases with growing number of

involved IoT devices. This aggravates reading and understanding of the process

models as well as the IoT-driven business rules. With increasing number of busi-

ness rules and increasing complexity of the rule logic, the flexibility, scalability,

and maintainability of the resulting process model and IoT-driven business rules

is impaired. The complex nesting and ambiguous involvement of IoT makes any

later extensions or changes difficult. As IoT-driven business rules are hard-coded

in the business process in form of gateways, aggregation and combination of IoT

low-level data into high-level one is not appropriate with BPMN 2.0. Obviously,

the IoT-driven businesses rules cannot be reused in a different context. When

using BPMN, both process and decision logic are defined in one and the same

process model. As a result, the modeled logic is hard-coded and constrained to

a local location. Therefore, reusability is impaired [5].

Possible solution: There exist several works [2,19,33, 34] that introduce exten-

sions for representing IoT devices in the context of BPMN 2.0. These extensions

enable the explicit modeling of IoT participation by introducing IoT specific el-

ements. They propose a visual discrimination between regular BPMN elements

and IoT elements in the modeling phase [14]. In [36], two approaches for mod-

eling IoT-driven business rules are presented. The first one extends the BPMN

Modeling, Executing, and Monitoring IoT-driven business rules 7

2.0 standard by providing specific IoT decision modeling elements. Note that

BPMN 2.0 is a rather complex language and any extension constitutes a de-

viation from the standard [26] [14]. By extending BPMN with additional IoT

elements complexity might increase. In turn, this might effect model compre-

hensibility. The second approach proposes an IoT-specific drag&drop modeler,

which separates the business rules from the process logic. As the drag&drop

modeler outsources the business rules from the BPMN process model, the struc-

ture of the IoT-driven business rule cannot be viewed in BPMN. This makes it

difficult to extend, maintain, and troubleshoot the IoT-driven business rules.

Load patient file

Patient

files

Query EMG

sensor

Query

respiratory

sensor

Query skin

Sensor

respiration

skin

temp

heart

rhythm

Administer

oxygen mask

muscle

activity

Administer

inhaler

heart

rhythm

muscle

Activity

Update patient

record

Reset

monitoring

system

Read ECG

sensor

heart

rate

Sound

emergency

alarm

fast

cold

fast

hyper

normal

fast normal

hyper

normal

else

6 7

910

Fig. 3. IoT awareness in BPMN-based process model (adopted from [2]).

C2 - Modeling IoT-driven business rules with BPMN 2.0 + DMN

Description: Combining BPMN and DMN can solve some of the problems and

gaps mentioned above. For example, DMN is suited for aggregating and combin-

ing business rules as it uses appropriate techniques such as decision tables. Since

DMN does not provide any explicit elements for modeling IoT, the modeling of

business rules based on IoT data constitutes a challenge.

Example: Consider Figure 4. The results of the queried IoT sensors are stored

in data objects which then flow into the business rule task Check COPD severe-

ness (cf. Figure 4). Representing the received IoT data as data objects increases

the complexity and the number of modeling elements in the business process.

If IoT data is not represented in terms of data objects, such as in the process

model depicted in Figure 3, it will be unclear which IoT data actually concern

the business rules in DMN. This, in turn, affects model readability and compre-

hensibility. In addition, it is impossible to distinguish between IoT data objects

on the one hand and BPMN data objects on the other. Note that the Check

COPD severeness business rule task represents the decision modeled and exe-

cuted in DMN. As DMN outsources the decision logic from the BPMN process

model, the structure of the IoT-driven business rule cannot be directly viewed

in BPMN. Typically, the business rule task only provides the final decisions. In

addition, DMN does not officially support the modeling of IoT-driven business

rules. Consequently, it cannot be distinguished between IoT input data (cf. Fig-

ure 3 [2]) and, for example, input data from a database (cf. Figure 3 [3]). When

8 Y. Kirikkayis et al.

using DMN, decision logic is captured in decision tables (cf. Figure 3 [4]). With

increasing number of IoT devices, however, the complexity of the decision table

increases as well. Accordingly, the error detection becomes more difficult.

Possible solution: Several authors have argued that DMN is capable of modeling

IoT-driven business rules [5] [50]. For example, [5] shows how DMN elements can

be used to model different IoT-driven business rules (e.g., smart transportation,

smart ventilation, and smart healthcare). Thus, no discrimination between reg-

ular DMN elements (e.g. input data) and IoT-related DMN elements is present.

One possible solution to close this gap would be to extend the DMN standard

with IoT decision elements.

Load patient file

Query skin

sensor

Query

respiratory

sensor

Query EMG

sensor

Administer

oxygen mask

Administer

inhaler

Update patient

record

Reset

monitoring

system

Check COPD

severeness

Sound

emergency

alarm

Read ECG

sensor

EMG

value

Patient

files

severeness

none

severe

attach

mild

attack

Skin

temperature

Respiratory

value

heart

rate

else

ECG

Skin

temperature

Sensor

[2]

COPD

severeness

Heart status

Respiration Muscle activity

Respiration dataSkin sensor EMG data Blood pressure

sensor

Final result

Data

EMG sensor

Decision Model and Notation

(DMN)

[3] Data

[3] Sensor

[2]

Sensor

[2]

[1]

Input 1 Output

Skin temp

Input 2

Respiratory

Heart status

"cold"

"fast"

"severe"

"cold"

"fast"

"severe"

"normal"

"fast"

"severe"

"normal", "cold"

"fast"

"mild"

"normal"

"none"

Decision table

[4]

Input 3

EMG

"normal", "hyper"

"hyper"

"normal"

"normal", "hyper"

Fig. 4. Using BPMN 2.0 and DMN for modeling an IoT-driven business rule.

C3 - Reducing the complexity of IoT-driven business rules

Description: As discussed in the context of C1 and C2, the complexity of both

the process and the decision model increases when involving IoT devices. Mod-

eling IoT-awareness for a process and decision model is a complex undertaking,

and the resulting model often turns out to be difficult to understand due to the

potentially ambiguous use of modeling elements [2]. IoT-driven business rules

become more complex when modeling them with BPMN 2.0 as the involvement

of IoT is not supported by the standard. In turn, this has a negative repercussion

on modeling IoT-aware processes and IoT-driven business rules.

Example: As processes running in an IoT setting are often data- and decision-

intensive, the modeled process might be too extensive and, thus, too complex

Modeling, Executing, and Monitoring IoT-driven business rules 9

to be understandable. When modeling IoT-driven business rules in BPMN (cf.

Figure 3), the number of gateways and control flow paths grows as additional

business rules are introduced. This leads to a large number of branch conditions

and control flow elements, resulting in a complex structure [14]. When combining

BPMN and DMN we can encapsulate this complexity by defining the business

rules in decision tables. This significantly reduces the number of gateways and

sequence flows on one hand. On the other, the business rules are hidden in

decision tables. At the BPMN level, it is impossible to see how the IoT-driven

business rules are defined and how they depend on each other. In the following,

we consider metrics proposed in literature to evaluate the complexity of the

models described in the previous sections [27] [14]. The metrics are used for

identical processes with different modeling approaches. The BPMN metrics NOA

(number of activities), NOG (number of gateways), and NOF (number of flows)

were defined in [27]. The metrics for DMN are taken from [28] and consist of

TNR (total number of rules), NOD (number of decisions), and TNDO (total

number of data objects). Table 2 shows that the complexity of modeling IoT-

driven business rules in BPMN is larger compared to BPMN + DMN due to

the higher number of sequence flows (NOF) and gateways (NOG). As opposed

to BPMN, the complexity in BPMN + DMN is shifted to the decision tables

(TNR).

Table 2. Evaluation of complexity through the application of metrics

Case NOA NOG NOF TNR NOD TNDO

BPMN 10 11 31 - - -

BPMN+DMN 11 6 23 17 5 5

Possible solution: A possible solution for reducing complexity related to IoT-

aware processes is presented in [2]. The authors introduce new modeling ele-

ments that, for example, merge individual sensor artifacts into one sensor group

artifact in order to increase the abstraction level. Another possible solution is the

definition of guidelines for IoT-driven business rules. For example, [38] proposes

seven process modeling guidelines (7PMG). However, these do not consider the

modeling of IoT-driven business rules. Another approach is the definition of pat-

terns for modeling IoT behavior. These patterns could, for example, reduce the

number of message flows between the central pool and the IoT-aware pools by

using the computing capacities of the IoT devices [51].

3.2 Execution Challenges

C4 - Extending process log with IoT data

Description: The sensors used in a business process record the physical world

and transform it into the digital world. The data generated by IoT devices allows

10 Y. Kirikkayis et al.

for the continuous monitoring and provision of opportunities for analysing and

optimizing the performed processes, e.g., through process mining or real-time

monitoring [42] [43]. Furthermore, as a data source IoT can improve the verifi-

cation of the conformance between the actual execution of a business process in

the physical world and its execution as recorded by the Business Process Man-

agement System (BPMS) based on a secondary log of sensor data [1]. Through

the use of common business process engines (e.g. Camunda [20]), which are un-

aware of IoT involvement, extending the process log with IoT data is difficult

and complex. As standard BPMN elements are used for modeling the IoT-driven

business rules, the business process engines is unaware of the involvement of IoT,

whereby the event log cannot be extended by the engine with IoT data.

Example: Process mining has become an important research area in Computer

Science, which aims to extract knowledge from event logs to discover, monitor,

and improve business processes [44] [45]. To allow for a finer grained discovery,

monitoring, and improvement of IoT-aware business rules, the event log needs

to be extended with IoT-related data collected from smart objects. Extending

the event log to include IoT data requires an IoT-aware process engine and a

suitable architecture. Most IoT infrastructures are based on isolated IoT devices

and integrated with applications that are not necessarily process-aware. Fur-

thermore, such applications are often based on proprietary control software with

non-standard interfaces [43].

Possible solution: A possible solution to enhance a business rule log with IoT

data is to use or develop an IoT-aware business rule engine as well as an embed-

ding architecture. The IoT-aware engine should be able to detect IoT actions

and record them in the log. Another way to extend the log with IoT data is to

capture the IoT actions in a separate event log. Then the IoT event log may be

merged with the process event log.

C5 - Treatment of IoT data outliers

Description: Outliers constitute irregularities or behavioral deviations of the IoT

devices and the delivered IoT data. IoT sensors are responsible for capturing,

collecting, and transmitting data. The data collected from the physical environ-

ment, however, might be prone to outliers [29]. The treatment of outliers is very

important in relation to IoT-driven business rules to avoid erroneous decisions

being made based on the faulty sensor data.

Example: The IoT is used in a wide variety of environments and scenarios, e.g.

environmental monitoring, smart cities, disaster warning, and agriculture. In this

context, sensors are often installed in harsh environments. As a consequence, the

sensors are susceptible to malfunction, rapid wear, and tampering. This, in turn,

can lead to outliers [29]. In Table 3, fault categorizations in IoT implementations

are mentioned and summarized [24].

Modeling, Executing, and Monitoring IoT-driven business rules 11

Table 3. IoT fault categorizations [24].

Fault Definition

Short An IoT data point deviates significantly from the expected temporal or spa-

tial trend of the data.

Stuck-at A series of data points has zero or almost zero variation for a period of time

greater than expected.

Noise Sensor data exhibiting an unexpectedly high amount of variation .

Illustrating example: Consider an environmental monitoring station that con-

sists of temperature sensors, humidity sensors, and brightness sensors. If the

brightness and temperature sensors are manipulated such that they are directly

hit by the sun, the IoT sensor will provide distorted values. This, in turn, leads

to the faulty execution of corresponding IoT-driven business rule.

Possible solution: There are different approaches [46–48] and techniques (e.g.

machine learning) for detecting and treating of outliers. Most approaches, how-

ever, do not equip their IoT-aware business rule engine or architecture with the

techniques for handling outliers. One possible solution is to equip the architecture

with middleware that detects the IoT outliers and handles them accordingly.

C6 - Treatment of defective IoT devices

Description: Handling defective IoT devices constitutes another major challenge

to be tackled in the context of IoT-driven business rules. As discussed in C5,

IoT-driven business rules need real-time processing. In this context, IoT sensors

must provide a result within an acceptable amount of time. The challenge is to

detect and handle defective or non-reachable IoT sensors.

Example: As IoT devices constitute electronic components, they might suddenly

fail and then no longer function [30]. Such an IoT device failure result in the

evaluation of IoT-driven business rules with missing measured values or zero val-

ues. Note that this might lead to several runtime issues in the business process

relying on these rules. In literature, such failures are referred to as binary fail-

ures [31]. The use of IoT devices in harsh environments and limited computing

capacity can lead to failures as well. Other reasons include limited battery life,

hardware failures, and human mistakes [32]. The failure of IoT devices involved

in IoT-driven business rules might have dire consequences. For example, if a

queried IoT sensor is not accessible, no decision can be made, which can lead

to deadlocks as the workflow engine continuously checks whether the condition

is met or not. Another possible sequence would be the execution of an incorrect

business rule. If an IoT sensor suffers from a binary failure, i.e., it returns a zero

value by default or the occurrence of an error, it might result in an incorrect

business rule executed by mistake. Assume, for example, that the business rule

”temperature sensor < 25◦C“ returns a zero value if the referred temperature

sensor is defective; the condition would still be met.

12 Y. Kirikkayis et al.

Possible solution: Several techniques [30] [31] exist for detecting of defective IoT

devices. However, there is no specific approach that deals with such failures in

the context of IoT-driven business rules. One approach is to use a middleware

that enables fault tolerance based on redundant IoT devices or the replacement

of faulty IoT data by accessing historical records, depending on the duration of

the outage. Another solution is to introduce a prioritization mechanism for IoT

devices. The process modeler can assign priority levels for IoT devices. Depend-

ing on the priority level of the defective IoT device, the process execution may

be aborted or the IoT devices be ignored for decision making.

3.3 Monitoring Challenges

C7 - Traceability of IoT-driven business rules

Description: When monitoring IoT-driven business rules (during both process

and business rule execution) traceability constitutes a fundamental challenge.

Traceability refers to the understanding of the decision resulting from the evalu-

ation of an IoT-driven business rule. To understand what triggered an IoT-driven

business rule, it is crucial to comprehend which IoT devices were queried how

and why. A monitoring approach should address this challenge and present both

the modeled and the executed IoT-driven business rules in a structured and

understandable way.

Example: With increasing context-intensity of the environment in which the IoT-

driven business rule is performed, the number of monitoring challenges increases.

Moreover, an IoT sensor may be used by multiple business rules. Furthermore,

the results of one IoT-driven business rule can be utilized by other rules, resulting

in a complex nesting. Therefore, the traceability of IoT-driven business rules

must be monitored in an appropriate manner during execution.

Possible solution: [10] introduced a BPMN 2.0 extension that represent business

rules graphically in combination with a truth table. This approach further uses

overlays and color highlighting to visualize the result of a business rule execu-

tion. Although it is possible to determine how the decision was reached, it is

not possible to reconstruct the exact decision-making process of a business rule

in retrospect, as detailed temporal information is missing, especially for time-

critical sensor queries. One possible solution is to time-stamp the incoming data

of the queried IoT devices and each evaluation of a business rule, and to visualize

it accordingly in the process.

C8 - Fault monitoring in IoT-driven business rules

Description: Monitoring errors constitute another challenge. If IoT devices are

used in business rules without receiving any feedback on faults, the IoT-driven

business rule process might suffer. Detecting and monitoring IoT devices involved

in business rules provides the visibility needed to understand exactly what went

wrong and where it went wrong, to subsequently ensure that this error is avoided

in the future.

Modeling, Executing, and Monitoring IoT-driven business rules 13

Example: IoT sensors generate large amounts of data and operate automati-

cally and continuously. In order to ensure that sensors properly work in the

context of IoT-driven business rules, a precise monitoring system is needed to

check the behavior and performance of the IoT sensors. As a key monitoring

challenge faulty sensors need to be detected during runtime. Another challenge

is to detect and monitor anomalies and outliers. Moreover, the monitoring of IoT

devices supports the confidence of the data collected by the sensor and, thus, the

quality of the business rules. The higher the confidence of the received IoT data

is, the better and more precise the resulting IoT-driven business rule will be.

For example, if four temperature sensors are used in a smart home to calculate

the average temperature, but only two sensors provide a value as the other two

are defective, the confidence of the data is compromised. As a consequence, the

resulting output of the IoT-driven business rule will be not accurate and pos-

sibly an incorrect action be performed. Without a monitoring system detecting

defective IoT devices, the discovery of such scenarios would be not possible.

Possible solution: [10] uses color highlighting (e.g., green, orange, and red) of

IoT devices (sensor artifact) involved in the business rule in case an error such

as a timeout occurs while polling sensor data. In addition, the corresponding

error message is displayed in the execution log. However, the detection is limited

to communication errors and errors in the source code. A possible solution is to

realize a component that not only detects defective IoT devices, but outliers and

insufficient data quality as well. Such a component can mark and visualize the

respective errors and provide additional information about the kind of error.

C9 - Real-time monitoring of IoT-driven business rules

Description: Real-time monitoring of IoT-driven business rules enables contin-

uously updated information streamed with low latency. In turn, the continuous

streaming of up-to-date IoT data allows for the immediate detection of prob-

lems, i.e., based on the real-time monitoring of IoT-driven business rules alerts

can be forwarded more quickly to systems for mitigation in the event of a failure

of IoT devices.

Example: Monitoring IoT-driven business rules and relevant IoT devices in real-

time constitutes another challenge. In particular, this monitoring shall provide

additional information about the current processing state of the respective rule.

For example, it needs to be monitored which IoT-driven business rules are cur-

rently running, in what state they are (e.g., are there faulty IoT devices?), and

what will happen next. For the real-time monitoring of IoT-driven business rules,

the monitoring system needs to communicate with all IoT devices involved in

process and rule execution. Furthermore, the monitoring system should be ex-

tensible and scalable to be able to add IoT sensors and IoT-driven business rules

on the fly.

14 Y. Kirikkayis et al.

Possible solution: Several works [10, 40, 41, 49] exist for monitoring IoT-driven

business rules after their execution. In context, during execution it is only pos-

sible to monitor the final result of the IoT-driven business rules, but not how

they are composed and which intermediate results (of sub-rule evaluation) exist.

Moreover, it is not indicated in what state the rules are (e.g. ready, executing, or

finish). One possible solution is to develop a suitable architecture that delivers all

existing information about the IoT-driven business rules to the monitoring sys-

tem in real-time. Note that this requires appropriate communication protocols

and a real-time capable monitoring component in the architecture.

4 Conclusion

In this paper, modeling, execution, and monitoring IoT-driven business rules was

examined by either exploring corresponding rules with BPMN 2.0 or BPMN in

combination with DMN. The IoT adds value to BPM through its ability to

transform the physical world to its digital twin. Integrating BPM with IoT ca-

pabilities should exploit the complete potential of IoT and cover all relevant use

cases in this context. Current research related to the combination of IoT and

BPM is concerned with the integration of IoT with BPM as a resource. Less at-

tention is paid to the integration of IoT into BPM for decision making through

IoT-driven business rules. As a result, several challenges exist with respect for

the modeling, execution, and monitoring of IoT-driven business rules. In this

paper, in addition to exploring the current support for IoT-driven business rules

in BPMN and BPMN + DMN, we have derived these challenges through study-

ing literature, real-world IoT-driven business rules, and hands-on experiences in

our IoT-related projects BPMN Extension for IoT (BPMNE4IoT) [2] and IoT

Decision Making for BPMN (IoTDM4BPMN) [10]. Existing solutions were de-

scribed based on a literature review, including a discussion of their strengths and

weaknesses. If no solution was found in literature, a possible solution approach

was discussed.

The identified challenges should be addressed in future with the goal of en-

abling the integration of IoT and BPM for executing of IoT-driven business

rules.

References

1. Janiesch et al.: The Internet of Things Meets Business Process Management: A

Manifesto. In: Systems, Man, and Cybernetics Magazine, (2020)

2. Kirikkayis, Y., Gallik, F. and Reichert, M.: Towards a Comprehensive BPMN Ex-

tension for Modeling IoT-Aware Processes in Business Process Models. In: 16th Int

Conf on Research Challenges in Inf Science (RCIS), (2022).

3. Ashton, K.: That ’Internet of Things’ Thing. In: RFID journal, vol. 22, no. 7, pp.

97–114, (2009).

4. Chang, C., Srirama, S., Buyya, R.: Mobile Cloud Business Process Management

System for the Internet of Things: A Survey. In: ACM Computing Surveys 49(4),

pp. 1–42. (2016).

Modeling, Executing, and Monitoring IoT-driven business rules 15

5. Hasi´c, F., Serral, E. and Snoeck, M.: Comparing BPMN to BPMN + DMN for

IoT process modelling: a case-based inquiry. In 35th ACM/SIGAPP Symposium on

Applied Computing, (2020).

6. Valderas, P., Torres, V. and Serral, E.: Modelling and executing IoT-enhanced busi-

ness processes through BPMN and microservices. In: The Journal of Systems and

Software 184, (2022).

7. Dumas, M., La Rosa, M., Mendling, J., Reijers, H.: Fundamentals of Business Pro-

cess Management 2nd edn. Springer-Verlag GmbH, Germany (2018).

8. Song, R., Vanthienen, J., Cui, W., Wang, Y. and Huang, L.: Context-Aware BPM

Using IoT-Integrated Context Ontologies and IoT-Enhanced Decision Models. In:

Conf on Commerce and Enterprise Computing, (2019).

9. Koschmider, A., Mannhardt, F. and Heuser, T.: On the Contextualization of Event-

Activity Mappings. In: Business Process Management Workshops, (2018).

10. Kirikkayis, Y., Gallik, F. and Reichert, M.: IoTDM4BPMN: A IoT Decision Making

Framework for Business Processes in BPMN. In: Int Conf on Service Science, (2022).

11. Krishnamurthi et al.: An OVerview of IoT Sensor Data Processing, Fusion, and

Analysis Techniques. In: Sensors, (2020).

12. OMG: Decision Model and Notation (DMN) 1.2, (2018).

13. OMG: Business Process Model and Notation (BPMN) 2.0, (2010).

14. Hasi´c, F. and Serral, E.: Executing IoT Processes in BPMN 2.0: Current Support

and Remaining Challenges. In: 13th Int Conf on Service Science, (2019).

15. Zhou et al.: An Internet of Things based COPD managing system: Its development,

challenges and first experiences. In: Clinical eHalth, (2019).

16. Guiling et al.: Clinical guidelines on the application of Internet of Things (IoT)

medical technology in the rehabilitation of chronic obstructive pulmonary disease.

In: Journal of thoracic disease, (2021).

17. Bazhenova et al.: From BPMN process models to DMN decision models. In: Infor-

mation Systems 83, (2019).

18. Combi et al.: Seamless Design of Decision-Intensive Care Pathways. In: Int Conf

on Healthcare Informatics (ICHI), (2016).

19. Yousfi et al.: uBPMN: A bpmn extension for modeling ubiquitous business pro-

cesses. In: Information and Software Technology, (2016).

20. Camunda: Process Engine. In: https://docs.camunda.org/manual/7.8/user-

guide/process-engine/, accessed: 2022-20-04.

21. SAP Signavio: SAP Signavio Process Governance In:

https://documentation.signavio.com/suite/en-us/Content/workflow-

accelerator/userguide/intro.htm, accessed: 2022-20-04.

22. Gruhn et al.: BRIBOT: Towards a Service-Based Methodology for Bridging Busi-

ness Processes and IoT Big Data. In: Service-Oriented Computing: 19th Int Conf

(ICSOC), (2021).

23. Torres et al.: Modeling of IoT devices in Business Processes: A Systematic Mapping

Study. In: Conf on Commerce and Enterprise Computing (CEC), (2020).

24. Chakraborty et al.: Fall-curve: A novel primitive for IoT Fault Detection and Iso-

lation. In: Embedded Networked Sensor Systems, (2018).

25. Bazhenova et al.: From BPMN process models to DMN decision models. In: Infor-

mation Systemms, (2019).

26. Indulska, M., Muehlen, M., Recker, J.: Measuring Method Complexity: The Case

of the Business Process Modeling Notation. (2009).

27. De Oca, IMM. and Snoeck, M.: Pragmatic guidelines for business process modeling,

In: SSRN 2592983, (2014).

16 Y. Kirikkayis et al.

28. Hasic, F. and Vanthienen, J.: Complexity Metrics for DMN Decision Models, In:

Computer Standards and Interfaces, (2019).

29. Gaddam et al.: Detecting Sensor Faults, Anomalies and Outliers in the Internet of

Things: A Survey on the Challenges and Solutions. In: Electronics, (2020).

30. Pachauri, G., and Sharma, S.: Anomaly Detection in Medical Wireless Sensor Net-

works using Machine Learning Algorithms. In: Procedia Computer Science, (2015).

31. Ye, J-. Stevenson, G. and Dobson, S.: Detecting abnormal events on binary sensors

in smart home environments. In: Pervasive Mob. Comput., (2016).

32. Choi et al.: Detecting and Identifying Faulty IoT Devices in Smart Home with

Context Extraction. In: Conf on Depandable Systems and Networks, (2018).

33. Sungur et al.: Extending BPMN for Wireless Sensor Networks. In: Conf on Business

Informatics, (2013).

34. Meyer, S., Ruppe, A., Hilty, L.: The Things of the Internet of Things in BPMN.

In: Conf in Advanced Information Systems Engineering Workshops, (2015).

35. Cheng et al.: Modeling and Deploying IoT-Aware Business Process Applications

in Sensor Networks, (2019).

36. Kirikkayis, Y., Gallik, F. and Reichert, M.: Visual Decision Modeling in IoT-

Aware Processes. In: Central European Workshop on Services and their Compo-

sition, (2020).

37. Polak, P.: BPMN Impact on Process Modeling, (2013).

38. Mending, J., Reijers, H. and van der Aalst, w.: Seven Process Modeling Guidelines

(7PMG), In: Information and Software Technology, (2020).

39. Gallik, F., Kirikkayis, Y. and Reichert, M.: Modeling, Executing and Monitoring

IoT-Aware Processes with BPM Technology. In: Int Conf on Service Science, (2022).

40. R. Song, J. Vanthienen, W. Cui, Y. Wang and L. Huang: Context-Aware BPM

Using IoT-Integrated Context Ontologies and IoT-Enhanced Decision Models. In:

Conf on Commerce and Enterprise Computing, (2019).

41. R. Oliveira et al.: An intelligent model for logistics management based on geofenc-

ing algorithms and RFID technology. In: Expert Systems with Applications, (2015).

42. M. Pegoraro and W. M. P. van der Aalst: Mining uncertain event data in process

mining. In 2019 Int Conf on Process Mining (ICPM), (2019).

43. Seiger et al.: Towards IoT-driven Process Event Log Generation for Conformance

Checking in Smart Factories. In: Int Enterprise Distributed Object Computing

Workshop, (2020).

44. van der Aalst: Process mining: Data Science in Action. Springer, (2016)

45. van der Aalst et al.: Process mining Manifesto, In: Lecture Notes in Business

Information Processing vol 99, (2012).

46. Samara et al.: A Survey of Outlier Detection Techniques in IoT: Review and Clas-

sification. In: Journal of Sensor and Actuator Networks, (2021).

47. Jiang et al.: Outlier Detection Approaches Based on Machine Learning in the

Internet-of-Things. In: Wireless Communications, (2020).

48. Bhatti et al.: Outlier detection in indoor localization and Internet of Things (IoT)

using machine learning. In: Journal of Communications and Networks, (2020).

49. Celestrini et al.: An Architecture and Its Tools for Integrating IoT and BPMN in

Agriculture Scenarios. In: Symposium on applied computing, (2019).

50. Song et al.: A DMN-Based Method for Context-Aware Business Process Modeling

Towards Process Variability. In: Business Information Systems, (2019).

51. Domingos et al.: Automatic Decomposition of IoT Aware Business Processes - a

Pattern Approach. In: Int Conf on ENTERprice Information Systems (CENTERIS),

(2019).