THE ADIWA PROJECT
On the Way to Just-in-Time Process Dynamics based on Events
from the Internet of Things
Markus Schief, Christian Kuhn, Birgit Zimmermann, Philipp R
¨
osch
SAP Research, Bleichstraße 8, Darmstadt, Germany
Walter Waterfeld, Jens Schimmelpfennig, Dirk Mayer
Software AG, Darmstadt, Germany
Heiko Maus
German Research Center for AI (DFKI GmbH), Kaiserslautern, Germany
J
¨
orn Eichler
Fraunhofer SIT, Darmstadt, Germany
Keywords:
Event-based system, Event processing, Context awareness, Dynamic business processes, Internet of things.
Abstract:
In this paper, we introduce a concept, which focuses on innovative commercial system implementations re-
flecting process-embedded events from the Internet of Things. The developed concepts are derived from
experiences applying recent research advances to industry scenarios. The rationale behind the overall con-
cept is twofold: while transparency is increased by event-based methodologies in the context of the Internet
of Things, the agility of business processes is fostered by enhanced business process models, orchestration
support, execution control, and user assistance.
1 INTRODUCTION
In recent years, the speed of change in business envi-
ronments has been accelerating. To enable agile busi-
ness processes, transparency is a crucial prerequisite.
While a great deal of research has been performed in
the field of dynamic business processes, a lack of ac-
ceptance can be constituted as, up to now, the data
basis has not been sufficient (Pesic and van der Aalst,
2006). Our concept enables dynamic design, plan-
ning, and execution of business processes based on
the identification, processing, and usage of real world
events from the Internet of Things (IoT).
By analyzing four industry scenarios in the area
of logistics, business services, retail, and production,
we discovered requirements for our technical compo-
nents and the concept architecture. With regards to
transparency, the challenge is to identify the relevant
real world events and transform them into consum-
able information pieces for business processes, e.g.
(Heusler et al., 2006). A systematic and consistent
concept is required to transform the IoT information
and, hence, to increase the information value within
business processes. In terms of business processes,
information can be leveraged within the generic de-
sign of business processes as well as during the plan-
ning and execution of specific business process in-
stances. Internal triggers, such as business process de-
viations, as well as external triggers, such as changed
environmental factors, need to be covered by the tar-
geted event-based information system.
To alleviate these challenges, an extension of busi-
ness process modeling and execution is needed, which
integrates IoT data as well as the description of pro-
cess variance. Orchestration tooling as well as knowl-
edge worker support have to be provided in order to
371
Schief M., Kuhn C., Zimmermann B., Rösch P., Waterfeld W., Schimmelpfennig J., Mayer D., Maus H. and Eichler J..
THE ADIWA PROJECT - On the Way to Just-in-Time Process Dynamics based on Events from the Internet of Things.
DOI: 10.5220/0003417503710377
In Proceedings of the 13th International Conference on Enterprise Information Systems (ICEIS-2011), pages 371-377
ISBN: 978-989-8425-56-0
Copyright
c
2011 SCITEPRESS (Science and Technology Publications, Lda.)
realize the full potential of IoT information. Support-
ing users is important for two main aspects: To en-
sure that humans are involved in business processes
that are closer to the real world and, secondly, in or-
der to foster process dynamics, process participants’
know-how from the actual process execution needs to
be exploited. The second aspect ensures a steady pro-
cess evolution and adaptation to the real world.
2 FUNCTIONALITIES
2.1 Event Processing
Event processing plays a central role in the overall
ADiWa approach of connecting business processes
to the real world via IoT technologies. It is essen-
tial to complement the traditional top-down realiza-
tion of business processes with a bottom-up approach
of events. Only in this way can business processes
cope with the unpredictability of the real world.
The connecting of business processes with IoT
technologies results in a wide spread of the technol-
ogy stack in ADiWa. It reaches from business pro-
cess management, via service and event management,
to IoT sensors and devices. Events occurring at all
layers are often intertwined. Thus, there is a strong
need to process and aggregate low-level IoT events
to high level business events (Schmidt and Schief,
2010). This challenge usually requires not only a rel-
atively complex multi-step processing of events into
business events but also complex structured events
during and as the result of the processing. Thus,
there is a strong need for a complex event process-
ing (CEP) component. In our event concept, we dis-
tinguish between two main event processing compo-
nents: Firstly, the CEP component generates and pro-
cesses complex structured events and, secondly, the
event bus provides a scalable distribution via publish-
subscribe-mechanisms.
Our work leverages state of the art research in
these areas. The field of complex event processing
has been initiated by (Luckham, 2002). (Hinze et al.,
2009) gives an overview of the currently-developed
technologies and identifies characteristics of a large
number of applications. It shows that there exist two
approaches for defining event processing language
(EPL) rules: streaming queries, e.g., according to the
SQL-style, and rule-based approaches. (von Ammon
et al., 2008) describes an integration of event process-
ing with Business Process Management (BPM) and
gives examples of practical applications. Although
it describes the first approach for the integration of
event processing with BPM, it does not address the in-
tegration with IoT technologies. There is also no ap-
proach for an integrated event tooling for the business
user from BPM to the IoT. Based on the Java Messag-
ing Service (Hapner et al., 2002), a lot of work has
been done on message bus functionality for the dis-
tributing and brokering of events. (Muhl et al., 2006)
gives an overview of many relevant systems. Though
quite a large number of techniques exists in different
systems, they focus on the level of a messaging sys-
tems and not completely on events.
Beyond the general complex event processing, we
additionally consider several quality criteria as well
as analysis capabilities to further enhance the value of
information provided by the complex events. To un-
derstand and trust the output of the system, the user
requires reliability and guarantees. In ADiWa, we
achieve these goals by a holistic quality determination
and propagation approach. We consider four different
layers associated with quality. At the lowest layer,
the event producers emit events with a certain pro-
duction quality. Typical event producers are sensors
or devices (e.g., RFID readers); corresponding qual-
ity dimensions are, for example, accuracy and com-
pleteness of the measured values. Next, the events are
transmitted and processed with a certain notification
and processing quality, respectively. We determine
the notification quality based on a large set of fea-
tures of the event bus (Hinze et al., 2009). Examples
for such features and their impact on the quality are:
does the event bus support validity intervals of events,
and what is the granularity? Is the event bus capable
of early filtering, and how efficient and accurate is this
filtering? Is there support for distributed processing?
Is there support for privacy? The processing quality
mainly targets the properties of the CEP engine like
the above-mentioned throughput, latency, or accuracy
of complex event detection. Finally, the events af-
fect dynamic business processes with a certain con-
trolling quality. Here, key performance indicators
(KPIs) given by the business process management in-
dicate the success of the dynamic adaptation of busi-
ness processes utilizing events.
To realize the holistic quality determination and
propagation, we make use of an extended metadata
schema. There, we track the individual quality cri-
teria as well as lineage information to additionally
provide an uninterrupted and persistent trail of events
for downstream analyses. For efficiency reasons, only
the event-specific metadata are directly attached to the
events (such as its age), while the remaining informa-
tion is stored in external repositories (quality criteria
of the event producers and event bus). Further, by
benchmarking and monitoring, we evaluate, observe,
and ensure the quality of the event transport.
ICEIS 2011 - 13th International Conference on Enterprise Information Systems
372
With several analysis capabilities, we provide both
the event processing and the user with valuable sup-
plementary functionalities. These functions include
the handling of missing events as well as the de-
tection of new complex events. Further, we enable
the proactive detection of occurring complex events;
here, a user is notified with an alert-event if the occur-
rence probability of a specific complex event exceeds
a given threshold. This alert allows the user to dynam-
ically initiate appropriate actions at an early stage.
2.2 Process Modeling and Execution
Usually, only a low degree of flexibility is supported
in business process management systems. There are
a few approaches dealing with the question of how to
enhance the flexibility of business process execution
(Dadam and Reichert, 2009; Hallerbach et al., 2009).
In addition to offering change operations, rules can
be used to offer a more flexible business process ex-
ecution. An overview of the current state of the art
can be found in (Paschke and Kozlenkov, 2009) and
(Graessle et al., 2006). Based on the enhanced trans-
parency provided by IoT events, ADiWa enables com-
panies to support flexible adaptation and dynamic ex-
ecution of business processes. Four phases in the life-
cycle of a process are considered in our concept: de-
sign, planning, execution and controlling.
The lifecycle of business processes starts with
the design phase. In this phase, business processes
are documented as semi-formal process models. In
ADiWa, we design business processes using an ex-
tended event-driven process chain (Keller et al., 1992)
method, allowing the definition of complex events
and providing a means of integrating real-world IoT
events into business processes. This approach allows
process owners to define process-relevant events and
to specify appropriate process responses. The result-
ing process models are transformed into technical rep-
resentations and are used as templates for both the
planning of individual process instances and the con-
figuration of a process-controlling component.
In the planning phase, results of the modeling
phase are instantiated (respectively adjusted) for spe-
cific process instances. Planning is an iterative pro-
cess in which a large amount of intermediate results is
created and interpreted. For example, all data occur-
ring for one process instance and the matching rules
have to be taken into account. In addition, it must be
possible to adjust plans at a later time, for example,
to adapt them to a suddenly-occurring event. Plan-
ers must be able to specify the required operations.
They expect immediate reactions in order to be able
to continue with planning directly after the adapta-
tion. To enable this rapid interaction, the planning
component must be able to read, analyze, and write
data as quickly as possible. The goal is to enable users
to plan processes flexibly. Thus, it must be possible
to define several alternative planning sequences based
on potential variations of (strategic and operative) ob-
jectives. Likewise, required activities and resources
required need to be adjusted to reflect IoT events.
In the execution phase, the focus is on a flexible
process execution in terms of IoT intelligence. To be
able to offer this functionality, an execution environ-
ment is needed that is based on the results of both the
design and the planning phase. In addition, the ex-
ecution environment must provide a rule-based pro-
cess execution as well as the interpretation of events
from the IoT. All this information must be combined
in order to respond to events showing up during run-
time. As such, during runtime, the execution of a pro-
cess instance can be adapted to its actual context. If
an upcoming event necessitates a change of planning
or even execution of certain process steps, the exe-
cution environment supports this flexibility, for ex-
ample, by offering change operations similar to the
ones defined by (Hallerbach et al., 2009). Adaptations
in process execution or process planning are commu-
nicated as process-relevant events. According to the
event context, three kinds of business process-relevant
events can be distinguished: execution-, planning-,
and design-relevant events. If the execution or plan-
ning of a certain process needs to be changed fre-
quently during runtime, this fact may indicate that
the original process design, on which planning and
execution is based upon, is insufficient or outdated.
Thus, the design needs to be adjusted to ideally fit to
the runtime requirements. Such optimization poten-
tial is detected by a CEP engine (Section 2.1) react-
ing to both execution- and planning-relevant events,
which are aggregated to the above-described design-
relevant events. Design-relevant events trigger gover-
nance processes, which support the process owner in
the optimization of business process models, provid-
ing an automated change management approach for
execution-driven process evolution.
Another key stage for flexible processes is the
controlling phase. The results of monitoring and
controlling affect all process lifecycle phases. The
ADiWa monitoring and controlling component col-
lects event data from the execution platform as well
as identified business process-relevant events from the
CEP engine. It reassembles single as-is process in-
stances and visualizes them as event-driven process
chains. By combining and aggregating the data of IoT
objects belonging to business process instances with
other process-related data (e.g., semantic data), the
THE ADIWA PROJECT - On the Way to Just-in-Time Process Dynamics based on Events from the Internet of Things
373
controlling component is the process memory. Pro-
cess analysts and process managers are able to an-
alyze single process instances as well as aggregated
instances and even process types. In case of devia-
tions in process performance, an alert-event is gener-
ated and communicated to the person in charge (Sec-
tion 2.3) or other subscribers. The employee in charge
can react to the identified issues to control the process
in a flexible way, or change the process.
2.3 User Process Interaction
One major topic in ADiWa is the process-embedded
support of users and the preservation of their process
know-how. Leveraging the users’ know-how from
daily process work is one of the challenges for busi-
ness process management (Riss et al., 2005). There-
fore, several projects tackle this issue, such as Apos-
dle
1
, ACTIVE
2
, and Nepomuk (Riss et al., 2007), to
name a few. However, no approach has yet to con-
sider exploiting know-how from processes traceable
via the IoT.
Nonetheless, supporting users through the IoT,
such as smart items in instrumented environments,
has recently gained momentum. Applications range
from product tracing, such as in SemProM (Schnei-
der and Kroener, 2008), to knowledge-based assis-
tance, such as memory externalization (Kawamura
et al., 2007). The novelty in ADiWa consists of the
comprehensive approach to the use of the IoT to let
users intuitively participate in, intervene, and modify
business processes. This approach is complemented
by means for process-embedded information support
based on IoT-enriched user and process context.
Investigating the project’s industry scenarios, we
find processes which are often coarsely modeled (if at
all), partly unspecific, dynamic, and human-centric,
as well as human-individual parts. In those cases, pro-
cess participants must react to new situations (e.g.,
those which can be tracked by events from the IoT)
and often have to adapt a running process. Partici-
pants range from office workers to process-embedded
mobile workers interacting with smart objects in an
instrumented environment. ADiWa empowers users
with tools to understand processes and provides de-
tailed information on involved objects in order to al-
low controlling, optimization, or different kinds of
intervention such as in exceptional cases. It sup-
ports users in coping with the vast amount of newly-
available information from the IoT.
ADiWa’s interactive user workspace (see Fig-
ure 1) enables process participants to flexibly work on
1
www.aposdle.tu-graz.at
2
www.active-project.eu
process steps by introducing agile task management,
collaboration, and knowledge work support. Here,
a configurable process information cockpit provides
required information, such as process instance data,
data from legacy systems, and occurred events rel-
evant to the user. This approach is combined with
proactive information delivery to satisfy the presumed
information need inferred from the process and user
context. The workspace presents relevant informa-
tion objects (such as documents, notes, guidelines,
and data from smart objects and their product mem-
ory) as well as alternative actions from process know-
how (such as process guidelines, former process in-
stances, and similar problem solutions). Besides sup-
porting knowledge reuse, the environment is designed
to automatically capture process knowledge during
process execution.
ADiWa explicitly focuses on enabling users to ex-
ecute processes in reality while tracking their work-
ing activities automatically, and thus, keeping up-to-
date process information in the system. This func-
tion ranges from giving feedback (e.g., noting reasons
for exceptions) and supporting solution-finding in the
task management environment (e.g., making a check-
list for handling an exception) to directly interacting
with smart objects (e.g., to solve an exception). More-
over, it is also tracked as process audit data. For in-
stance, in the retail scenario, rearranging goods in a
store due to an exception handling does not require
a manual reporting of transferred goods, because this
is already traced by the IoT events sent by the instru-
mented shelves. Comparing the process model to the
actual execution based on traced data, provides a valu-
able basis for performance analysis and optimization.
ADiWa provides a platform for multi-modal inter-
action with a wide range of usage paradigms and in-
terfaces. Within the latter, participants see their tasks,
required information, and supposed actions, as well
as any changes in the dynamic processes. They are
able to interact with smart objects, retrieve additional
information, and provide feedback on process steps.
Another innovative paradigm applied is the digital
pen and paper interaction in mobile scenarios. Writ-
ing with the digital pen on a specially-prepared pa-
per can be interpreted by software either after man-
ual synchronization or by direct streaming (via Blue-
tooth). Applications range from simple note-taking
or filling out a form with handwriting recognition
to selecting commands on a command sheet to con-
trol software. By introducing this as IoT source in
ADiWa, users are able to do their process work on pa-
per while being seamlessly embedded in the business
process and providing direct feedback. In conjunction
with the controlling phase explained in Section 2.2,
ICEIS 2011 - 13th International Conference on Enterprise Information Systems
374
Figure 1: ADiWa big picture of components.
process participants are also able to intuitively com-
bine context data from various sources including IoT
and process data during the planning and execution
phase. They can aggregate performance indicators
and include them in the interactive user workspace as
widgets. These widgets are used as working tools and
can be shared between users and jointly evolved. A
widget configuration can be done without the inter-
vention of IT departments, as it is required in business
process management nowadays.
3 ARCHITECTURE AND
SECURITY
The core architectural paradigm of ADiWa is event-
driven business process management. To enable
this vision, a reference architecture (see Figure 1)
is defined, which defines and integrates the most
important and influential architectural concepts and
mechanisms. All modules within the reference archi-
tecture are self-contained and modular entities with
clearly-defined external interfaces. The interoperabil-
ity will be maintained through the development of
Web services conforming to WS-I
3
guidelines. Part-
ner projects of ADiWa within the Digital Product
Memory Innovation Cluster play an important role.
The SemProM
4
project focuses on the collection and
representation of information during an object’s life-
cycle. The Aletheia
5
project provides a holistic view
of an item’s state. The ADiWa reference architecture
consists of some key components which are essential
for event-driven processes.
The event bus and the CEP enable the routing,
distribution, and the event processing algebra-
based analysis of events.
3
www.ws-i.org
4
[http://www.semprom.org/]
5
[http://www.aletheia-projekt.de/]
The service bus serves as central orchestration
and integration entity of the different components
based on a system-wide service registry.
The process modeling environment is the main
component for design-time business process man-
agement, including maintenance of business rules.
The process execution environment is the com-
ponent for business process run-time, especially
for planning and execution of process instances.
An interactive user workspace assists the human
with appropriate context information and consum-
able information for effective decision-making.
The analytical and evaluation component col-
lects information from events and processes to de-
rive meaningful representations.
Dynamic business processes consume a large
number of events from intelligent objects and base
decisions on those events. Planned as well as ad hoc
changes in complex environments must not violate se-
curity policies. The architecture needs to enforce se-
curity polices despite of weak security mechanisms of
intelligent objects and dynamic environments. Secu-
rity properties of events must be checked in spite of
their large number and high frequency.
There are several business process-based ap-
proaches for the analysis and specification of secu-
rity requirements and controls. Rodriquez et al. (Ro-
driguez et al., 2007) present their method for secure
business process specification to annotate UML activ-
ity diagrams with an UML profile. Modeling security
goals in business processes more explicitly enables
Wolter et al. (Wolter et al., 2009) to adapt their policy
transformation process to different target security in-
frastructures. Basin et al. (Basin et al., 2006) present
with SecureUML a flexible approach developing spe-
cial purpose UML dialects for security specification
and analysis. Neither of them seamlessly integrates
multiple models horizontally or incorporates support
for cross-phase modeling as is proposed in ADiWa.
Accordingly, we define a guideline for the sys-
tematic analysis of security requirements within the
context of dynamic business processes. Further, we
derive a framework for an integrated security mod-
eling of dynamic business processes. Using weav-
ing models (Bezivin et al., 2006), we link security-
related artifacts of dynamic business processes. The
interconnectivity allows for the analysis of security-
related dependencies and serves as a basis for im-
mediate impact analysis, policy derivation, and se-
curity status evaluation. Moreover, we formulate a
method to infer and enforce secure interactions with
smart objects at runtime. Low-level security polices
are decoupled from security expert knowledge and
THE ADIWA PROJECT - On the Way to Just-in-Time Process Dynamics based on Events from the Internet of Things
375
(frequently changing) implementation details utiliz-
ing description logics and ontologies. Thus, we are
developing a process security evaluation engine pro-
viding operators with insight concerning the current
security status of a process. The engine analyzes the
events consumed or generated by the dynamic busi-
ness process using the integrated security model and
the derived security policies. It is capable of simu-
lating possible execution paths and detecting poten-
tial security breaches in the near future. Trustworthi-
ness of the events to be processed is at the heart of
secure event processing. In ADiWa, we are develop-
ing methods to establish trusted relationships between
event processing components to avoid the overhead
of authenticating each event individually. Plausibility
checks are used to detect compromised components.
4 OUTLOOK
In this paper, we provide a very brief overview of
the motivation, progress, and goals of the ADiWa
project. In line with the overall concept, the differ-
ent facets of the research activities deliver a holistic
solution framework that covers the various challenges
in the field of event-based system implementations.
While the described event processing functionalities
increase the needed transparency, the enhanced busi-
ness model design and execution improve the flexibil-
ity of business process adaptations. By that, we aim to
close the gap between technical possibilities and their
business usage and, thus, to demonstrate the business
value of our overall concept. In this respect a deep and
thorough evaluation of the concept and the functional-
ities is our benchmark. Leveraged by our research ac-
tivities, we intend to raise the research field of event-
based system implementations one step higher and to
provide a fertile ground for future research projects.
ACKNOWLEDGEMENTS
The project was funded by means of the German Fed-
eral Ministry of Education and Research under the
promotional reference 01IA08006. The authors take
the responsibility for the contents.
REFERENCES
Basin, D., Doser, J., and Lodderstedt, T. (2006). Model
driven security: From uml models to access control
infrastructures. ACM Transactions on Software Engi-
neering and Methodology, 15(1):39–91.
Bezivin, J., Bouzitouna, S., Del Fabro, M., Gervais, M.,
Jouault, F., Kolovos, D., Kurtev, I., and Paige, R.
(2006). A canonical scheme for model composition.
Lecture Notes in Computer Science, 4066:346–360.
Dadam, P. and Reichert, M. (2009). The ADEPT project:
a decade of research and development for robust and
flexible process support. Computer Science-Research
and Development, 23(2):81–97.
Graessle, P., Schacher, M., Grc, P., et al. (2006). Agile Un-
ternehmen Durch Business Rules. Springer.
Hallerbach, A., Bauer, T., and Reichert, M. (2009). Config-
uration and management of process variants. Hand-
book on BPM.
Hapner, M., Burridge, R., Sharma, R., Fialli, J., and Stout,
K. (2002). Java Message Service Specification v1. 1.
Sun Microsystems, Inc.
Heusler, K., Stolzle, W., and Bachman, H. (2006). Sup-
ply Chain Event Management Grundlagen, Funk-
tionen und potenzielle Akteure. Wirtschaftswis-
senschaftliches Studium, 35(1):19.
Hinze, A., Sachs, K., and Buchmann, A. (2009). Event-
based applications and enabling technologies. In Pro-
ceedings of the Third ACM International Conference
on Distributed Event-Based Systems, page 1.
Kawamura, T., Fukuhara, T., Takeda, H., Kono, Y., and Ki-
dode, M. (2007). Ubiquitous memories: a memory ex-
ternalization system using physical objects. Personal
and Ubiquitous Computing, 11(4):287–298.
Keller, G., Nuettgens, M., and Scheer, A. (1992). Semantis-
che prozessmodellierung auf der grundlage ereignis-
gesteuerter prozessketten (epk). Veroeffentlichungen
des Instituts fuer Wirtschaftsinformatik, 89.
Luckham, D. (2002). The power of events: an introduction
to complex event processing in distributed enterprise
systems. Springer.
Muhl, G., Fiege, L., and Pietzuch, P. (2006). Distributed
Event-Based Systems. Springer.
Paschke, A. and Kozlenkov, A. (2009). Rule-Based Event
Processing and Reaction Rules. In Proceedings of the
2009 International Symposium on Rule Interchange
and Applications, pages 53–66. Springer.
Pesic, M. and van der Aalst, W. (2006). A declarative ap-
proach for flexible business processes management.
LNCS, 4103:169.
Riss, U., Rickayzen, A., Maus, H., and van der Aalst, W.
(2005). Challenges for Business Process and Task
Management. Journal of Universal Knowledge Man-
agement, 0(2):77–100.
Riss, U. V., Jarodzka, H. M., and Grebner, O. (2007).
Pattern-based task management & implicit knowl-
edge. In 4th Conference on Professional Knowledge
Management. GITO Verlag Berlin.
Rodriguez, A., Fernandez-Medina, E., and Piattini, M.
(2007). M-BPSec: A method for security requirement
elicitation from a UML 2.0 business process specifica-
tion. Lecture Notes in Computer Science, 4802:106.
ICEIS 2011 - 13th International Conference on Enterprise Information Systems
376
Schmidt, B. and Schief, M. (2010). Towards Agile Business
Processes Based on the Internet of Things. Advanced
Manufacturing and Sustainable Logistics.
Schneider, M. and Kroener, A. (2008). The smart pizza
packing: An application of object memories. In IE’08.
von Ammon, R., Greiner, T., Paschke, A., Springer, F.,
and Wolff, C. (2008). Event-Driven Business Process
Management. OBJEKTSpektrum.
Wolter, C., Menzel, M., Schaad, A., Miseldine, P., and
Meinel, C. (2009). Model-driven business process se-
curity requirement specification. Journal of Systems
Architecture, 55(4):211–223.
THE ADIWA PROJECT - On the Way to Just-in-Time Process Dynamics based on Events from the Internet of Things
377