Knowledge-driven architecture composition

Burzlaff, Fabian

PDF
Dissertation.pdf - Veröffentlichte Version
Download (7MB)

URL:	https://madoc.bib.uni-mannheim.de/59125
URN:	urn:nbn:de:bsz:180-madoc-591257
Dokumenttyp:	Dissertation
Erscheinungsjahr:	2021
Ort der Veröffentlichung:	Mannheim
Hochschule:	Universität Mannheim
Gutachter:	Stuckenschmidt, Heiner
Datum der mündl. Prüfung:	29 März 2021
Sprache der Veröffentlichung:	Englisch
Einrichtung:	Fakultät für Wirtschaftsinformatik und Wirtschaftsmathematik > Practical Computer Science II: Artificial Intelligence (Stuckenschmidt 2009-) Außerfakultäre Einrichtungen > Institut für Enterprise Systems (InES)
Fachgebiet:	004 Informatik
Freie Schlagwörter (Deutsch):	Wissensbasierte Systeme , Software Architekturen , IoT , Semantische Interoperabilität , Softwareentwicklung
Freie Schlagwörter (Englisch):	knowledge-driven architecture composition , software architecture , IoT , semantic interoperability , software engineering
Abstract:	Service interoperability for embedded devices is a mandatory feature for dynamically changing Internet-of-Things and Industry 4.0 software platforms. Service interoperability is achieved on a technical, syntactic, and semantic level. If service interoperability is achieved on all layers, plug and play functionality known from USB storage sticks or printer drivers becomes feasible. As a result, micro batch size production, individualized automation solution, or job order production become affordable. However, interoperability at the semantic layer is still a problem for the maturing class of IoT systems. Current solutions to achieve semantic integration of IoT devices’ heterogeneous services include standards, machine-understandable service descriptions, and the implementation of software adapters. Standardization bodies such as the VDMA tackle the problem by providing a reference software architecture and an information meta model for building up domain standards. For instance, the universal machine technology interface (UMATI) facilitates the data exchange between machines, components, installations, and their integration into a customerand user-specific IT ecosystem for mechanical engineering and plant construction worldwide. Automated component integration approaches fill the gap of software interfaces that are not relying on a global standard. These approaches translate required into provided software interfaces based on the needed architectural styles (e.g., client-server, layered, publish-subscribe, or cloud-based) using additional component descriptions. Interoperability at the semantic layer is achieved by relying on a shared domain vocabulary (e.g., an ontology) and service description (e.g., SAWSDL) used by all devices involved. If these service descriptions are available and machine-understandable knowledge of how to integrate software components on the functional and behavioral level is available, plug and play scenarios are feasible. Both standards and formal service descriptions cannot be applied effectively to IoT systems as they rely on the assumption that the semantic domain is completely known when they are noted down. This assumption is hard to believe as an increasing number of decentralized developed and connected IoT devices will exist (i.e., 30.73 billion in 2020 and 75.44 billion in 2025). If standards are applied in IoT systems, they must be updated continuously, so they contain the most recent domain knowledge agreed upon centrally and ahead of application. Although formal descriptions of concrete integration contexts can happen in a decentralized manner, they still rely on the assumption that the knowledge once noted down is complete. Hence, if an interoperable service from a new device is available that has not been considered in the initial integration context, the formal descriptions must be updated continuously. Both the formalization effort and keeping standards up to date result in too much additional engineering effort. Consequently, practitioners rely on implementing software adapters manually. However, this dull solution hardly scales with the increasing number of IoT devices. In this work, we introduce a novel engineering method that explicitly allows for an incomplete semantic domain description without losing the ability for automated IoT system integration. Dropping the completeness claim requires the management of incomplete integration knowledge. By sharing integration knowledge centrally, we assist the system integrator in automating software adapter generation. In addition to existing approaches, we enable semantic integration for services by making integration knowledge reusable. We empirically show with students that integration effort can be lowered in a home automation context.