[Paper Review] Interaction semantics and its implications for an interaction oriented architecture of IoT-type applications

Johannes Reich|arXiv (Cornell University)|Jan 29, 2017

IoT and Edge/Fog Computing3 citations

TL;DR

This paper proposes an interaction-oriented software architecture for IoT applications based on a formal reference model of interaction semantics. It enables loosely coupled, reusable components by defining semantic alignment, layered abstraction, and clear separation of concerns, significantly reducing adaptation effort in dynamic interaction networks.

ABSTRACT

Several synergistic trends, subsumed under the phrase "Internet of things (IoT)" massively drive the increasing importance of networking applications. In the past, the exponential growth of the Internet was mainly due to semantically agnostic transport protocols. In the future it is to be expected that, because of the increasing autonomy of technical systems, it becomes necessary to better understand the nature of the semantics of these interaction networks to create appropriate networking applications. Appropriate means that the architecture of these applications allows to minimize the effort to adapt these applications to the permanently changing interaction networks. The proposed interaction oriented architecture is based on a reference model of interaction semantics. It provides guiding principles on how to design networking applications. The reference model of interaction semantics provides: a unifying description of the things in the physical, the information and the human world; an interaction model that is of direct runtime relevance; an understanding for how hierarchical structured components can cooperate loosely coupled; a concept to determine how much semantics has to be common to enable components of different semantic domains to cooperate loosely coupled; and a data type model. The software reference architecture provides: a definition of software layers; means to express vertical interactions, that is interactions which demarcate a software layer; means to express horizontal interactions, that is, between processes in the same software layer; a definition of a component and how to distinguish it from other entities like systems or objects; and a model how to separate reusable from non-reusable parts of an application's functionality.

Motivation & Objective

• Address the growing complexity of IoT applications driven by autonomous, interconnected systems.
• Overcome limitations of semantically agnostic protocols by embedding semantic awareness in system design.
• Minimize adaptation overhead in evolving interaction networks through structured semantic modeling.
• Provide a unifying framework for interoperability across physical, information, and human domains.
• Enable modular, reusable software components through clear separation of functionality and semantic alignment.

Proposed method

• Introduces a reference model of interaction semantics to unify descriptions across physical, information, and human worlds.
• Defines a runtime-relevant interaction model that supports dynamic, loosely coupled component cooperation.
• Establishes a semantic alignment concept to determine minimal common semantics required for cross-domain component interaction.
• Proposes a layered software architecture with distinct vertical and horizontal interaction patterns between components.
• Introduces a component model that differentiates components from systems or objects based on functional and semantic boundaries.
• Applies a separation of reusable and non-reusable functionality through semantic and structural decomposition in the architecture.

Experimental results

Research questions

• RQ1 How can semantic alignment be formally defined to enable interoperability between components in different semantic domains?
• RQ2 What architectural principles support loose coupling while maintaining semantic coherence in dynamic IoT interaction networks?
• RQ3 How can software layers be structured to support both vertical (layered) and horizontal (peer) interactions effectively?
• RQ4 What criteria distinguish reusable components from non-reusable ones in an interaction-oriented IoT architecture?
• RQ5 How can a unified semantic model bridge the physical, information, and human worlds in IoT applications?

Key findings

• The reference model of interaction semantics enables a unifying description of entities across physical, information, and human domains.
• The architecture supports direct runtime relevance of interaction semantics, improving dynamic system adaptability.
• Hierarchical components can cooperate loosely coupled through defined semantic alignment thresholds.
• The model quantifies the minimal required semantic commonality for cross-domain component interaction.
• The software architecture clearly separates reusable from non-reusable functionality, enhancing maintainability.
• Vertical and horizontal interaction patterns are formally expressed, enabling structured communication across layers and peers.