The Internet of Things (IoT) is a technological topic with a very important societal impact. IoT application domains are various, such as smart cities, precision farming, smart factories, smart buildings, etc, and the diversity of these application domains is the source of the very high technological heterogeneity in the IoT, leading to interoperability issues. The semantic web principles and technologies are more and more adopted as a solution to these interoperability issues, leading to the emergence of a new domain, the Semantic Web Of Things (SWoT). Scienti c contributions to the SWoT are many, and the diversity of architectures in which they are expressed complicates comparison. To unify the presented state-of-the-art architectures, we propose an architectural pattern, Lower, Middle and Upper Node (LMU-N). LMU-N provides a reading grid used to classify processes to which the SWoT community contributes, and to describe how the semantic web impacts the IoT. A survey, based on this reading grid, capturing the integration of semantics into the IoT is the core of this paper. Then, the evolution of the semantic web to adapt to the IoT constraints is described as well, in order to give a twofold view of the convergence between the IoT and the semantic web toward the SWoT.
The Internet of Things (IoT) is both a scienti c, technological and societal
challenge. The networks of connected devices are becoming an everyday reality for
citizens across the world. The de nition of the notion of IoT has evolved since
it was rst proposed in 1999 by Kevin Ashton4. An up-to-date de nition is
proposed by [
This heterogeneity is a cause to the vertical integration of solutions: the same
vendors distribute an application-dedicated hardware communicating with a
potentially proprietary protocol with a dedicated application [
However, interoperability remains an issue at the semantic level. [
The contributions of the semantic web community toward the IoT are many: we counted 1426 publications at the end of 20169, which contributions are expressed in various architectures, and with complementary or competing visions. We focused on 71 scienti c publications10 for the survey this abstract advertises for, chosen for their quality, their innovative aspect, and for the balance in their content between semantic web and IoT. An extra attention was given to publications proposing semantic web contributions in explicitly and precisely de ned IoT architectures.
The heterogeneity of the contributions presented makes their comparison harder. That is why we propose LMU-N, a unifying architectural pattern designed to describe the convergent contributions of the semantic web and the IoT communities. The LMU-N description is a preliminary to the survey at the core of the paper: LMU-N is used as a framework to contextualize semantic web contributions to the IoT. The use of a uni ed architectural pattern allows to study how the IoT architectures, and the characteristics of their nodes, in uences the way semantic web approached can be integrated.
The remaining of the paper if organized as follows: rst, an overview of the LMU-N pattern is provided. Then, we explain how LMU-N can be used as a reading grid for a survey of SWoT contributions. Especially, it allowed the identi cation of data and system oriented processes instantiated in IoT 5 http://onem2m.org/ 6 https://openconnectivity.org/ 7 Hardware and software interoperability, e.g. shared communication protocols 8 Data format interoperability (XML, JSON...) 9 After a study on http://ieeexplore.ieee.org, http://www.sciencedirect.com and http://dl.acm.org, looking up the keywords "semantic web" and "internet of things" 10 Which complete list is available in the full survey paper (link in the next page) architectures enriched by semantic web principles. The papers of the survey are classi ed according to their contributions to these processes, situated in the LMU-N pattern. For further information, the survey described by this abstract is available in open access11. 2
LMU-N is an architectural pattern issued from a bottom-up analysis conducted during the redaction of the survey this abstract introduces. LMU-N was derived from recurring patterns in already existing architectures presented in the surveyed papers. LMU-N is constituted of two main components, the nodes and the ows. The present section provides an overview of LMU-N, and references some of the architectures from which this pattern was extracted.
The node is a communicating entity on the network, an abstraction of both device and service, which play similar roles in the papers we surveyed. This abstraction aims at focusing on the common features of nodes, especially their intrinsic capabilities, and on how they process data. This design choice can be found in [4{8]. Based on their characteristics (memory, processing power, communication capabilities), the nodes can be clustered into three categories: { Upper Node (UN) have high processing power, extended communication capabilities, and large storage capabilities (typically servers). { Middle Node (MN) are very often referred to in the literature as gateway.
They typically have medium processing power, extended communication capabilities, and restricted memory storage. { Lower Node (LN) are typically connected devices, with very limited power source, processing and communication capabilities, and very limited to no storage capabilities.
The other important element of the LMU-N is the ow. In the representation
of an IoT network as a graph, the edges are communications between nodes,
instantiated by messages ows. As shown on g. 1, ows can be directed in three
general directions: horizontal for nodes of the same level, such as the work done
in [
Contributions of the semantic web to IoT processes The core hypothesis in the present survey is that contributions from the semantic web to the IoT can be clustered into recurring patterns, called processes, and that these processes are constrained by the nodes and the ows that support them. This analysis aims at identifying how the semantic web principles can be weaved into IoT architectures. LMU-N is used as a reference frame where the contributions of surveyed papers are situated in terms of nodes and ows involved (e.g. upstream from lower to middle node), and in terms of processes (e.g. enrichment, node selection, abstraction...).
Processes can be separated in two categories, based on their topic of interest: the content-oriented processes, supporting application-dedicated processing, and the node-oriented processes, where the exchanged and processed data describes the graph of nodes itself. These two types of processes are both contributing to interoperability, at di erent applicative levels. We identi ed 11 content-oriented processes, covered by 38 publications and 6 node-oriented processes covered by 25 publications.
{ Content-centric processes have been classi ed according to three aspects: Representation transformation: Processes where the expressiveness of the content is modi ed, but not its core meaning. This category includes Enrichment and Lowering.
Transport/provisioning: Processes focused on content distribution across the IoT network, without modi cation of said content. This category includes Noti cation/Dissemination, Control, Routing, and Querying.
Processing: Processes using content for applicative purposes and value creation. This category includes Abstraction, Consistency enforcement, Aggregation, Visualization and Decision support. { Node-centric processes have been classi ed according to two aspects: Awareness: Processes dedicated to the representation for a node of other nodes of the network, and of their capabilities. This category includes Discovery, Exposition and Selection.
Homogeneity: Processes dedicated to solving issues related to nodes heterogeneity. This category includes Abstraction, Composition and Speci cation/Con guration.
This classi cation, with a description of each process and of its instantiations by surveyed papers, is the main contribution of the survey.
Moreover, the comparison of the contributions within a reference architecture allowed the identi cation of architectural trends in node communication patterns, and these trends are structuring the integration of the semantic web into the IoT. For instance, the hierarchical topology of IoT architectures, as well as the di erence of processing capacities of nodes of di erent levels, are an incentive to push the resolutions of an issue for a given level to a node of an upper layer. This trend is one of the causes to the predominance of upstream processes observed in the survey. The identi cation of trends is supported by the analysis of the contributions studied in the complete survey.
This analysis of how the semantic web contributes to the IoT is completed by a survey of the evolution of the semantic web to adapt to the characteristics and constraints of the IoT. Transformations include the integration of semantic web technologies into dedicated protocol (such as CoAP [14]) and formats. The adaptation to IoT data is also driven by streaming approaches, and OBDA [15]. This twofold analysis emphasizes the convergence of both the IoT and the semantic web toward the SWoT. 4
To face the interoperability issues intrinsic to the expansion of the IoT, semantic web principles and technologies are more and more commonly weaved into IoT architectures, leading to the emergence of the SWoT. This paper proposes a survey of this convergence of IoT and semantic web. The use of a unifying architectural pattern allowed us to propose a state-of-the-art of the SWoT based on a reusable structure. This analysis supports our claim that processes in an IoT architecture are constrained by the nodes and ows they involve, entailing identi able trends noticeable in the survey. The study of the reciprocal interactions between the IoT and the semantic web shows that not only does the semantic web provide solutions to the interoperability and complexity issues of the IoT, but the IoT also challenges the semantic web principles and technologies to evolve to be compliant with its constraints.
Some issues identi ed by the literature remain open challenges, such as consistency of data across the network or scalability to face the constant increase of the number of nodes, and therefore the constant increase of the volume of processed data. Future works include the study of a decentralized approach to face scalability without compromising consistency, and compliant with the di erent nodes and ows described by LMU-N.
A promising opportunity opened by the development of the SWoT as identied in this survey is the evolution of IoT traditional machine-centric data into richer, more expressive content, via its description with vocabularies connected to natural language resources. Machine-to-Machine (M2M) communication being by design unsuitable to human, natural language resources can be used to enable meaningful user interactions. Moreover, many ontology alignment techniques are based on natural language processing, and they represent a next step in semantic interoperability. 10. A. Sheth, C. Henson, and S. S. Sahoo, \Semantic Sensor Web," in IEEE Internet
Computing, vol. 12, pp. 78{83, 2008. 11. S. Poslad, S. E. Middleton, F. Chaves, R. Tao, O. Necmioglu, and U. Bugel, \A Semantic IoT Early Warning System for Natural Environment Crisis Management," IEEE Transactions on Emerging Topics in Computing, vol. 3, no. 2, pp. 246{257, 2015. 12. J. Rowley, \The wisdom hierarchy: representations of the DIKW hierarchy," Journal of Information Science, vol. 33, no. 2, pp. 163{180, 2007. 13. N. Seydoux, K. Drira, N. Hernandez, and T. Monteil, \Autonomy through knowledge: how IoT-O supports the management of a connected apartment," in Semantic Web Technologies for the Internet of Things, pp. 67{78, 2016. 14. G. Loseto, S. Ieva, F. Gramegna, M. Ruta, F. Scioscia, E. D. Sciascio, and B. I, \Linked Data ( in low-resource ) Platforms : a mapping for Constrained Application Protocol," in ISWC, (Kobe), 2016. 15. J.-p. J. Calbimonte, H. Jeung, O. Corcho, and K. Aberer, \Semantic Sensor Data Search in a Large-Scale Federated Sensor Network," Semantic Sensor Networks, pp. 14{29, 2011.