An ontology for the automated deployment of applications in heterogeneous IoT environments

Paper Title: 
An ontology for the automated deployment of applications in heterogeneous IoT environments
Authors: 
Konstantinos Kotis and Artem Katasonov
Abstract: 
In the near future, Internet of Things should integrate an extremely large amount of heterogeneous entities. To tackle heterogeneity, these entities will need to be consistently and formally represented and managed (registered, aligned, composed and queried) trough suitable abstraction technologies. Two distinct types of these entities are a) sensing/actuating devices that observe some features of interest or act on some other entities (call it ‘smart entities’), and b) applications that utilize the data sensed from or sent to the smart entities (call it ‘control entities’). The aim of this paper is to present an ontology as the key technology for the abstraction of these entities, towards supporting the automated deployment of control entities in settings where smart entities have been already deployed. In specific, the paper presents the use of this technology to support the following required distinct tasks of this deployment process: a) the semantic registration of heterogeneous IoT entities, b) the alignment of IoT entities’ metadata and use of these alignments for the matchmaking between smart and control entities, and c) the alignment of the semantics of the data of the messages that are exchanged between these IoT entities during device-to-application communication. The paper presents a use case scenario and the formal definitions of the main concepts and properties of the proposed ontology. Furthermore, the paper presents a proof-of-concept-based ontology evaluation approach.
Full PDF Version: 
Submission type: 
Ontology Description
Responsible editor: 
Decision/Status: 
Reject
Reviews: 

Revised submission after a "reject and resubmit", now rejected. Previous title: "An IoT-ontology for the Representation of Interconnected, Clustered and Aligned Smart Entities", reviews beneath the second round reviews.

Solicited review by Michael Compton:

This paper discusses an ontology for the internet of things (IoT) and alignment between ontologies representing IoT concepts.

This paper has been substantially rewritten since I reviewed it previously. However, I feel that much of my introductory comments to the previous version still apply to this paper: in that, there seem to be good ideas here for an IoT ontology based on SSN and DUL, and that in the IoT there is likely to be some heterogeneity in descriptions as much as devices and so a mediation or alignment may be required; but, the paper falls short of clearly articulating the ontology design or the alignment mechanisms.

The paper claims its contribution as the ontology as a means to support semantic registration, metadata/matchmaking alignment and message alignment. The paper doesn't describe these in enough detail or technical clarity.

The paper is still verbose, taking 8 1/2 pages before the ontology is introduced. The description itself covers only four concepts from the ontology and the text merely repeats the information in the OWL definitions. More interesting would be a description of how the whole ontology works together, descriptions of why it is designed in the way it is (the are many points here worthy of discussion, such as why an iot:IoT_Entity is a dul:Situation) and what reasoning or technical benefit there is from the given structure.

Alignment and message alignment are very much glossed over, leaving the reader to guess at the details.

The example didn't motivate the need for this infrastructure (it's goal and results are something that happens already without IoT), nor was it carried through the paper in a way that clearly explained the technology presented.

Solicited review by Sara Hachem:


In this paper, the authors address the problems of diversity of things within IoT and the modeling of things registry /gateways to support better interoperability. As a solution, they introduce an ontology that models knowledge about things in the Internet of Things (IoT) along with the way they should interoperate (for producers to easily register those things, and consumers to easily coordinate and retrieve them).

In more detail, authors start by motivating their work in terms of heterogeneity if things in the IoT. Then they present a scenario where new heterogeneous devices can be introduced without adding large overhead to required human effort. After which they position their work with respect to existing solutions, mostly ontologies developed in the context of IoT related solutions.

The authors also define the aim of the their ontology which is providing a means of disambiguation of terminology and aligning entities in the IoT. The authors then describe the modeling of entities and their alignment along with the ontologies they extended. Finally they introduce their evaluations, which are represented through SPARQL queries.

Technical contribution:
The main contribution is:
- A semantic model of IoT entities through an ontology that extends DUL ontologies. The most important concepts are: Smart entities and Control entities. Smart entities have four types: embedded, actuating, computing and attached devices. Control entities represent software components.

Strengths

+ The main topic of the paper, which is ontologies for the Internet of things, is of high importance.
+ The main contribution in the paper, which is an approach to support interoperability of things has not been completely solved in the literature yet
+ The related work section is recent and covers existing solutions.

The idea of the work is interesting and appropriate for a semantic web journal. However, there are still some modifications and clarifications to be made and I can not recommend an accept before the issues are addressed.

Weaknesses
- There is no description in the contribution of data mapping which makes the mapping between entities incomplete. How does the ontology model the mapping between the type of data provided by the smart entity and the type of data acquired by the control entity?
- The example is still too long, especially that it is only referred to in the evaluation section.
- The evaluation section should be a validation section as it does not show any evaluations. A large part of it could have been provided as an illustration while presenting the contribution.
- The language is too redundant and repetitive, specifically the text before the main contribution section. As a result, only a short space is attributed for the ontology model itself.

Solicited review by Anna Fensel:

Considering the motivation of the work i.e. to suggest the ontology to have it used as a registry, I have a problem as I do not see the evidence of the present or future impact in this work. Technically registries existed in other fields e.g. UDDI for web services, but it is difficult to get commitment for these.
The ontology suggested by the authors has no backing, involvement or use from industry/stakeholders (e.g. device manufacturers or similar IoT players committed to use it). Ontologies should be shared, but here I fail to observe that.
One of the platforms that has been relatively successful to serve as an open IoT registry has been Pachube (bought by LogMeIn a year or so ago). Even though strictly speaking they did not use semantics, they had a good ontological impact in IoT data sharing.

The ontologies also could be published on publicly available repositories such as CKAN to be available for the communities, instead of the difficult-to-find private websites of the authors.

Technically, I also have problems with the published ontology and could not review it thoroughly. When I tried to open the latest ontology version - v2.1 - in Protege v3.4.4, it has been impossible - 519 errors were reported.

First round reviews:

Solicited review by Michael Compton:

This paper presents an ontology for the Internet of Things. It discusses the purpose and setting of the ontology along with new concepts introduced and alignment with other ontologies (DUL,SSN, foaf).

Firstly, I found the language in the paper verbose and repetitive, which I think somewhat hampered the reading. That aside, the topic and general argument of the paper are aligned with the cfp. Further, the idea of a generic ontology as a mediation between the ontologies of various parties in the IoT has quite some merit. The paper adds a further interesting layer with the idea that the mediation and alignment might need to be carried out automatically and might therefor contain error and must be regarded with an uncertainty bound. The choices of DUL, the SSN ontology and foaf are well motivated.

I think, however, that the presented ontology is flawed and rather works against the claimed purpose of the paper. There are some restrictions that are questionable and the alignment to foaf, SSN and DUL seems incorrect.

There seems to be some mistakes in the description: such as the text stating that a smart entity having a Person as a owner - the ontology doesn't say this (nor does the fragment in Figure 1).

Some of the restrictions are questionable. For example, that a smart entity (or a device as stated in the ontology) be owned by a Person is overly restrictive - why not owned by a company or a government?

The general design of the ontology is also not correct, nor does it foster reuse and alignment. To take the concepts of supposedly upper ontologies and place them as subsumed by concepts in the presented IoT ontology, or to otherwise redefine the concepts of upper ontologies (as is done with ssn:Device) alters the established meanings of those concepts. This creates a number of problems from increasing the chance of errors to making alignment with other ontologies more difficult and certainly defeats the arguments presented in the paper of reuse and for using these existing ontologies because of their acceptance and agreement on them.

A better strategy is to simply use the upper ontology as a basis for new concepts: for example, why create Physical_entity when all its concepts could be classified under appropriate children of PhysicalObject. In cases of redefining or adding a subsuming concept to an upper concept, it is better to define out part (a subconcept) of the upper concept that has the required property.

The example given in the paper was long and not really used in enhancing the remainder of the paper.

The purl URL for the ontology didn't load. I was able to download from the dropbox, however the IoT ontology there contained errors and wouldn't classify.

While I did find some promising aspects to this paper, the flawed modelling, verbose and unclear description and the problems with the ontology itself mean that I cannot recommend accepting it. The modelling needs to be either reconsidered or explained more clearly and the paper needs to be rewritten to make it clearer and less verbose.

Solicited review by Andriy Nikolov:

The paper describes the design of an ontology aimed to serve as a mediating schema in ubiquitous computing environments. The ontology extends existing schemas covering foundational concepts (DUL) and common aspects relevant to information exchange in networks of smart devices (e.g., SSN and QUDT) and introduces concepts and properties representing IoT specific information: taxonomy of device types, network composition, etc. This is an important domain, where the emergence of a reusable ontology would be beneficial. One aspect, which differentiates the proposed ontology from some existing ones dealing with the ubiquitous computing domain (e.g., SOUPA, CONON, etc.) is the design decision to build on top of existing standard ontologies. This is promising as it makes reuse of both schema and data potentially easier.

However, there are several issues in the paper which do not look fully clear. In particular, this relates to the envisaged scenarios of actual interoperability within networks of smart devices. The paper mentions automatic or semi-automatic run-time matching of schema ontologies provided by different devices vendors, maintaining produced alignments, and using them for actual integration of data provided by different devices. Although the actual matching is outside the scope of the paper, this assumption makes the whole scenario look unrealistic given the limited reliability of existing automatic ontology matching tools, as well as limited expressivity of alignments which can be produced. From the description of the use case in section 3.1 it even looks like the ontologies from different vendors have to be aligned pairwise rather than to a single common "denominator" schema. Making the process semi-automatic would not help in this case, as the end-user (Mary) cannot be relied on to refine the schema matching results. In general, the role of the proposed ontology in the described scenario is unclear, if it cannot serve as a shared high-level schema by different vendors than it is not clear what is its added value for the use case. These issues should be clarified.

I could not check the actual ontology: the server to which the address http://purl.org/IoT/iot redirects was down. But from the provided description it seems to focus primarily on the taxonomy of devices rather than on defining the structures for data exchanged in the network: the only pattern of collaboration is the reuse of sensor observations provided by other devices in the network. Are more complex ways to inter-operate envisaged: e.g., joint handling of a task or combining data for situation awareness?

What is the principle for representing the device hierarchy (Sensing_device, Actuating_device, Embedded_device, Attached_device, etc.): do these carry different information about different types of devices or are used only for classification?

The ontology currently includes both generic and domain-specific concepts and properties. While domain-specific ones are primarily intended for illustration, this can be inconvenient for reuse. Perhaps, domain-specific ones should be separated and use a different namespace.

A minor thing: the example uses Heat as a FeatureOfInterest, while LivingRoomTemperature as its quality. Not sure if this is a proper way of defining it, as heat is not an object itself, but rather a quality (i.e., temperature itself) of the living room.

Some typos:

- p.7: "a Software agent that me be responsible"
- p. 8: "The related to Control entity classes" -> "The classes related to Control entity"
- p. 9: "To better support this decision" -> "To support this decision better"

Solicited review by Sara Hachem:

Summary
In this paper, the authors address the problems of diversity of things within IoT and the modeling of things registry /gateways to support better interoperability. As a solution, they introduce an ontology that models knowledge about things in the Internet of Things (IoT) along with the way they should interoperate in order to address the diversity of the billions of devices that should constitute the IoT. They also focus on modeling semantic registries and gateways for producers to easily register those things, and consumers to easily coordinate and retrieve them.

In more detail, authors start by motivating their work in terms of heterogeneity of things in the IoT. Then they position their work with respect to existing solutions, mostly ontologies developed in the context of IoT-related solutions. After which they present their motivating scenario that is similar to a smart-home scenario where new heterogeneous devices can be introduced and can cooperate without adding large overhead to required human effort.
After the scenario, authors define the aim of their ontologies, which is providing a means to disambiguation of terminology, representing interconnected, clustered, and aligned entities in the Web of Things, and finally modeling support for smart gateways. Authors then describe the modeling of IoT entities, the ontologies they extended, and the model they use for coordination and alignment between ontologies. Finally, they introduce the evaluations, which are represented through SPARQL queries.

Technical contribution
The contribution can be divided into two main parts:
- A semantic model of IoT entities through an ontology that extends DUL, QUDT, and FOAF ontologies, and is aligned with SWEET ontologies. The most important concepts in the ontology are: Smart entities and Control entities. Smart entities represent 4 types of devices: embedded, actuating, computing and attached devices. Control entities represent software components. The model also includes clusters of entities, which are comprised of entities that belong to similar domains, in addition to domain specific entities such as sensor models.
- The modeling of a smart gateway that supports learning of conceptual schemas of new entities and alignments of ontologies to allow better registration and retrieval of things. The importance of the semi-automatic alignment of smart entities definitions at run time is highlighted, however, how the actual alignment takes place is not introduced as it is considered to be out of the scope of the paper.

The contribution is interesting but it is not appropriate for a semantic technologies journal, unless much more details on the alignment of ontologies where provided (given that it is presented as one of the most important contributions in the paper), which is not the case. The most details in the contribution section are provided for the ontology modeling things, which is not a novelty and therefore decreases the quality of the paper, and also decreases the quality and relevance of the described ontology.

Strengths
+ The structure of the paper, on a high level, is coherent and provides a proper flow of information.
+ The main topic of the paper, which is ontologies for the Internet of things, is of high importance.
+ The main contribution in the paper, which is an approach to support interoperability of things has not been completely solved in the literature yet which makes it an important contribution, as it indeed satisfies the requirements of the IoT to incrementally grow (in terms of devices and software gradually), to interconnect devices and allow them to communicate and cooperate with each other, to provide applications for IoT environments by 3rd parties, in addition to providing generic software that is not bound to a specific type of devices.
+ The presented ontologies extend existing standard ontologies.
+ The motivating scenario shows clearly how the contribution can provide better interoperability.
+ In general, the English is fluent and makes the paper easy to read.

Weaknesses
-Throughout the paper, the scale issue is ignored, as it is not taken into consideration in the presented approach to answer things-related queries. This risks being a bottleneck and renders the contribution not very useable in the large scale IoT, unless it is used on small scales (such as few devices in the user's home), which then cannot be represented as a solution for the IoT.
- The motivating scenario is not sufficiently referred to throughout the paper, which makes the contribution harder to visualize (for example, regarding clustering and data integration), especially that the modeling of those two steps is not described in details in the paper. This decreases the illustration and clarity of the paper.
- The analysis in the related work section regarding the contribution by Hachem et al is not accurate. In more detail, in the described ontology, authors do not only focus on devices, but also on composition and estimation of measurements.
- The title of the paper focuses on clustered and aligned entities, however clustering and alignment modeling is not well described in the paper, which raises a big issue given that it is repeatedly represented as one of the most important contributions in the paper.
- Figures 1, 2 and 4 are too dense and are better represented as graphs.

Typos etc.
- Page 3, column 1: "[...] loosely-coupled interoperability than any other form [...]" -> "[...] loosely-coupled interoperability as opposed to any other form [...]".
- Page 4, column 1: "[...] Hackem et al in [...]" -> "[...] Hachem et al in [...]".
- Page 7, column 2: "Software agent that me be responsible" -> "Software agent that may be responsible".
- Page 9, column 2: "[...] put effort in the alignment of them to common and widely used concepts as much as possible such as [...]" -> "[...] put as much effort as possible in their alignment to common and widely used concepts such as [...]".
- Page 11, column 1: "we demonstrate the this case" -> "we demonstrate this case".

Solicited review by Anna Fensel:

The paper presents a generic Internet of Things (IoT) ontology, that incorporates and combines some of the existing vocabularies being developed in the area (e.g. SSN), as well as adds new concepts.

The motivation and purpose of the ontology is not so clearly articulated: if the ontology is to serve as yet another possibility to link other ontologies, it is not enough to justify its existence. Also from the data point of view, the presence of the datasets or specific use cases proving the need for such ontology is not demonstrated.

The paper also includes a vague description the middleware which could mediate between the proposed ontology and other IoT ontologies.
Clearly, such approach overlaps with from similar developments a number of other IoT projects, which e.g. the paper itself mentions on page 4.

Formally, from the paper presentation point of view, only the requirements for the ontology design are being spread over several sections, which makes it difficult to follow.
In the Appendix, it would be better not to use the color mark up (such as light green) - many people print out and read the paper files in black and white.

Tags: 

Comments

This paper was submitted as part of the 'Big Data: Theory and Practice' call.