BGSD: A SBERT and GAT-based Service Discovery Framework for Heterogeneous Distributed IoT
Introduction
The current IoT architecture is mainly based on the cloud platform, and the IoT devices access the network and connect to the cloud platform through 4G, WiFi or other technologies [1], [2]. Thanks to the powerful storage and computing capabilities of the cloud platform, under its control, heterogeneous IoT devices with different capabilities can be interconnected to offer complex services that a single device cannot provide. However, the architecture based on the centralized cloud platform also has its inherent shortcomings [3], for example, it requires reliable network and cloud platform infrastructure. In situations where communication is unstable, such as emergencies, disaster relief, etc., these “offline” IoT devices will be out of service, which may lead to unexpected consequences. Therefore, for these scenarios with dynamic topology and heterogeneous devices, it is very necessary to form flexible and autonomous collaboration among them in a distributed manner, which is called SIoT (Social IoT) [4], [5].
The SIoT decouples smart devices through a distributed service architecture with semantic services, so that each device can search for the required services in the network on demand, which can effectively enhance the flexibility of the system [6]. There are three key technologies to realize this idea: (i) how to improve the matching accuracy between requests and services, which is the focus of researchers, (ii) how to reduce the cost of searching for the required services in a distributed network, and (iii) how to reduce the difficulty of compatibility development between devices of different manufacturers or different periods.
To solve the above problems, this paper proposes an efficient service discovery method based on SBERT [7] and GAT [8], named as BGSD, which can improve the service search accuracy while reducing the overhead of system resources. BGSD allows the nodes in SIoT to simulate people’s social processes to perceive the local topology environment, and manage and discover services in a distributed manner, making SIoT highly navigable and scalable [4], [9]. Its few online computations and model parameters will also allow BGSD to be applied to a wide range of heterogeneous SIoT devices.
The main contributions of this paper include the following aspects:
- (a)
A distributed intelligent IoT architecture based on OWL-S (Web Ontology Language for Services) and MANET (Mobile Ad Hoc Network) is constructed to ensure that the network has good flexibility and robustness.
- (b)
A service matching method based on SBERT is proposed. The key text information in the node service description is extracted, and embedded into the feature vector based on the SBERT model offline. This can effectively improve the matching accuracy and reduce the amount of online calculation.
- (c)
A GAT-based topology awareness and network navigation method is proposed. The nodes in the network learn the network topology features based on GAT to improve the navigability of the network, thereby speeding up service search and reducing overall communication costs.
The remainder of this paper is organized as follows: we first conclude the related work in Section 2 and introduce the preliminary knowledge in Section 3. Then the SBERT and GAT based service search and matching method is described in Section 4, and the dataset and simulation results are presented in Section 5. Finally, we conclude the paper in Section 6.
Before going further into different sections, we give the full forms of the abbreviations that will be used throughout the paper in Table 1.
Section snippets
Related work
Since service discovery in distributed networks includes route search and service matching, we summarize the related work from the two perspectives search respectively.
Sentence BERT
SBERT [7] is a fine-tuned model of BERT(Bidirectional Encoder Representations from Transformers) that could derive the deep semantic features of sentences. BERT is a well performed NLP model that uses a cross-encoder to compute the semantic relevance of two sentences [20], which requires the two sentences to be compared to be fed into the model simultaneously. However, it is time consuming to use this model to predict the similarity of a mass of sentence pairs because of its large parameter
Service discovery based on SBERT and GAT
Heterogeneous and distributed service discovery can be applied to emergency rescue, intelligent transportation, environmental monitoring, intelligent park and other fields.
In a IoT system composed of heterogeneous devices that (i) produced by different manufacturers and in different periods, (ii) can provide different services and (iii) in scenarios where there is no fixed network infrastructure, the use of distributed automatic service discovery technology can effectively improve the
Datasets and experimental settings
The services and queries with semantic descriptions used in this experiment are from OWLS-TC4, which contains 1083 semantically annotated services from 9 domains, and 42 test queries with associated service sets.
Further, based on NetworkX [26], we generated a batch of SIoT topology graphs. In any graph , randomly distribute nodes into Euclidean space, and each node in the graph selects the nearest neighbors to establish connections, where is selected randomly and
Conclusions and future work
This paper makes an in-depth study on service discovery in heterogeneous distributed scenarios. The service discovery method based on SBERT and GAT is proposed, called BGSD. Manufacturers of IoT devices only need to describe the capabilities and requirements of IoT devices in natural language, and then based on BGSD, IoT nodes can utilize semantic information to efficiently search services in the Ad Hoc network and accurately determine whether the service provider meets the query. Natural
Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Hanqiang Deng received the Bachelor’s degree in mechanical engineering and automation from the School of Mechanical Engineering and Automation, Beijing University of Aeronautics and Astronautics, Beijing, China, in 2012, and the Master’s degree in control science and engineering from the College of Mechatronics and Automation, National University of Defense Technology (NUDT), Hunan, China, in 2018. He is currently a Ph.D. student at the College of Intelligence Science and Technology, NUDT,
References (34)
- et al.
The social internet of things (sIoT) – when social networks meet the internet of things: Concept, architecture and network characterization
Comput. Netw.
(2012) - et al.
An efficient social-like semantic-aware service discovery mechanism for large-scale internet of things
Comput. Netw.
(2019) - et al.
Semantic web service discovery using natural language processing techniques
Expert Syst. Appl.
(2013) - et al.
Residual attention graph convolutional network for web services classification
Neurocomputing (Amsterdam)
(2021) - et al.
Graph neural networks: A review of methods and applications
AI Open
(2020) - et al.
Performance analysis of AODV, DSR, OLSR and DSDV routing protocols using NS2 simulator
Procedia Eng.
(2012) - et al.
Interoperability in internet of things: Taxonomies and open challenges
Mob. Netw. Appl.
(2019) An adaptive edge router enabling internet of things
IEEE Internet Things J.
(2016)- et al.
Toward the internet of things for physical internet: Perspectives and challenges
IEEE Internet Things J.
(2020) - et al.
Middleware for internet of things: A survey
IEEE Internet Things J.
(2016)
Semantics mining and indexing-based rapid web services discovery framework
IEEE Trans. Serv. Comput.
Research on LBS Service-oriented Architecture Based on Semantic Web and Its Service Discovery Algorithms
The semantic web: A new form of web content that is meaningful to computers will unleash a revolution of new possibilities
Sci. Am.
Cited by (1)
Meet User's Service Requirements in Smart Cities Using Recurrent Neural Networks and Optimization Algorithm
2023, IEEE Internet of Things Journal
Hanqiang Deng received the Bachelor’s degree in mechanical engineering and automation from the School of Mechanical Engineering and Automation, Beijing University of Aeronautics and Astronautics, Beijing, China, in 2012, and the Master’s degree in control science and engineering from the College of Mechatronics and Automation, National University of Defense Technology (NUDT), Hunan, China, in 2018. He is currently a Ph.D. student at the College of Intelligence Science and Technology, NUDT, China. His current research interests include Ad Hoc networks and heterogeneous IoT.
Jian Huang received the B.S. degree in control science and engineering from the Department of Automatic Control, NUDT, Hunan, in 1994, where she received the Master’s degree and Ph.D. degree in control science and engineering from the College of Mechatronics and Automation, NUDT, in 1997 and 2000, respectively. She now is a professor at the College of Intelligence Science and Technology, NUDT. Her current research interests include system simulation and artificial intelligence.
Quan Liu received the B.Eng. degree from Central South University, Changsha, China, in 2007, the M.Eng. degree from the National University of Defense Technology (NUDT), Changsha, in 2009, and the Ph.D. degree with the College of Computer Science, NUDT, in 2014. His research interests are distributed coordination, medium-access control protocol design, and cognitive networking in cognitive radio networks.
Cong Zhou received the bachelor’s degree in automation from Xiangtan University, Hunan, China, in 2014, and the master’s degree in control science and engineering from Xiangtan University in 2017. She is currently a Ph.D. student at the College of Intelligence Science and Technology, NUDT, China. Her research interests include knowledge graph and knowledge reasoning.
Jialong Gao was born in Hangzhou, Zhejiang, China in 1991. B.S. and M.S. degrees were received from National University of Defense Technology, Changsha, China, in 2014 and 2019, respectively. He is currently working toward the Ph.D. degree in the College of Intelligence Science and Technology, National University of Defense Technology. His research interests involve cooperative control, adaptive control, system simulation and their applications to unmanned aerial vehicle(UAV) and robotic systems.