Abstract
Internet of Things (IoT) has facilitated the emergence of various applications which require deploying a large number of sensors over a wide geographical area. For efficient communications, sensors with similar data within a vicinity are grouped into clusters. Compared to centralized sensor clustering, a distributed scheme is more scalable, can avoid traffic congestion, and does not suffer from single-point failure. However, due to a lack of global information, distributed sensor clustering may create more clusters than a centralized scheme. In this article, we propose to use an artificial neural network (ANN) as a tool to summarize the experience of a centralized clustering scheme in the presence of global information. Then, this experience becomes a learned knowledge to be transferred to distributed decision makers, which may subsequently approximate the centralized scheme in making the clustering decision even with access to only local information. To achieve the desired performance, we derive some secondary local information from original local information, and use it as input to ANNs to compensate for the loss of global information. Evaluation results show that it is feasible in achieving at a distributed decision maker, a same clustering solution as the centralized scheme, with about 5% clustering error probability.
Original language | British English |
---|---|
Pages (from-to) | 21851-21861 |
Number of pages | 11 |
Journal | IEEE Internet of Things Journal |
Volume | 9 |
Issue number | 21 |
DOIs | |
State | Published - 1 Nov 2022 |
Keywords
- Artificial intelligence
- artificial neural network (ANN)
- industrial sensor network
- Internet of Things (IoT)
- sensor clustering