A survey on Machine Learning Software-Defined Wireless Sensor Networks
(ML-SDWSNs): Current status and major challenges
Abstract
Wireless Sensor Networks (WSNs), which are enablers of the Internet of
Things (IoT) technology, are typically used en-masse in widely
physically distributed applications to monitor the dynamic conditions of
the environment. They collect raw sensor data that is processed
centralised. With the current traditional techniques of state-of-art
WSNs programmed for specific tasks, it is hard to react to any dynamic
change in the conditions of the environment beyond the scope of the
intended task. To solve this problem, a synergy between Software-Defined
Networking (SDN) and WSNs has been proposed. This paper aims to present
the current status of Software-Defined Wireless Sensor Network (SDWSN)
proposals and introduce the readers to the emerging research topic that
combines Machine Learning (ML) and SDWSN concepts, also called
ML-SDWSNs. ML-SDWSN grants an intelligent, centralised and
resource-aware architecture to achieve improved network performance and
solve the challenges currently found in the practical implementation of
SDWSNs. This survey provides helpful information and insights to the
scientific and industrial communities, and professional organisations
interested in SDWSNs, mainly the current state-of-art, ML techniques,
and open issues.