Context Aware Fault Tolerance Data Gathering in IoT Enabled WSN

Wireless sensor network (WSN) is utilized in various fields, including surveillance and artificial intelligence. WSN have become the key components of the Internet of Things (IoT). The communication between sensor nodes (SNs) and the base station plays a significant role in WSNs. In IoT applications, sensor devices receive data from their surroundings and sends it to the base station. However, sensors in IoT are vulnerable to faults, leading to a higher probability of failures in IoT-enabled WSNs. Faulty nodes and broken links can severely impact the reliability of IoT-enabled WSNs. This article aims to identify faulty nodes and gather data using a mobile sink located at the cluster head of each cluster. Various fault tolerance algorithms can improve network performance by enhancing reliability through multipath transmission, relay node placement, and backup node selection. However, these algorithms often face challenges such as significant data transmission delays, packet overhead, and lower detection accuracy. This work proposes a multi-objective deep reinforcement learning (DRL)-based algorithm for fault detection and tolerance in IoT-enabled WSNs. Additionally, we introduce a K means clustering algorithm as part of unsupervised learning. The objective of this research is to detect faulty nodes with high accuracy and minimal overhead. The focus is on reliable data transmission following fault detection. A mobile sink (MS) is employed for energy-efficient data gathering, which significantly improves network lifetime. Extensive simulations and theoretical analyses demonstrate that MODRL algorithm achieves outstanding results compared to contemporary systems by improving fault detection accuracy (FDA), false alarm rate (FAR), false-positive rate (FPR) in addition to enhancing network lifetime and overall throughput.