Intelligent Fault Diagnosis for AIT-Based Smart Farming Applications

In today’s era, the artificial intelligence of things (AIT) has revolutionized every sphere of human life. Artificial intelligence (AI) approaches can effectively manage the heterogeneity of IoT devices. The applications of AIT are seen in various verticals, including smart homes/offices, smart factories, Industrial IoT, precision agriculture, etc. In particular, AIT has been a huge asset to the farming sector. Smart farm monitoring requires a huge number of sensors deployed in the monitoring field that measure various physical parameters such as humidity, soil moisture, temperature, etc. However, the main challenge is to handle the vulnerability of sensor node failure due to any natural calamity. This leads to the malfunctioning of sensor modules, which degrades the performance of the network. Also, sensor nodes are battery-limited. It is difficult to recharge and replace the sensors frequently. This article proposes an AI-based hyperparameter-tuned least square support vector machine (HT-LS-SVM) for fault diagnosis that detects faults with high accuracy. Also, it proposes mobile sink that applies reinforcement learning (RL)-based algorithm to perform obstacle-aware path planning. This algorithm adds self-learning capability to the mobile sink, which makes the system autonomous and improves network lifetime significantly. The simulation and testbed results depict the outperformance of the proposed scheme in terms of false detection accuracy (FDA), false alarm rate (FAR), false-positive rate (FPR), F1 score, network lifetime, latency, and energy consumption.