Deep Reinforcement Learning with Fuzzy Inference System for Prediction of Crop-sowing Windows

Advanced farming techniques leverage Internet of Thing (IoT)-compatible crop monitoring to collect real-time data on crop conditions, enabling precision agriculture by optimizing resource usage and enhancing decision-making for increased crop yield and sustainability. An IoT monitoring system was instrumental in analyzing data on the correlation between crop yield, weather parameters, and insect reproduction, revealing the environmental factors conducive to increased pest breeding conditions. An Optimized Fuzzy Inference System (OFIS) was proposed for IoT-compatible crop monitoring to identify the crop-sowing window. In the beginning, a wireless sensor network was set up in the field to collect meteorological information, which was then evaluated using a Fuzzy Inference System (FIS). Within this analysis, optimization of linguistic variables was achieved through the utilization of a Multi-Objective Evolutionary Algorithm (MOEA). However, MOEA computationally expensive and require more manual parameter tuning to fine tune the linguistic variables in FIS. So in this article, Deep Reinforcement Learning (DRL) is introduced which adapt and learn complex linguistic variable optimizations with minimal manual parameter tuning, making it more versatile and potentially efficient for FIS optimization. DRL, a sub domain within machine learning, integrates Deep Neural Networks (DNN) with reinforcement learning techniques. Within the realm of DRL, an agent engages with its surroundings, makes decisions, gains rewards, and gradually refines its actions for optimization. The experimental results prove that the proposed DRL-FIS has high accuracy, True Positive (TP) rate and low False Positive (FP) rate than OFIS for identification of cropsowing window.