AI-driven hybrid framework for enhanced pest detection and resource optimization using graph networks and deep reinforcement learning.

Farming is challenging due to climatic changes, insects, and inefficient resource management. Therefore, there is a need for smarter and autonomous systems. Traditional AI face challenges in solving problems including real-time optimization of resources, adaptation to environments, and fusion of data. In this study, a new framework is introduced that combines Graph Convolutional Networks (GCN), AutoML, and Deep Reinforcement Learning (DRL) to provide a change in precision agriculture. The system enhances resource management, pest detection, and crop immunity against diseases by considering space and time information. GCNs model the interaction of agricultural fields with environmental conditions, such as soil moisture, humidity, and temperature, both spatially and temporally. They monitor stages of crop development and pests. This spatial optimization allows dynamic real-time optimization. AutoML minimizes human expertise by automatically adapting parameters and structure across different areas and situations in agriculture. DRL drives an environment-adaptive decision-making process that automatically optimizes the control of resources and pests via adaptation based on environmental feedback. Two datasets were utilized to test the framework: the IoT Smart Farm Dataset (temperature, health of crops, soil moisture) and the Precision Agriculture Crop Dataset (pest infestation, satellite images, climatic data). Stability in yield improved by 21.7% (± 2.3%) compared to baseline strategies, accuracy in crop health assessment improved by 96.8%, and accuracy in the detection of pests improved by 95.3%, according to the results. The DRL technology saved 14.2% on fertilizer use and 16.4% on water usage. The approach provides efficient and scalable solutions for short-term and long-term agricultural challenges while setting a new benchmark for AI-farming and climate change resilience.