Optimization of Vegetable Production in Hydroculture Environments Using Artificial Intelligence: A Literature Review

This review analyzes the role of artificial intelligence (AI) and automation in optimizing vegetable production within hydroculture systems. Methods: Following the PRISMA methodology, this study examines research on IoT-based monitoring and AI techniques, particularly Deep Neural Networks (DNNs), K-Nearest Neighbors (KNNs), Fuzzy Logic (FL), Convolutional Neural Networks (CNNs), and Decision Trees (DTs). Additionally, Recurrent Neural Networks (RNNs) and Long Short-Term Memory (LSTM) models were analyzed due to their effectiveness in processing temporal data and improving predictive capabilities in nutrient optimization. These models have demonstrated high precision in managing key parameters such as pH, temperature, electrical conductivity, and nutrient dosing to enhance crop growth. The selection criteria focused on peer-reviewed studies from 2020 to 2024, emphasizing automation, efficiency, sustainability, and real-time monitoring. After filtering out duplicates and non-relevant papers, 72 studies from the IEEE, SCOPUS, MDPI, and Google Scholar databases were analyzed, focusing on the applicability of AI in optimizing vegetable production. Results: Among the AI models evaluated, Deep Neural Networks (DNNs) achieved 97.5% accuracy in crop growth predictions, while Fuzzy Logic (FL) demonstrated a 3% error rate in nutrient solution adjustments, ensuring reliable real-time decision-making. CNNs were the most effective for disease and pest detection, reaching a precision rate of 99.02%, contributing to reduced pesticide use and improved plant health. Random Forest (RF) and Support Vector Machines (SVMs) demonstrated up to 97.5% accuracy in optimizing water consumption and irrigation efficiency, promoting sustainable resource management. Additionally, LSTM and RNN models improved long-term predictions for nutrient absorption, optimizing hydroponic system control. Hybrid AI models integrating machine learning and deep learning techniques showed promise for enhancing system automation. Conclusion: AI-driven optimization in hydroculture improves nutrient management, water efficiency, and plant health monitoring, leading to higher yields and sustainability. Despite its benefits, challenges such as data availability, model standardization, and implementation costs persist. Future research should focus on enhancing model accessibility, interoperability, and real-world validation to expand AI adoption in smart agriculture. Furthermore, the integration of LSTM and RNN should be further explored to enhance real-time adaptability and improve the resilience of predictive models in hydroponic environments.