In this chapter, an elaborated description of machine learning (ML)-based IoT system for smart and sustainable agriculture in modern perspective is presented. Idea for future perspective to advanced ML-IoT system development is emphasized, and a CNN and LightGBM-based crop recommendation system is suggested. Internet of things (IoT) is an emerging technology and dedicated platform to connect the remote systems to each other. Recently, IoT is widely adopted in smart and sustainable agriculture for environmental and crop data acquisition. The sensors data collected from IoT devices is analyzed using ML techniques for detection and further action is taken for improvement in farming. The ML-IoT solution assists farmers in deciding which state of action to be taken as per the analysis of IoT sensor devices data such as temperature, light intensity, humidity, ultraviolet range, and soil moisture and boost agriculture for sustainable goals. A comprehensive discussion is given of the present situation, applications, opportunities for study, constraints, and future issues.
Top1. Introduction
For thousands of years, farming has been a vital human endeavor that has supported civilization. Agriculture and humans are inextricably linked, agriculture has always played a vital role in human civilization. This fact contributes to the development and improvement of the conventional, ineffective, and time-consuming agricultural approaches (Earles 2005). The world is changing quickly, and new trends and technology have altered people's lifestyles. New and emerging technologies are starting to play a significant role in daily life (Obaisi et al. 2022), (Gyamfi et al. 2024).
Figure 1. Unified framework of IoT-enabled machine learning-based smart and sustainable agriculture
Optimizing agricultural operations to satisfy the demands of the world's rising population has become more crucial due to the high population growth and the resulting need for food (Obaisi et al. 2022), (Reddy and Dutta 2018), (de Pinto et al. 2020). The way we approach farming has changed dramatically in recent years due to technology breakthroughs, giving rise to a new method called smart farming (Obaisi et al. 2022), (Gyamfi et al. 2024). By incorporating technology into agricultural methods, smart farming is a cutting-edge method that helps farmers increase productivity, minimize waste, and maximize crop yields (Reddy and Dutta 2018), (Said Mohamed et al. 2021), (Bazzana, Foltz, and Zhang 2022). This method gathers information and delivers real-time insights regarding crop health, soil quality, and other important indicators using a variety of technologies (Akkem, Biswas, and Varanasi 2023), including as sensors (Ferrández-Pastor et al. 2016), drones, artificial intelligence, and the internet of things (IoT) (Navulur, Sastry, and Giri Prasad 2017). Additionally, there are several advantages to smart farming, such as higher output, lower labor costs, better crop quality (Said Mohamed et al. 2021), and agricultural strategy that is sustainable (de Pinto et al. 2020). Additionally, this method is more precise, allowing farmers to focus on trouble spots on their crops and use less fertilizer and pesticides (Figure 1) (Said Mohamed et al. 2021), (Yang et al. 2021). A concern regarding deployment of a IoT enabled framework is energy (Gyamfi et al. 2024), (Rajak et al. 2023). In IoT based systems, operation is constrained with the energy storage and power provided by battery (Ferrández-Pastor et al. 2016). Whole system gets failed as the no power or insufficient power is provided by the battery or at the condition of battery replacement (Srisruthi et al. 2017), (Sisinni et al. 2018). In multi-sensor IoT network power optimization and management are very important for uninterruptable operation (Ferrández-Pastor et al. 2016), (Quy et al. 2022). Power consumption of sensors used in agriculture can be optimized by utilizing the appropriate switching of nodes (Rezk et al. 2021) and make off or put in sleep mode the nonfunctioning devices specially transceivers (Quy et al. 2022), (Farooq et al. 2020).