Geopolitical events are causing supply chain strains, and climatic events are impacting the food systems’ resilience [1]. The challenges to ending hunger and food insecurities keep growing, and the COVID-19 pandemic has further highlighted fragilities in theour agrifood systems and inequalities in the our societies [2]. This scenario becomes more urgent with the growing food demand. The Food and Agriculture Organization (FAO) has stated that by 2050 there will be around 10 billion people, and the food demand will grow by 70% [3]. Artificial intelligence (AI) techniques applied in agriculture can optimize agricultural processes by food system resilience increases.

AI is an evolving set of technologies that are used to solve a variety of applied problems and has been extensively applied in farming recently ^[4][4]. This work’s purpose is to make a systematic review of the current studies and research in agriculture that employ the recent practices of AI technologies to solve several relevant problems.

Reviews of the literature are important for synthesizing the existing knowledge base: in [5], the authors conducted a review about crop yield prediction using machine learning; in [6], advanced agricultural disease image recognition technologies were explored; IoT solutions for smart farming were researched in [7]; big data in agriculture in [8]; and agriculture 4.0 in [9,10].

2. Descriptive Analysis (RQ1)

ReThis serachersction quantitatively describes the 176 selected studies considering the publications and citations volume; the method of the studies identified; and the most influential countries, journals, and institutions. Figure 1 shows the distribution of papers and citations per year. The publications number and citations in the area have tripled over the last three years; the data collected in December 2022, reached a peak of 2440, referring to the 2020 publications. That is, the 2021 and 2022, publications were already citing the 2020 publications, showing the dynamism of feedback from this line of research. Studies were classified as either theoretical or empirical (Table 1). Theoretical studies were classified as reviews or SLRs. Additionally, empirical studies were classified as modeling and simulation, surveys, or case studies. The number of studies was balanced, with 43% classified as theoretical and 57% as empirical. Theoretical studies had an emphasis on reviews of the literature, with 59 papers representing 34%, and in between empirical studies, these stood out in modeling and simulation, with 68 papers representing 39%. The most representative articles in quality appraisal steps reflect this distribution, such as ^[11][12][13,15] which developed, respectively, robot strawberry-picky and smart irrigation systems. When studying a research paper’s relevance, bibliometric analysis can consider several indicators. It This showed that thresearch shows the volume and impact of the publications concerning citation numbers. The Netherlands is the most influential country in the is research scope, with 5 publications and 1629 citations, closely followed by India, with 37 publications and 1499 citations. Greece had only 3 publications and 1002 citations, and China had 23 publications with 899 citations. Table 2 shows the countries with more than 100 citations. Table 4 shows the ten institutions with the highest citation volume. This ranking was led by universities from the USA, The Netherlands, three universities from China, India, Chile, and two universities from Malaysia and Brazil. Finding 1: The publications and citations of artificial intelligence techniques applied to agriculture increased almost six times over the last three years, demonstrating the importance and timelinesss of this research line. The most influential countries identified were among the world’s largest food producers, and there were different Journals with high-impact factors in publishing in this field.

4. Artificial Intelligence in Agriculture (RQ2)

Agriculture, meaning land cultivation, is the science of raising livestock and producing crops. The principal resource base for agriculture is the physical environment, and the cultivated crop plant is their production unit. The challenge of agriculture is to efficiently manage the physical environment to provide for the biological demands of the crop plant ^[13][27]. The principal factors that impact crop yield are soil productivity, the accessibility of water, climate, and pests or diseases ^[14][28]. Finding 2: In the reviewed literature, rwesearchers  identified seven main applications: crop management, water management, soil management, fertigation, crop prediction, crop classification, disease, and pests. And twenty-four different artificial intelligence technics, including more big data, IoT, and cloud computation, were identified. Applications that were more frequent included crop management, water management, diseases and pests. The technics used the most were machine learning, robotics, deep learning, and the Internet of Things.

5. Benefits, Challenges and Trends (RQ3)

Table 5 shows an analysis of the selected studies in the quality assessment stage with a focus on the benefits and challenges in agriculture. Modeling and simulation papers, in general, used machine learning in the development of algorithms and systems to apply crop, water, and fertirrigation management ^{[12][15][16][17]}[14,15,20,21]. In ^{[11][18][19][20][21]}[13,16,17,18,26], reswearchers used crop classification and disease, and pest management with machine learning and computer vision. Computers and Electronics in Agriculture led the ranking with 4206 citations and 55 publications, followed by Agricultural Systems (1195), Sensors (1012), and Artificial Intelligence in Agriculture (592). Table 10 shows the Journals with more than 100 citations; those with the greatest impact factor are Computers in Industry (11.245), IEEE the Internet of Things Journal (10.238), Computers and Electronics in Agriculture (6.757), Agricultural Systems (6.765), and Information Processing in Agriculture (6.409). Robotics, deep learning, and computer vision were used in the case study paper on kiwifruit harvesting ^[22][19]. The case studies showed in ^[25][26][24,25] developed systems using machine learning and IoT for crop management. In reviews of the literature, ^[5][24][5,23] used machine learning techniques for crop yield prediction, ^[23][22] used digital twins for crop management, [8] studied big data for smart farming, and [6] used deep learning and computer vision for disease and pest recognition. In the SRL analysis, rwesearchers identified terms that used various AI technologies to optimize agricultural processes; the main terms identified were precision agriculture, agriculture 4.0, and smart farming. Precision agriculture is an approach in farm management that uses information technology (IT) to optimize resource usage and reduce environmental impacts ^[26][27][25,49]. Precision agriculture uses remote sensing approaches in the aerial monitoring of agricultural fields and provides real-time images collected from satellites, UAVs, or manned aircraft ^[28][50]. Agriculture 4.0 or Digital Agriculture is a term referring to Industry 4.0. It represents a more efficient industry that makes full use of Big Data and new technologies to benefit the entire supply chain and produce a greater and better quantity, with less, in search of increasing food supplies and reducing waste ^{[10][29][30][31][32]}[10,51,52,53,54]. Described as precision farming evolution, agriculture 4.0 uses automated collection, integration, and data analysis [9]. Next-generation agriculture 5.0 and 6.0 uses a deep training data set and technological advancements through robots that can target achieving both production and environmental goals ^[25][24]. Already the term “Smart Farming” is to the application of intelligent systems and communication technologies such as sensors, IoT, cloud-based processes, artificial intelligence, and networking in the farming system to boost farm produce ^[33][55]. Finding 3: Artificial intelligence techniques applied in the main fields of agriculture were identified, with the main benefits being the optimization of agricultural management systems, irrigation, and the identification of diseases and pests. It was observed that the increase in intelligence in agriculture could be related to the digitization and manipulation of large volumes of data, enabling the use of intelligent techniques in system optimization and planning. Computer vision was used in conjunction with robotics and Unmanned Aerial Vehicles (UAVs) for classifying crops and identifying diseases and pests. The present section provides insights into the technologies most researched. Based on the technologies identified in the 176 articles analyzed, Figure 2 shows the Top 10 most frequent technologies and terms identified. There is a relationship between the identified technologies: machine learning is the most used technique, and this technique, like deep learning and computer vision, needs data to obtain good results. RWesearchers can see in the top 10: the Internet of Things, which is capable of collecting and transmitting data; big data, a knowledge area that studies how to treat, analyze and obtain information from large data sets; and cloud computing, which is a data center that makes data available over the Internet. Drones and unmanned aerial vehicles (UAVs) can collect a huge and complex amount of data, and using big data analytics tools and cloud computing could be utilized to increase data processing efficiency, provide data security and scalability, and reduce costs ^[26][25]. Machine learning, ANN-based, and deep learning techniques hold a promising future in crop prediction due to the amount of data from varied sources ^[24][23]. In addition to the most cited technologies presented in Figure 2, the emergence of new technologies was observed in the review, including Digital Twins (DT). Precision Agriculture (PA) was the term most frequent, but new terms like Agriculture 4.0 and smart farming are gaining space in reviews of the literature. ReThisearchers  work analyzed the relevant studies on AI in agriculture. These findings identified summarize the analysis and possible future research directions stand out:

• Research needs to be adapted to the climate and crop of application regions; food-producing countries like Brazil are still not very expressive in their scientific production in the area.
• AI technologies can be applied in several areas of agriculture; it is necessary to understand the production chain of the crop analyzed to identify the best technique to be applied and its interrelationship with terms such as agriculture 4.0 and smart farming seek which can integrate these various technologies for the optimization of a production chain.
• The most applied technologies have in common digitized data needs; they are at the digital revolution heart, and, for future research, the interaction and need for technologies to enable the application and reach of results must be observed.