Several reviews have been conducted concerning the application of fuzzy methods in agriculture. A study by Sannakki and Rajpurohit (2011) reviewed early papers on the use of fuzzy methods in agriculture. The study concluded that fuzzy logic effectively identifies plant diseases in various parts of the plant using image processing and soft computing techniques, particularly in handling vague image data. The paper by Makkar (2018) reviews the concept of fuzzy logic and its application in various fields such as chemical science, medical science, agriculture, and operations research. Meanwhile, Bannerjee et al. (2018) discusses challenges in agriculture, including disease and pest infestation, improper soil treatment, inadequate drainage, and irrigation. The paper highlights three primary artificial intelligence (AI) systems used to address these challenges, including fuzzy logic systems.

The application of fuzzy methods in agriculture has recently been reviewed by De and Singh (2021), highlighting the lack of efficient knowledge-based models in the agri-supply chain domains. The review specifically covers aspects such as land suitability, production techniques, irrigation, cold storage deficiencies, transportation, waste management, environmental and sustainability issues, and drought management.

In the paper by Tomasiello and Alijani (2021), the authors discuss papers addressing decision-making in agri-food supply chains, with a focus on green supplier selection, routing problems, and the most common fuzzy decision-making techniques employed for agri-food supply chains. Each of these reviews contributes to our understanding of the implementation of fuzzy logic in agri-food supply chains. However, they do not explicitly address the achievements in integrated and coordinated data distribution throughout the entire supply chain.

All activities within the agricultural chain are intricately linked and exert a collective influence on the final outcomes of agricultural products. Therefore, it is imperative to develop a model and conduct simulations incorporating various activities and parameters that impact the agri-chain in order to achieve optimal results. This paper intends to fill this gap and complement previous studies by focusing on papers discussing the agriculture part, specifically, farm processes such as cultivation and harvesting, which mostly has the potential to influence the mixing of products or goods.

Agri-chains are the chains of activities within the agricultural sector that transfer crops from farms to consumers and transform raw commodities into marketable goods. They involve both fresh products, which retain their inherent qualities, and processed products, where raw materials are transformed into higher-value items like canned goods and desserts (Alemany et al., 2021). The primary activities in agri-chains include cultivating, harvesting, handling, processing, storing, and transporting (distributing) (Taşkıner and Bilgen, 2021), as illustrated in Fig. 1.

The agricultural process begins with crop selection, land selection, planting, and nurturing (cultivating), which includes fertilizer application and pest and disease management (Barker, 2016). Harvesting occurs at maturity, but scheduling is challenging due to restricted time frames and resource restrictions (Johnson et al., 2019), weather, and resource constraints, leading to potential losses (Kumar and Kalita, 2017). As a result, harvest yields are inherently uncertain in terms of quantity, quality, and timeliness. After harvesting, crops are delivered to preprocessing sites where they may be washed, packed, or treated according to their shelf life requirements (Paltrinieri and Staff, 2014). Semi-processed stocks then proceed to processing factories for conversion into final products with a greater added value that meets consumer demand.

The complexity of the products will affect the preprocessing and processing locations, whether they occur at the same place or in different places. For example, cocoa beans may undergo fermentation and packaging at processing sites (Saltini et al., 2013). The cocoa bean can also be technically pressed into powder or liquid cocoa, which is then stored or delivered to other factories to be utilized as raw materials to produce final products (Kamphuis, 2009), meaning that preprocessing and processing are done at different locations. The complexity of end products required a different approach to preprocessing and processing facilities. Crop quality frequently degrades during storage and shipping (Kumar and Kalita, 2017). The storage of agricultural products is critical, as quality often declines during this phase, with post-harvest waste estimated at 20–60% (Lemma et al., 2014). Sufficient transportation and inventory management are vital due to fresh products’ perishability and processing equipment’s limited capacity, as indicated in Fig. 1. The key challenges include uncertainty of weather, the effectiveness of the fertilization process, control, and resource constraints that complicate harvest planning and scheduling.

Nevertheless, ascertained high-quality products are secure and efficient in traceability while minimizing losses; an integrated approach is necessary for harvest planning, which includes harvesting planning such as scheduling, routing, and resource allocation. In particular, the complexity features of the agri-chain aformentioned are highly influenced by product mixing and perishability, yielding the importance of effective traceability from farm to consumer consumption is required. Information technology solutions such as IoT, sensors (e.g. temperature, humidity, and soil pH), networking (e.g. WSN, Long Range Wide Area Network (LoRaWAN)), and positioning systems (e.g. GIS) are essential to tackle these challenges. However, storing large datasets on cloud platforms raises security concerns (Stojkoska and Trivodaliev, 2017). Utilizing fuzzy techniques and algorithms for data processing methods and analytical systems can leverage farm management’s effectiveness. A comprehensive assessment of prior research is needed to evaluate the integrated approach and available decision-making models.

The emergence of smart technologies and other advanced machinery in agriculture has led to remarkable improvements in both the quality and quantity of products. As the global population grows, the demand for efficient food production vastly increases. However, farmers are taking on this challenge and exploring innovative methods to meet the demand for agricultural products.

The classical decision-making process helps decision-makers evaluate and choose the best course of action from a set of alternatives. These alternatives may include actions, acts, or strategies, and the decision-maker must also consider the state of nature and probability distribution to make an informed choice and utilize it to determine the most suitable course of action. Decision-making is a complex mental process that implicates problem-solving to determine a desired outcome by considering various factors.

Fuzzy logic is a highly recommended approach for decision-making when faced with situations that are uncertain, vague, imprecise or ambiguous. It provides an effective means of dealing with such complexities and can prove to be a valuable tool in a variety of contexts, especially in smart farming. Bellman and Zadeh (1970) made a significant contribution to decision-making in fuzzy environments when they introduced it in 1970. Fuzzy decision-making in smart farming improves agricultural yield and quality. It optimizes resource utilization, reduces waste, and enhances efficiency and sustainability (Erdoğan, 2022).

The MCDM methods listed in Table 1 have both advantages and disadvantages, which vary depending on the field in which they are applied. Researchers have conducted comparative studies on each MCDM method. In a survey of MCDM applications in various fields, Aruldoss et al. (2013) found that the Technique for Order Preference by Similarity to Ideal Solution (TOPSIS) method was the most widely used. In the TOPSIS methodology, the ideal solution is defined as the one that has the shortest distance from the positive ideal solution (PIS) and the farthest from the negative ideal solution (NIS). During the process of TOPSIS, the performance ratings and the criteria weights are assigned as crisp values (Chen, 2000) (sees Table 2).

Various combinations of fuzzy algorithms and TOPSIS have been used in research to support the development of smart farming. For example, Ecer et al. (2023) utilized the TOPSIS approach combined with q-rung Orthopair fuzzy numbers (q-ROFNs) to select the most suitable Unmanned Aerial Vehicles (UAVs), commonly known as Drones, based on different features such as radar, power system, and camera capabilities. Similarly, in the fishery sector, Padma et al. (2022) used two Multi-Criteria Decision-Making (MCDM) methods, namely fuzzy AHP (FAHP) and TOPSIS, to compare five types of fish varieties that are popularly consumed, and provide fishermen or farmers with a vision of the optimal types of fish species to be fished or farmed.

Fuzzy techniques can also enhance traditional Multi-Criteria Decision-Making (MCDM) methods in dynamic environments. Notably, Dhumras and Bajaj (2023) introduced a picture fuzzy soft Dombi EDAS methodology that uses multiple aggregation operators and evaluates interrelationships as input arguments for traditional Evaluation based on Distance from Average Solution (EDAS) and tested in robotic farming to address various contemporary challenges.

A study conducted by Gichamo et al. (2020) used MCDM in the agricultural sector to select a processed wastewater control system for reuse that has a positive impact on the environment. They utilized fuzzy Preference Ranking Organization Method for Enrichment Evaluations (PROMETHEE) to provide recommendations for the application of the system. However, other MCDM methods, such as TODIM, are rarely employed by researchers for development in smart farming.

The significance of this systematic review lies in its aim to identify and analyse works related to fuzzy methods for smart farming or precision agriculture. It emphasizes the importance of decision-making systems in managing uncertainties in the agricultural chain, influenced by factors such as environmental conditions, weather, and the economic climate (Foley et al., 2011; Godfray et al., 2010). The review employs a systematic literature analysis methodology to achieve its objectives (Kitchenham, 2012).

The literature analysis for this review was conducted with thoroughness and rigour. Five databases, including Scopus, Google Scholar, ACM Digital Library, Springerlink, and IEEEXplore, were utilized. The search term was carefully defined as (‘Fuzzy Smart Farming’) AND (‘Fuzzy Precision Agriculture’). The review’s time frame ranges from 2017 to 2024, ensuring a comprehensive coverage of the previous seven years. The studies evaluated in this review were published in English in periodicals, with the exception of chapters of books and summaries of events and seminars.

The Preferred Reporting Items for Systematic Reviews (PRISMA) technique was carefully followed in this study, which included three stages: identification, screening, and eligibility. In the identification stage, relevant works were gathered using specific search terms, and duplicates from various databases were removed. During the screening stage, articles were assessed for relevance, while in the eligibility stage, they were evaluated for compliance with the criteria and the quality of their results and conclusions. Relevant information was then collected for each eligible study, including authors, publication year, objectives, and descriptions of the fuzzy methods used for smart farming or precision agriculture.

In the identification stage, we obtained 830 non-duplicate entries through metadata filtration. In the screening phase, we examined the titles and abstracts, leading to the selection of 710 items for a comprehensive review. In the screening phase, we examined the titles and abstracts, leading to the selection of 710 items for a comprehensive review. After assessing the eligibility criteria, we included 90 of the 120 reviewed publications in the systematic review. The selection process can be seen in the workflow depicted in Fig. 2. The selected papers, based on PRISMA criteria (see Fig. 2), were categorized by publication year and citation count from each search engine. Clustering and relationships among articles follow the methods in Waltman et al. (2010) and Van Eck and Waltman (2017), which outline how to create clusters and assign weighting values based on total citations. Figures 3 and 4 illustrate the resulting clusters and relationships. We used VOSviewer (Perianes-Rodriguez et al., 2016) to visualize these connections, accessible through both desktop and online platforms.

The density of relationships between articles in Fig. 3 shows the total number published annually from 2017 to 2024. Fig. 4 illustrates clusters based on data from various search engine databases, with the largest circle in each cluster representing citation counts. For instance, in the Scopus cluster, the article “Fuzzy Logic Based Smart Irrigation System Using Internet of Things” by Krishnan et al. (2020) has the highest citation total at 236. Furthermore, Fig. 5a illustrates the number of articles published by year. From 2017 to 2021, IEEEXplore was the leading source, with 19 publications and a peak of seven in 2020. In 2023, Google Scholar saw a significant rise, adding 18 qualifying publications. In contrast, the ACM Digital Library produced only three publications from 2017 to 2024. Figure 6 shows the total articles published from 2017 to 2024 across five indexing platforms.

Our findings are significant, particularly when considering the total number of citations, as shown in Fig. 5b. Publications published on Scopus have a higher number of citations than IEEEXplore, reaching almost 40% of the total citations of articles. Across the board of search results that have been carried out in this research, individually, the article written by Keswani et al. (2019), published by SpringerLink, is the article that has the highest number of citations, which has been cited 120 times. Several things underlie why many other researchers cited the article. More about the substance of the discussion of each article can be seen in the Section 5.

When gathering information on the Internet, users rely on the used search engines to find what they are searching for. Different search engines cater to various needs and use distinct algorithms (Ulloa et al., 2022; The free encyclopedia, 2024). Their main function is to gather and organize information in a results table, a process called indexing. For academic searches—such as scientific journals, research articles, patents, and abstracts—specific algorithms are employed for each type of indexed work.

This study utilized five scientific search engines. When users first visit a search engine, cookies are stored by their browsers to enhance the experience. Cookies play an essential role in indexing keywords and search results (Moz.com, 2024). However, these cookies are also widely used to provide feedback to third parties who work with search engine providers, which generally produce advertisements for any product correlated with frequently used keywords by users.

Search engines track user behaviour, impacting search results. To counteract this personalization and maintain the integrity of searches, we use an anonymous browsing method for default users, as shown in Fig. 7. We also established a closed cloud environment accessible via Remote Desktop Protocol (RDP). Virtual machines provide remote desktop services using XRDP and VNC, allowing multiple RDP clients to connect simultaneously. We choose the Brave browser as one of the best browsers that support good security and privacy controls, based on a review by Jennifer Simonson. Its Security and Privacy option cleans history, cookies, and cache upon closing, ensuring consistent default search conditions.

All papers in this category are mainly concerned with cultivation operations. Fifty-six papers focus on smart farming cultivation operations such as irrigation systems, watering monitoring, fertilizer controlling, crop monitoring, soil monitoring, and crop condition monitoring, as depicted in Fig. 8a. In contrast, six papers examine harvesting operations, such as crop maturity decisions. Three papers cover both cultivating and harvesting operations, while thirty-six papers focus on processing facilities. In the following section, we will delineate numerous factors that encompass the primary endeavours of agriculture overall, encompassing cultivation operations, harvesting procedures, processing facilities, and amalgamations of these undertakings.

As agri-chains had a complex activity constrained by natural problems, such as weather conditions and resource availability, this section examines model features, including time window, resource limitations, uncertainty, and sustainability, as seen in Fig. 8b.

Regarding the modelling approach (see Fig. 9), most studies utilize fuzzy logic for decision-making. For example, Cruz et al. (2017) used fuzzy control to optimize the distribution of water and electrical resources on farms. Dimatira et al. (2016) applied fuzzy logic to assess tomato maturity, whether the tomato is matured, half-matured, or immature, based on colour, size, and shape. Herman and Surantha (2019) implemented a Mamdani fuzzy controller for managing water and nutrients in hydroponics plants, e.g. bok choy and lettuce, while Tobias et al. (2020) classified lettuce growth stages using a Mamdani fuzzy inference system. Additionally, Alaviyan et al. (2020) employed a fuzzy controller to regulate temperature, light, and soil moisture for production optimization and cost efficiency.

Nevertheless, several studies have utilized fuzzy inference systems for agricultural innovations. Munir et al. (2019) implemented a Mamdani fuzzy model in their Smart Watering System (SWS) to control irrigation. Mendes et al. (2019) created prescriptive maps for central pivot speed control using fuzzy inference. Cagri Tolga and Basar (2022) developed fuzzy MCDM methods to assess vertical farm options. Castañeda-Miranda and Castaño-Meneses (2020) employed various fuzzy systems to manage water pumps, predict greenhouse temperatures, and activate anti-frost irrigation. Saggi and Jain (2020) applied fuzzy-genetic algorithms for estimating crop coefficients and evapotranspiration in wheat and maize. Anter et al. (2019) combined the Crow Search Algorithm and Fast Fuzzy C-Means (FFCM) for identifying greenness in agricultural images and generated an accurate alternative approach based on optimization for the segmentation of green plants. Chouhan et al. (2021) used a Fuzzy Based Function Network (FBFN) with IoT to detect plant leaf diseases.

Furthermore, Bahri et al. (2020) developed a multi-agent smart farm platform using fuzzy cognitive maps for fertilization. Acharjya and Rathi (2021) applied a fuzzy rough set and real-coded genetic algorithm (RCGA) for optimal crop identification prediction. Remya (2022) utilized neuro-fuzzy inference to assess soil quality based on limited inputs like organic carbon. Cai et al. (2019) employed a fuzzy adaptive PID control algorithm for managing greenhouse parameters such as temperature and moisture. Kokkonis et al. (2017) introduced a neuro-fuzzy algorithm for controlling irrigation water valves.

This section discusses the actuators as variables that perform decision-making control. Actuators are responsible for performing specific actions based on instructions given by a control system. These variables are crucial in decision-making, allowing farmers to automate and efficiently complete tasks. The actuators should ideally take into account, i.e. data source, acquisition, controller, and effects.

Each stage of the agricultural supply chain presents unique challenges that can be effectively addressed through various approaches, particularly with the aid of computer science. In Section 4, our analysis shows that majority of articles focus on challenges during the cultivation stage, as seen in Fig. 8a, where over half highlight cultivation issues with blue lines and dots. However, there’s a lack of emphasis on harvesting and processing challenges. While six articles address harvesting operations—Deepanayaki and Vidyaathulasiraman (2024), He et al. (2024), Bahri et al. (2020), Dimatira et al. (2016), Huang et al. (2020), and Tobias et al. (2020)—they rely solely on simulations and datasets. As Damerum et al. (2020) emphasize, the quality of product is highly depending to harvesting, post-harvesting, cultivation, and processing facilities. The timing of the harvest significantly affects quality, and proper storage conditions, including temperature and humidity, are essential to mitigate damage from pests and bacteria, as indicated in Liu et al. (2022).

The features of the agricultural chain described in Section 4 are associated with the operations shown in Fig. 8b. This figure highlights the key aspects for analysis, including time windows, resource limits, uncertainty, and sustainability. Notably, 73% of the reviewed articles focus on resource limits, followed by sustainability and time windows. Most studies provide proof of concept through experiments that implement methods in prototypes or real-world applications, as shown in Fig. 8c. In contrast, only 15 out of 90 articles examined uncertainty in farming, primarily related to natural factors such as weather and soil conditions. Furthermore, smart farming encompasses the depreciation of agricultural equipment and the crucial knowledge of farmers, in addition to devices like sensors and computing systems (Chavas and Nauges, 2020).

The other two features, sustainability, and time window, are very important as well. Smart farming effectively ensures sustainable agricultural systems that are harmonized with the quality functions of crops. It consists of using IoT devices, sensors, and edge computing to validate parameters that can induce plant growth. As shown in Fig. 8c, automated irrigation and stabilizing soil pH are crucial in crop production; also, the automation of spraying pesticides and applying fertilizers is necessary. However, it is important to perform all the stages of agriculture at appropriate time frames so that quality production and customer satisfaction requirements can meet sustainability alongside nature.

Conventional farming data acquisition has been transformed into data-driven approaches since the evolution of IoT technologies has become a game changer. The upcoming advancement will incorporate more sophisticated technologies for gathering agricultural and environmental data for systematic studies. Fig. 10 indicates that the majority of articles highlight sensors as the main data acquisition tool linked to devices such as edge computing systems or microcontrollers. The data from these sensors inform control systems that implement appropriate environmental measures.

Another fact is dataset usage ranks second among researchers at 29% of total articles, followed by Wireless Sensor Networks (WSN) at 13%. Analog sensors represent the most minor portion with just one article. This trend emphasizes the strong contributions of researchers in smart farming, as they leverage various technologies to enhance agricultural quality and quantity while promoting global food safety. A wireless sensor network (WSN) is a sophisticated system comprising multiple sensors embedded in a microcontroller with a wireless communication module. In essential, WSNs are designed to gather information in remotely located areas and transmit this information wirelessly, it enable receivers to monitor remotely (Mahbub, 2020). Whereas, deploying and configuring of WSNs is a complex issue which requires huge amount energy resources.

Researchers often use datasets to investigate various dimensions of plant commodity expansion, aiming to drive innovation and improve quality. This dataset is gathered from sensors or cameras at the fields, where sowing takes place. Using datasets, researchers can simulate various scenarios to address plant issues, such as diseases or better irrigation systems.

Upon comprehensive examination of global food security shows that industrial technology 4.0 has brought significant advancements, especially in the implementation of smart farming. The interconnection of agricultural activities significantly affects production effectiveness and efficiency. Therefore, establishing a system to record, monitor, and control these activities, whether through manual or automated process —is essential.

In this paper, we present a perspective on the use of fuzzy logic and control in smart farming, primarily as a control system to improve the efficiency of agricultural irrigation. Significant advancements have been made in smart farming, which is set to revolutionize traditional practices. In the future, adopting smart farming will be more accessible, with many global companies like IBM (Gomstyn and Jonker, 2023) and Microsoft (FarmBeats, 2024), along with innovative startups, showing strong interest in this field. Smart farming involves land mapping with GIS technology, global monitoring, and analytics. The resulting feedback aids decision-making and directs control systems for various agricultural tasks. Implementing smart farming will no longer be a difficult task by subscribing to a pay-per-use smart farming platform within the integration with edge computing devices. Farmers can conveniently monitor and efficiently manage their operations through various media and learn from successful crop cultivation case studies. This potential for advancement emphasizes sustainable resource use in agriculture.

Nevertheless, our study had some limitations, primarily because we were unable to use a Multi-Criteria Decision-Making (MCDM) method to analyse the survey parameters collected and yields as we may not have captured potential variations that could have been identified through the application of the MCDM method.

To address this issue, we plant to conduct further simulations using a fuzzy linguistic model and the 2-Tuple Linguistic (2-TL) MCDM approach. This will enhance our analysis of survey parameters. Furthermore, integrating a fuzzy logic controller with deep learning algorithms will improve decision-making regarding plant health issues that are not well-explored. Some other issues arise on the sustainability of smart farming systems involving various sensors, controllers and any other computing devices that needs to be addressed. A key question is whether these systems offer an economical solution for farmers who invest in equipment and software. It’s important to consider how long these devices will function effectively, factoring in their depreciation. This approach will provide a clearer understanding of survey results and help identify overlooked variations in our study.