Smart Irrigation Systems in Agriculture: A Systematic Review

: This research aims to carry out a systematic review of the available literature about smart irrigation systems. It will be focused on systems using artiﬁcial intelligence techniques in urban and rural agriculture for soil crops to identify those that are currently being used or can be adapted to urban agriculture. To this end, a modiﬁed PRISMA 2020 method is applied, and three search equations are formulated. From those ﬁlters, and after a screening process, 170 articles are obtained. These articles are analyzed through VantagePoint, a text processing software. After this, they are taken through a detailed analysis phase in which 50 sources are selected as the most relevant to be read and analyzed by topic. Finally, the different phases of the analysis are used to draw conclusions that might be interesting for researchers working in this speciﬁc ﬁeld or for the general public interested in rural and urban agriculture and its automation.


Introduction
According to FAO projections, by 2050 the world population will be around 9.1 billion people, which will significantly increase the already high global demand for food.In the coming years, it is expected that food production will increase by 70% worldwide, and if only developing countries are considered, it could increase by up to 100% [1][2][3].Currently, 70% of the freshwater extracted from aquifers, streams, and lakes is used for crop irrigation [1,4], and that percentage would need to increase to sustain the expected rise in food production.
The depletion of land, available water, and biodiversity, combined with climate change, are already slowing the growth of agricultural productivity, leading to fears that in the future agricultural productivity will not be sufficient to meet global food demand [4][5][6].These previous factors, among others, demonstrate the urgency of improving aspects of agricultural production and enable areas that are not always used for agricultural activities, such as urban areas.
In this scenario, the availability of greater technological options becomes a driver for the transition to more intelligent agricultural systems [7], which, when implemented, help to improve the quality and quantity of food while optimizing the use of the resources necessary for its production [8,9], such as irrigation.This, together with the revival of activities such as urban agriculture, which has gained popularity in recent years and has attracted the attention of various researchers [10], may be the right way to alleviate the projected future food crisis.
Agronomy 2023, 13, 342 3 of 25 AI has different study fields, and machine learning (ML) is considered one of the most important.The ML systems allow computers to learn through algorithms that process datasets and make predictions about them [22,23], without the need to be explicitly programmed, and depending on the type of application and its complexity, these algorithms are classified into sub-fields.Within ML, the artificial neural network (ANN) sub-field stands out, being inspired by the human brain and seeking to emulate its complex functions, including in pattern generation, cognition, learning, and decision-making [24,25].ANNs are based on simplified versions of the biological neuronal network, which mainly consists of interconnected and organized processing units following a specific topology, which is regularly established by a series of nodes set out in multiple layers [25].
ANNs are usually used for the regression and classification of problems; language translation and pattern learning are typical examples of their usage [26].Deep learning (DL) is a sub-field of ML with certain similarities to the ANN method.This technique broadens the classic ML by adding complexity [18,19] to the mathematical models used for learning, thereby achieving highly accurate results, even with highly complex data.Data handling in DL algorithms involves transforming the data to enable its representation hierarchically through several abstraction layers.For example, in images, abstraction layers of colors and shapes, among others, are made for pattern identification and classification [27,28].These layers make it possible to provide accuracy and speed to model predictions, as well as to analyze complex data, such as image, video, audio, voice, and natural language data [27,29].
Fuzzy logic is also considered an AI field.This term was introduced in 1965 by Lofti Asker Zadeh as a name for a systematic way of addressing problems that human beings usually solve through common sense and experience [30].Fuzzy logic emerges as an alternative solution to problems that classic Boolean logic cannot resolve, given that this approach can only cope with variables adopting only binary values, which are usually mathematically represented by zeros and ones.When a problem requires using variables that can take a continuum of values, the result is expressions that cannot be considered entirely false or true [26].By introducing the notion of "degree" in the verification of a condition associated with a problem, it is possible that this condition takes values other than true and false.Therefore, this provides valuable flexibility for reasoning and enables one to consider inaccuracies and uncertainties linked to real physical situations [31].Fuzzy logic is classified as an AI field because it is used to model the behavior of human decisions obtained from experience [32] by using linguistic variables and IF-THEN rules, which simulate the human thought process.By using fuzzy logic formulations, the variables of interest can be classified into sets defined by fuzzy characteristics, such as very high, high, medium, low, or very low humidity or temperature.Therefore, both the system behavior to be controlled and the desired response of the control system can be linguistically described to model and design them, even if there are no traditional mathematical models in the system.On the other hand, the Internet of Things (IoT), which is interrelated in many applications with AI, has been defined as an interrelated system of computing devices, mechanical and digital machines, objects, animals, or people equipped with unique identifiers.It is a system with the ability to transfer data through a network, without the need for human-to-human or human-to-computer interaction.IoT seeks to integrate the physical world with the virtual one, using the Internet as a mean to communicate and extract information [33].
The fields of application of the IoT technologies are both large and diverse, because IoT solutions are spreading more and more to almost all areas of daily life [34].Agriculture is one of the sectors expected to be highly influenced by progress in the IoT field [35].Even now, IoT is significantly contributing to agriculture, including in irrigation and fertilizer systems, meteorological monitoring, soil monitoring, and disease and pest control, among others; it is expected to continue to influence even more areas in the future [35,36].
The search for applications that incorporate artificial intelligence into irrigation systems regarding modeling and control is the basic interest of this systematic review; as such, we do not concentrate on the type of irrigation delivery technique used (pivot, drip, flood, etc.).We are particularly interested in identifying which of these technologies have been used in urban agriculture and in what proportions, as well as to identify which technologies and methods are being used in rural agriculture that can be emulated in urban agriculture.We are also interested in identifying the variables considered in the different irrigation systems.The methodology of this systematic review is based on the PRISMA 2020 guidelines [37].
Section 2 sets out the methodology applied to the review.Section 3 presents the results and general discussion.Section 4 establishes the conclusions of the review.

Methodology
As previously stated, the methodology used in this review was the PRISMA 2020 guidelines [37].PRISMA 2020 was adopted in view of the clear guidelines it offers to ease robust systematic reviews.Therefore, this review article follows the recommendations of the guidelines, especially in the sections on the article selection method and discussion.In addition, a previous stage of the preliminary search was added to this core methodology to adequately reflect in the systematic review the authors' research focus on artificial intelligence techniques applied to irrigation control and modeling, more specifically in urban crops.
In this previous stage, a preliminary search of sources reporting irrigation systems that included some element of artificial intelligence was conducted.This preliminary search was supported by the free database Google Scholar.Among the results, the most interesting ones for the researchers were selected and read in depth.In this way it was possible to identify the leading technologies used in smart irrigation systems, which, together with the authors' background knowledge, served as the basis for a relevant proposal for the search equations to be used later in the systematic review.This review was conducted with the entire collection of the Scopus database in August of 2021.
How the final results of the systematic review are presented was also modified concerning the PRISMA 2020 methodology.Instead of analyzing the entire group of articles in a single stage, it was decided to subdivide the work into two parts.In the first part, an automatic analysis of all the articles was conducted using VantagePoint, a text mining software.In the second part, a detailed analysis and reading of the articles considered to be the most relevant according to the results of the first part was carried out.These changes were implemented due to the authors particular interest in some specific aspects of the results obtained in the review, as shown in the Results and Discussion sections.
Furthermore, according to the authors' criteria, three search equations were used to search articles, as it was intended to reveal the three specific subjects of greatest interest for the authors.These subjects had been identified in the preliminary search conducted during the previous stage.
Hereinafter, the methodology based on PRISMA 2020, which includes the proposed modifications, will be referred to as modified PRISMA 2020.
To systematically explore the content of the articles used in the review, VantagePoint was used.It is a professional-grade desktop text mining application that offers a wide set of powerful tools for scientific, technical, market, and patent refinement; analyses; and reporting [38].Through this tool, results were obtained in the form of different diagrams, tables, and images.To obtain the articles whose information was analyzed with the software, three search equations were established.The first equation focused on irrigation systems, using as terms those referring to IoT and ML and the main sub-fields, along with those referring to irrigation and agriculture (Table 1), with the objective of identifying the differences between ML algorithms in irrigation systems.The second equation included terms associated with the fuzzy logic, as well as those related to agriculture and irrigation (Table 2).The third equation used the terms of the previous equations but specified them in the context of urban agriculture (Table 3).In Tables 1-3, every search term in each column (OR) was searched with each of the terms in the other columns in the AND form.Note that the connector OR implies that the results contains at least one of the search terms, some, or all of them, while AND implies that all considered terms must be present in the results.
Irrigation "Fuzzy Logic" Agriculture "Irrigation System" "Fuzzy System" Farming "Automatic Irrigation System" "Fuzzy Model" The following were the criteria for the inclusion and exclusion process after the application of the search equations: The tittle-abs-key search (titles, abstracts, and keywords), as shown in Tables 1-3, was carried out in Scopus, and a total of 596 articles were obtained.The three search equations were applied separately, and then the results were grouped into a single dataset.It is worth noting that by the time of finishing the first version draft of this review article (in May 2022), a new search with the same equations was carried out, and this time 697 articles were found.This shows a comparative increase in publications on this topic.
From the 596 articles of the original search, the duplicate ones were eliminated.The resulting set of articles was screened using the inclusion and exclusion criteria explained above.As a result, 170 articles were obtained, for which a database [39] was automatically created using the free Scopus tools.The created database [39] has the following fields: "title", "abstract", "keywords", "year of publication", "names of the authors", "affiliations", and "title of source".To the database [39] obtained from Scopus, three additional fields were added: "agriculture type", "technologies", and "scope."These fields were created based on what was identified in the "title", "abstract", and "keywords" of all articles and the results of the reading during the preliminary stage.
The purpose of the database [39] was to gather and organize the large amount of information that was obtained from the articles found in relation to the searches made and process it automatically using the VantagePoint natural language processing software.The only objective was not to obtain conclusions from the software results, but to conduct a detailed analysis of the articles considered most relevant by the authors regarding smart irrigation systems.
The articles analyzed in depth were intended to have a variety of technologies applied directly to the irrigation systems.Therefore, the analysis of the software allowed the trends and uses of technologies to be identified in advance to assess the relevance of each article.The above is the next stage in the systematic review process, which enabled us to deepen the analysis of the subjects of greatest interest, namely urban agriculture, IoT, ML and fuzzy logic.In total, 50 articles were read in detail.The entire process of inclusion of articles for the review is schematized in the PRISMA 2020 flowchart (Figure 1).

Results and Discussion
This section of the results and the later discussion is presented in two parts.The first part focuses on the analysis of the total group of 170 articles obtained using the VantagePoint software.The results are presented in the form of diagrams, tables, and images, which act as a source for the analysis.The results of the analysis are then described in detail for the 50 articles selected as being of greatest interest for the authors of this review, and which may be useful for those with competing or complementary interests.

Results and Discussion
This section of the results and the later discussion is presented in two parts.The first part focuses on the analysis of the total group of 170 articles obtained using the VantagePoint software.The results are presented in the form of diagrams, tables, and images, which act as a source for the analysis.The results of the analysis are then described in detail for the 50 articles selected as being of greatest interest for the authors of this review, and which may be useful for those with competing or complementary interests.

Results Obtained Using VantagePoint
Figure 2 shows the behavior regarding the number of publications on smart irrigation systems over time.It can be noted how, since 2001, publications began to appear in limited quantities.In 2015, the publications began to increase significantly, and by 2020, a total of 50 publications on smart irrigation were detected.The publications in 2021 already totaled 32 by the time the article selection was carried out (August of that year).This showed a significant increase in interest in the last years that raises the possibility of applying smart irrigation systems for irrigation.

Results Obtained Using VantagePoint
Figure 2 shows the behavior regarding the number of publications on smart irrigation systems over time.It can be noted how, since 2001, publications began to appear in limited quantities.In 2015, the publications began to increase significantly, and by 2020, a total of 50 publications on smart irrigation were detected.The publications in 2021 already totaled 32 by the time the article selection was carried out (August of that year).This showed a significant increase in interest in the last years that raises the possibility of applying smart irrigation systems for irrigation.At the time of concluding the first version of this review article, a new search was conducted using the same equations, which provided 101 more publications than those obtained in the search performed in August 2021.It is likely that if these articles were screened, the number of publications in 2021 would be more than the 32 previously identified and the number of publications in 2022 would continue increasing, showing the growing interest in the subject of smart irrigation systems.
From the data provided by VantagePoint, as shown in Figure 3, the country with the greatest interest in the subject is India (71 publications), shown in red.It surpasses by far the next country, because India produced 51% of all publications during the studied period.India is followed, in order, by the USA (12 publications), Indonesia (9 publications), Brazil (8 publications), and China (8 publications), all marked in orange in Figure 3.At the time of concluding the first version of this review article, a new search was conducted using the same equations, which provided 101 more publications than those obtained in the search performed in August 2021.It is likely that if these articles were screened, the number of publications in 2021 would be more than the 32 previously identified and the number of publications in 2022 would continue increasing, showing the growing interest in the subject of smart irrigation systems.
From the data provided by VantagePoint, as shown in Figure 3, the country with the greatest interest in the subject is India (71 publications), shown in red.It surpasses by far the next country, because India produced 51% of all publications during the studied period.India is followed, in order, by the USA (12 publications), Indonesia (9 publications), Brazil (8 publications), and China (8 publications), all marked in orange in Figure 3.A particular capability of VantagePoint is that in addition to classifying the keywords of the articles, it allows the most used phrases to be identified.Figure 4 shows a cloud of words composed of sentences that are considered a significant combination of concurrent words that are repeated in the articles of Scopus.The largest elements in the figure are those that occur most frequently.For the analysis, the sentences that are not part of the search equations but are relevant in smart irrigation systems were considered, as follows: "neuro-fuzzy system," "real time," "water usage," "water resources," "climate change," and "low cost".These sentences served as criteria to select articles for detailed analysis.The hardware and software programs used to design, simulate, or implement the reported smart irrigation systems were also identified through the analysis of phrases and keywords.The relevant terms most frequently appearing in the articles were "Arduino", "ESP32", "Raspberry Pi," and "Zigbee", all of which were embedded plates in terms of hardware and software, as well as "MATLAB", "MQTT," and "WSN" (wireless sensor network).
Another essential element identified in the analysis of sentences and keywords was related to the variables most commonly used in the control or characterization of irrigation systems.The most common variables identified in the articles processed by Vantage-Point were "air humidity," "relative humidity," "soil temperature," and "air temperature," and the most recurrent of them was "soil moisture", which was found in relation to all those mentioned before.Therefore, this selection showed the most important variables to consider in the control of irrigation systems.A particular capability of VantagePoint is that in addition to classifying the keywords of the articles, it allows the most used phrases to be identified.Figure 4 shows a cloud of words composed of sentences that are considered a significant combination of concurrent words that are repeated in the articles of Scopus.The largest elements in the figure are those that occur most frequently.For the analysis, the sentences that are not part of the search equations but are relevant in smart irrigation systems were considered, as follows: "neuro-fuzzy system", "real time", "water usage", "water resources", "climate change", and "low cost".These sentences served as criteria to select articles for detailed analysis.A particular capability of VantagePoint is that in addition to classifying the keywords of the articles, it allows the most used phrases to be identified.Figure 4 shows a cloud of words composed of sentences that are considered a significant combination of concurrent words that are repeated in the articles of Scopus.The largest elements in the figure are those that occur most frequently.For the analysis, the sentences that are not part of the search equations but are relevant in smart irrigation systems were considered, as follows: "neuro-fuzzy system," "real time," "water usage," "water resources," "climate change," and "low cost".These sentences served as criteria to select articles for detailed analysis.The hardware and software programs used to design, simulate, or implement the reported smart irrigation systems were also identified through the analysis of phrases and keywords.The relevant terms most frequently appearing in the articles were "Arduino", "ESP32", "Raspberry Pi," and "Zigbee", all of which were embedded plates in terms of hardware and software, as well as "MATLAB", "MQTT," and "WSN" (wireless sensor network).
Another essential element identified in the analysis of sentences and keywords was related to the variables most commonly used in the control or characterization of irrigation systems.The most common variables identified in the articles processed by VantagePoint were "air humidity," "relative humidity," "soil temperature," and "air temperature," and the most recurrent of them was "soil moisture", which was found in relation to all those mentioned before.Therefore, this selection showed the most important variables to consider in the control of irrigation systems.
Figure 5 shows a nodal relationship diagram.In this type of diagram, each of the topics of interest is represented by a color and the relationships between publications The hardware and software programs used to design, simulate, or implement the reported smart irrigation systems were also identified through the analysis of phrases and keywords.The relevant terms most frequently appearing in the articles were "Arduino", "ESP32", "Raspberry Pi", and "Zigbee", all of which were embedded plates in terms of hardware and software, as well as "MATLAB", "MQTT", and "WSN" (wireless sensor network).
Another essential element identified in the analysis of sentences and keywords was related to the variables most commonly used in the control or characterization of irrigation systems.The most common variables identified in the articles processed by VantagePoint were "air humidity", "relative humidity", "soil temperature", and "air temperature", and the most recurrent of them was "soil moisture", which was found in relation to all those mentioned before.Therefore, this selection showed the most important variables to consider in the control of irrigation systems.
Figure 5 shows a nodal relationship diagram.In this type of diagram, each of the topics of interest is represented by a color and the relationships between publications dealing with joint topics are represented as nodes.The yellow dots within the nodes represent the numbers of common publications.Unconnected colored circles represent those publications that do not relate to topics other than their own.The diagram in Figure 5 shows a representation of the correlations between the most relevant technologies associated with artificial intelligence and the control or characterization of irrigation systems.This is of interest to identify the technological trends and possibilities for synergic combinations of technologies that lead to new solution proposals for irrigation systems.
From the results shown in Figure 5, a very close relationship between the systems with fuzzy logic and neural networks can be identified.This relation is coherent with what was previously identified in terms of the recurrent appearance of the phrase "neuro-fuzzy System" shown in Figure 4.A strong relationship was also identified between fuzzy logic and IoT, which was evident even when the implemented search equation did not predetermine coexistence (Table 2).
Figure 6 presents a comparative diagram between the results of the "agriculture type" field from the search database [39].As explained in the methodology section, this field was one of those added to the database [39] automatically created by Scopus.
In this diagram, both in the columns and in the rows, the different alternatives of the "agriculture type" are presented.A circle with a number inside represents the intersection of two alternatives and determines the number of times these two appear in the same publication.The junction of an alternative with itself is the total of publications in which it appears.When comparing the proposed greenhouse and outdoor systems (nongreenhouse in Figure 6), the outdoor systems are the majority with 51 publications, as compared to 15 for greenhouses.The diagram in Figure 5 shows a representation of the correlations between the most relevant technologies associated with artificial intelligence and the control or characterization of irrigation systems.This is of interest to identify the technological trends and possibilities for synergic combinations of technologies that lead to new solution proposals for irrigation systems.
From the results shown in Figure 5, a very close relationship between the systems with fuzzy logic and neural networks can be identified.This relation is coherent with what was previously identified in terms of the recurrent appearance of the phrase "neurofuzzy System" shown in Figure 4.A strong relationship was also identified between fuzzy logic and IoT, which was evident even when the implemented search equation did not predetermine coexistence (Table 2).
Figure 6 presents a comparative diagram between the results of the "agriculture type" field from the search database [39].As explained in the methodology section, this field was one of those added to the database [39] automatically created by Scopus.The systems implemented in pots, plant baskets, or crops in experimental small zones are considered "small-scale irrigation systems", and "large-scale crops" are those marketscale systems without discrimination as to whether they are outdoors or in greenhouses.Keeping this in mind, it can be observed that the reported urban agriculture irrigation systems are mostly small (13 publications), as compared to one single publication on large-scale crops.
If we compare the publications on urban agriculture irrigation systems with those on rural agriculture, most of them are rural agriculture studies (136 publications).In contrast, 28 publications are urban agriculture studies.
It is worth noting that organic agriculture was one of the alternatives considered in the review, which we were interested in highlighting given the appeal it represents for certain commercial and consumer sectors.Non-organic agriculture was included as a "conventional" alternative.However, the number of publications mentioning organic agriculture was very low, with only 3 publications, compared to the 162 publications on conventional agriculture.
Urban agriculture was a subject of special interest in this review; therefore, an analysis was carried out of this particular type of agriculture.Figure 6 shows that there are 28 publications related to urban agriculture out of 170. Figure 7 shows a nodal diagram in which the proportions of the two most used technologies nowadays for urban agriculture become apparent.The thickest relational line corresponds to IoT, evidencing the utility of this technology in urban agricultural environments and confirming the hypothesis of the ease of use of networked technologies within cities. Machine learning and fuzzy logic are the AI technologies most reported for use in urban agriculture environments.In this diagram, both in the columns and in the rows, the different alternatives of the "agriculture type" are presented.A circle with a number inside represents the intersection of two alternatives and determines the number of times these two appear in the same publication.The junction of an alternative with itself is the total of publications in which it appears.When comparing the proposed greenhouse and outdoor systems (non-greenhouse in Figure 6), the outdoor systems are the majority with 51 publications, as compared to 15 for greenhouses.
The systems implemented in pots, plant baskets, or crops in experimental small zones are considered "small-scale irrigation systems", and "large-scale crops" are those marketscale systems without discrimination as to whether they are outdoors or in greenhouses.Keeping this in mind, it can be observed that the reported urban agriculture irrigation systems are mostly small (13 publications), as compared to one single publication on large-scale crops.
If we compare the publications on urban agriculture irrigation systems with those on rural agriculture, most of them are rural agriculture studies (136 publications).In contrast, 28 publications are urban agriculture studies.
It is worth noting that organic agriculture was one of the alternatives considered in the review, which we were interested in highlighting given the appeal it represents for certain commercial and consumer sectors.Non-organic agriculture was included as a "conventional" alternative.However, the number of publications mentioning organic agriculture was very low, with only 3 publications, compared to the 162 publications on conventional agriculture.
Urban agriculture was a subject of special interest in this review; therefore, an analysis was carried out of this particular type of agriculture.Figure 6 shows that there are 28 publications related to urban agriculture out of 170. Figure 7 shows a nodal diagram in which the proportions of the two most used technologies nowadays for urban agriculture become apparent.The thickest relational line corresponds to IoT, evidencing the utility of this technology in urban agricultural environments and confirming the hypothesis of the ease of use of networked technologies within cities. Machine learning and fuzzy logic are the AI technologies most reported for use in urban agriculture environments.Figure 8 illustrates a pie chart showing the technologies that appear most frequently in the articles selected for the VantagePoint analysis.It can also be seen that fuzzy logic is the most mentioned technology in the articles, followed by IoT and ML.The proportion of articles discussing IoT highlights its importance as an applied and complementary technology to those technologies mentioned above.Figure 8 illustrates a pie chart showing the technologies that appear most frequently in the articles selected for the VantagePoint analysis.It can also be seen that fuzzy logic is the most mentioned technology in the articles, followed by IoT and ML.The proportion of articles discussing IoT highlights its importance as an applied and complementary technology to those technologies mentioned above.If the ML sub-fields are grouped, considering them as a single technology (ML, ANNs, and DL), it can be seen that, in comparison, the total number of articles in that group is very similar to that of the group addressing fuzzy logic.The emergence of "big data", a term that was not used in the search equations, identifies it as a complementary technology used in smart irrigation systems.

Highlights
Based on the results of the modified PRISMA 2021 core methodology and the analysis through VantagePoint, a selection of the most relevant sources was carried out, both on urban agriculture with the application of technologies (Table 4) and with technologies applied to agriculture (Table 5).These tables list the publications selected for the detailed analysis along with some highlights that allow the identification of the most relevant elements in the publications.

Highlights Urban Agriculture
Ref. [40] proposes an urban agriculture monitoring system with IoT implementations, with a chatbot system to notify the user of the different parameters of the plant.
Ref. [41] implements an irrigation system with fuzzy logic in crops in pots.The system uses IoT technology for data collection, and then applies the control process in a server using fuzzy logic.
In Ref.
[42], a flexible urban agriculture system model is proposed for different implementations focused on the use of IoT technologies and the cloud service.
Ref. [43] carries out a simulation of urban and rural agriculture crops, using data from different regions and climates in the United States.The studies are simulated with traditional and smart irrigation methods.In Ref. [44], an implementation of a smart irrigation system using Arduino, Raspberry Pi, and Node-RED is proposed.
In Ref.
[45], a system for indoor farming that is controlled by a fuzzy logic system for irrigation and light is proposed.
Ref. [46] proposes an analysis of which controllers and sensors are most relevant and efficient for an urban agriculture irrigation system.
In Ref. [47], the authors implement a smart system in urban farming, where it is used to calculate the best time of day for irrigation.Four different ML algorithms are applied to find the best performing one.
Ref. [48] proposes a system for indoor agriculture in which the water requirements of a crop are intended to be calculated using DL algorithms.
Table 5.The highlights of the different technologies.

IoT
Refs. [49][50][51] present automatic irrigation systems using IoT and fuzzy logic as the basis for smart irrigation.
Ref. [55] proposes an anti-frost irrigation system with IoT and an ANFIS (adaptive neuro fuzzy inference) model.The neural networks are responsible for predicting the internal temperature of the greenhouse, while the fuzzy logic system is responsible for the actuator control.
Refs. [56][57][58] take advantage of the ease of transmitting information through IoT and the ability to manage large amounts of data in the cloud.
Refs. [59,60] propose systems that use neural network technologies with IoT for the assembly of a smart irrigation system.

Machine Learning
Refs. [61,62] use ML algorithms to produce an alert based on the crop data to warn the farmer to irrigate.
In Ref. [63], a neuro-fuzzy system is proposed, in which predictions of the soil moisture and its differential with the current value of the moisture, which functions as an input to a fuzzy logic system, are carried out.
Refs. [64,65] implement irrigation systems with low-cost elements that implement different communication protocols along with IoT and ML.
Refs. [66][67][68] are systems that propose neural network models to predict the soil moisture at a future time point and implement control using fuzzy logic.The climate conditions are considered for irrigation.
Ref. [69] propose a system using data fusion to obtain better values from the sensor network.It uses two ML models, one with crop soil moisture and the other with evapotranspiration.
Ref. [70] implements a system based on the impact of plant water stress in the short-term, using ML technology for image processing, and according to this predicts the appropriate times for irrigation.
Refs. [71,72] propose irrigation systems in which DL algorithms participate.In both cases, the long short-term memory network (LSTM) is used.
Refs. [73][74][75] propose irrigation systems using ML algorithms that use data obtained from crops, together with environmental data, to predict the soil moisture.

Fuzzy Logic
Ref. [76] applies the concept of an intelligent agent to optimize the values of a sensor network for an irrigation system using fuzzy logic.
Ref. [77] applies a fuzzy logic system that has the soil moisture and CWSI (crop water stress index) as input variables.The latter is obtained from a plant temperature calculation using an infrared temperature sensor.
Refs. [78][79][80][81][82] are systems in which controllers are developed based on fuzzy logic with IoT to calculate the irrigation needs of crops.
Ref. [83] proposes a system that uses a fuzzy logic system to determine the weather conditions.This and the soil moisture are used for deciding the amount of water in a crop.
Refs. [84][85][86][87] are publications that propose ANFIS models or variants and the optimization of this type of systems to predict the evapotranspiration (ETO).
Refs. [88,89] propose the use of machine vision and fuzzy logic approaches to calculate the water requirements of the crop.

Irrigation Techniques and Fertigation
Figure 9 shows the irrigation techniques identified in the publications analyzed at this stage, where N/A represents the publications in which it is not clear which irrigation technique is used and the publications in which modeling or simulations are proposed.Hose refers to irrigation using this element in papers where simple prototypes are presented.
As shown in Figure 9, most of the publications do not make clear which irrigation technique is used, only mentioning the operation of solenoid valves or water pumps.
Table 6 shows the publications by type of irrigation technique.The most used are hose, drip, and sprinkler.The particular cases included [82], where the control of an irrigation gate was applied, and [88], where irrigation was performed by spraying.

Irrigation Techniques and Fertigation
Figure 9 shows the irrigation techniques identified in the publications analyzed at this stage, where N/A represents the publications in which it is not clear which irrigation technique is used and the publications in which modeling or simulations are proposed.Hose refers to irrigation using this element in papers where simple prototypes are presented.Intelligent systems that apply both irrigation and fertilization are of great utility and interest.Those identified at this stage of the review were [51,53,78].

Urban Agriculture
Even though the analysis in VantagePoint allowed the coexistence of urban agriculture with ML and DL technologies to be detected in several articles, as shown in Figure 6, it is interesting to note that only two publications with these types of technologies in irrigation systems for soil crops are reported [47,48].
It can be seen that ML and DL technologies are being applied preferably in aquaponic and hydroponic systems.Moreover, in those publications with irrigation systems for soil and where these technologies are mentioned, such technologies are being used for disease identification and growth stages and not for irrigation.This could mean the identification of an interesting niche, because even though aquaponic and hydroponic systems are promising and widely studied methodologies that can be rigorously controlled, soil cultivation is a method that allows an easier and cheaper approach for many urban growers that could be positively impacted using adequate technologies, and it could be interesting for researchers to work to overcome the inherent challenges of this farming method.
In urban agriculture systems, there is no standard type of crop assembly; they can be indoor or outdoor assemblies, rooftop crops, crops in baskets, conventional ground crops in the urban area, hydroponic crops, aquaponic crops, or vertical crops.In [42], these particularities and the differences between urban and rural agriculture are addressed.An example is the crop area; in rural agriculture, the crops cover an average of 7000 m 2 , while in urban agriculture, the cultivation areas can range from 9 m 2 to 1000 m 2 in wide public areas.Among the papers selected for the detailed analysis, in [40,41,47,48], physical assemblies are proposed.The others propose the use of modeling [42,45], simulations of the irrigation systems [43], or experimental assemblies of the hardware elements of the system, but without implementation in soil [44,46].
Moreover, from the review, we found that several papers do not consider the final implementation for reporting the technologies applied to irrigation systems.Instead, the papers merely describe, simulate, and implement the proposed systems in laboratory-scale prototypes.In the listed publications, the software and hardware assemblies are described, as well as the necessary tools for implementing the irrigation system.However, they do not go into the final results of the operational implementation of these systems.
On the other hand, the hardware elements used in the implementation of irrigation systems of urban agriculture are of special interest, as they serve as the basis for guiding the approach to practical implementations and future research.Several studies [44,[46][47][48] present proposals of systems with all their components specified, from the necessary sensors to the communication protocols.
One study [44] presents an assembly for an automatic irrigation system that uses elements such as Raspberry Pi, Arduino, an FPGA (field-programmable gate), a DHT11 humidity and temperature sensor, a light-dependent resistor (LDR), and a soil moisture sensor.In [47], a more complex system that uses an ESP32 is implemented, along with a lowcost Arduino module used for IoT applications as a microcontroller, a SI7021 temperature and humidity sensor, a DS18B20 temperature sensor, and a capacitive soil humidity sensor.In this paper, an RFM95 LoRa module is also used for exchanging messages between the sensors and the microcontroller.
Another study [46] proposes a simpler prototype of irrigation system assembly, using an ESP8266, an Arduino module very similar to the ESP32, along with a soil moisture sensor and a DHT11 (temperature and humidity sensor).Similarly, in [48], an assembly using an ESP32, a FC28 soil moisture sensor, and a DHT11, additional to the Raspberry Pi for the processing of ML algorithms, is implemented.
Regarding the software, [44] reports the use of the cloud-based visual programming tool Node-RED to program the system control.In [40], a chatbot application is created in LINE to ease crop data collection and the delivery of information to the user.In [45], a MATLAB tool is used for implementing a control with fuzzy logic.
The research in [47] uses the MQTT protocol to transmit information from the sensors to a server in the cloud and implements a mobile application to monitor and control certain aspects of the crop.Another study [48] develops an application for Android where it is possible to receive alerts and notifications on the crop status.Other studies [47,48] use Python for training the ML models.
The most recurrent technology applied to urban agriculture in the articles selected for the detailed analysis is IoT [40][41][42]44,[46][47][48], confirming the results of the VantagePoint analysis section.The use of this technology in urban agriculture systems is eased due to the greater possibility and reliability of the Internet connection in the urban area, providing a better connection compared to what is traditionally available in rural areas.Some of the analyzed publications show the advantages of IoT over other architectures.For example, [46] compares M2M (machine-to-machine) and IoT and concludes that the IoT technology, in the right environment, is far superior to the M2M technology.This is due to features such as the scalability, lower price, and simplicity of the algorithms in terms of their implementation.
Together with IoT, other technologies emerge in the publications.For example, [41] also implements fuzzy logic to control the irrigation system, while [40], in addition to using fuzzy logic, implements a chatbot in LINE.In [47,48], the authors use ML algorithms to predict the best irrigation timing and soil moisture level.
However, according to the review, and despite the advantages already pointed out, regarding the ease of access to communication tools and other public services in cities, the publications on smart irrigation systems, specifically applied to urban agriculture, are not very abundant, at least in systems that focus on the irrigation of soil crops, on which this review has been centered.
Only a few publications describe the use of fuzzy logic [41,45], and only two publications use ML technology [47,48].No irrigation systems that use machine vision to make decisions on irrigation were found.However, this technology is reported in articles focused on disease identification or crop growth monitoring [90].
This relatively low number of publications on smart irrigation systems in urban agriculture in soil crops means that there is an opportunity for future researchers to propose solutions for irrigation in such systems.One possible alternative is to adapt irrigation system solutions that have already been tested in rural agriculture in urban agriculture.

Internet of Things (IoT)
IoT technology is highly applicable to agricultural systems as it allows physical elements such as moisture sensors, pots, irrigation valves, and plants, among others, to be transformed into online objects on the Internet, represented by unique identifiers or tags.In this way, such elements can be monitored or controlled on the Internet, enabling the remote control of a crop and easing tasks that typically require the worker's physical presence, such as irrigation, fertilization, and visits to check the crop's status.
IoT is a disruptive technology in many sectors, including agriculture [35].The implementation of IoT in irrigation systems is combined with many technologies due to the benefits involved in representing physical elements in the form of data, as well as the ease of obtaining and collecting data from sensors and transmitting them either to the cloud [56][57][58] or between embedded systems.Embedded systems are computer systems that perform a specific task within a machine or a more extensive electric system.Data transmission is also possible through that use of low-cost boards such as Arduino and Raspberry Pi [49,53,54,60] for subsequent data processing or monitoring and for controlling irrigation systems.
The combinations of IoT with other technologies for irrigation applications are very diverse.For example, [49] uses an IoT system with the real-time monitoring of variables, which is implemented in a mobile application.In addition, a controller for an automated irrigation system that uses fuzzy logic is designed and developed.Likewise, [50,51] implement smart irrigation systems that jointly use IoT and fuzzy logic, with the difference being that the latter also performs the fertilization task.
In [11], an anti-frost irrigation system with IoT and an ANFIS (adaptive neuro-fuzzy inference system) is used.In this publication, the neural network model is responsible for predicting the internal temperature of the greenhouse, and the fuzzy logic system is responsible for activating the irrigation at the right time to prevent crops from freezing.Another study [59] applies a system using IoT and a CNN (convolutional neural network).This system uses machine vision technologies to analyze the physical state of the plant, and together with soil moisture, temperature, and relative humidity data, decides on the right time for irrigation.
In turn, [52] proposes the use of ML and IoT to implement an irrigation system that considers the variables of soil moisture, temperature, relative humidity, and pH for its operation.Moreover, the system provides a prediction of which crops can be planted according to the soil and weather conditions to avoid preharvest losses.Finally, this article shows the ease with which IoT enables the monitoring and recording of data from the sensors for later control.
In [56], a network of sensors that uses IoT and big data and generates a large volume of data which grows exponentially with time is implemented, requiring non-traditional computer processing applications to properly deal with the data.For example, it is used to irrigate an open field crop and to compare three ML algorithms to predict soil moisture.In these models, a fuzzy logic system is responsible for controlling irrigation.In a similar way, [53] uses an ML algorithm for the crop irrigation system.Additionally, an IoT platform is used to connect the system's physical devices to a mobile application to visualize the data of interest.

Machine Learning (ML)
As mentioned in the introduction, ML is useful in agriculture because it allows computers to learn from available data, such as the different weather variables continuously measured by weather stations or the measurements resulting from monitoring a crop for a considerable time.ML takes advantage of these data by using them to nourish mathematical algorithms that intend to predict or classify some variable of interest.For example, the evapotranspiration value can be used to estimate the crop's required irrigation periods.
ML algorithms are very varied, and depending on their application and complexity they are classified into different sub-fields.Examples of their application in irrigation systems can be identified in the analyzed publications.For instance, [66] proposes a system based on ANN to predict soil moisture with a timeframe of one hour from the time of measurement.In addition, this system uses a feed-forward neural network algorithm (a bioinspired ranking algorithm) with optimization of its training by using gradient descent and variable learning rate gradient descent, both algorithms that solve optimization problems through first-order iterations.
Likewise, [67,68] propose similar systems, except that they use the resilient back propagation and scaled conjugate gradient optimization algorithms.Another study [63] implements another an ANN model that allows timeframe predictions of soil moisture in the next hour.Additionally, the prediction is compared with the required soil moisture, and the difference is used for irrigation control.In this case, the radial basis function model is used; it is an ANN algorithm used to align functions.
Moreover, [65] implements a system that seeks to predict the irrigation timing using the K-nearest neighbor (KNN) ranking algorithm.This is a simple algorithm that uses the initial data associated with four soil conditions (from dry to humid) to compare them with new data from the system's sensors and to make irrigation decisions.
Another study [62] proposes a system that creates an alert for the irrigation time through an ML decision tree algorithm.This is a method inspired by the tree structure.Each decision path starts at a root node, which represents a sequence of data divisions, until it reaches a Boolean result following different branches.Another study [64] proposes a system using a conventional neural network algorithm that chooses the proper time for irrigation after obtaining data from different sensors.
Likewise, [73] proposes a smart irrigation system that searches for the best time of day for irrigation.In this case, a random forest algorithm is used, which groups several independent decision trees developed for data classification and regression purposes.Finally, [69] implements a model in which a refinement process is applied to the data using data fusion.A combination of multiple sources is used for optimized data quality.These data then feed the DL support vector machine algorithm developed for classification, regression, and outlier detection of the data.
Another study [74] implements an irrigation system that predicts the soil moisture through a combination of support vector regression (an algorithm similar to support vector machine) and K-means data grouping algorithms.Another study [75] implements different ML algorithms to observe which one obtains the best results for soil moisture prediction.The gradient boosting or gradient boosting regression tree (GBRT) algorithm obtains the best results and creates a strong predictive model based on a set of weak predictive models, typically decision trees, which enables predictions of evapotranspiration.
Another study [70] uses image processing algorithms to represent the impact of water stress on the plants in the short term by comparing the physical features of the crop leaves before and after irrigation, proposing a control system that decides the best time for irrigation.
The DL long short-term memory (LTSM) algorithm is part of the irrigation system in [71,72].It can learn long-term dependencies, especially in sequence prediction problems.
Another study [71] seeks to optimize the irrigation scheme and suggest the best crop for the next rotation.In comparison, [72] seeks to predict the soil moisture value one day in advance.
In all systems presented here, there is a particular interest-the control method implemented for the irrigation valves.For [63,[66][67][68], the fuzzy logic system is the one responsible for irrigation control, which receives the values obtained from the ML models as inputs, along with other variables.
In [64,65,71,74,75], the irrigation control is managed by the algorithms receiving the predicted soil moisture values to turn the irrigation system on or off.In [72,73], irrigation control is provided by the ML system itself, while in [61,62], an alert is produced so the grower can start the irrigation.
The objectives for ML implementation in irrigation systems differ among the publications.In [62, 63,66,68,72,74], the aim is to predict the soil moisture value in a given time range.Other studies [61,62] predict the soil moisture value but seek to generate an alert for the grower, so that they can make decisions.
Likewise, [69] aims to identify crop water needs according to soil moisture and also proposes an irrigation model based on evapotranspiration.Other studies [64,65,73,75] aim to meet the water requirements of the crop using an irrigation control system based on information from different sensors.In [70], the aim is to obtain information on crop water requirements based on leaf image processing.

Fuzzy Logic
Fuzzy logic allows empirical knowledge to be transformed into control systems through fuzzy rules and linguistic variables.This approach to the control system allows the use of the grower's accumulated knowledge from years of farming to determine the variables of interest for irrigation, e.g., the proper time of the day for irrigation.Then, these variables can be used as part of the decision system in a possible smart irrigation controller.
Fuzzy logic is a widely used method in irrigation systems.Its use cases are diverse and it is commonly combined with other technologies.One study [76] uses an irrigation system based on the concept of an intelligent agent, a software entity capable of perceiving the environment and acting upon this information, which is responsible for taking measurements of the different system sensors, including the soil humidity, soil temperature, luminosity, air temperature, and rainfall data, and determining the proper time for irrigation.
Another study [78] uses both fuzzy logic and IoT for decision-making related to fertilization and crop irrigation in a greenhouse, considering the pH and electrical conductivity variables of the system's drainage water.Similarly, [77] uses soil moisture and CWSI (crop water stress index) data as inputs into a fuzzy logic system to program the smart irrigation system, which uses low-cost sensors.
Moreover, [83] proposes a fuzzy logic system that develops a fuzzy variable called the weather condition using the values of different meteorological variables obtained from the Internet.This variable is jointly used with soil moisture data to decide the amount of water required to irrigate a bean crop in Ecuador.
Another study [79] implements an irrigation system that uses soil temperature and soil humidity as input signals to the fuzzy logic system.The purpose of this system is to calculate the time duration for opening valves at the time of irrigation, which is carried out constantly.It has autonomous irrigation valves for different opening times.
Meanwhile, [82] considers soil moisture the most relevant variable for an irrigation system.Therefore, this variable and the available water levels are the two inputs to the fuzzy system.Additionally, data are stored in the FireBase (NoSQL) database and viewed in an Android application.Another study [81] implements a fuzzy logic system that also has solar energy panels for the power supply of the system elements, thereby obtaining an electrically autonomous irrigation system.
In the publications about smart irrigation systems, we identified a recurrent combination of fuzzy logic and neural networks.For example, [85] suggests using fuzzy logic as a method for calculating evapotranspiration, in which neural networks are used for the continuous training of the membership functions of the fuzzy system.This model is called the ANFIS model.
Another study [84] suggests a co-active neuro-fuzzy inference system (CANFIS).The difference between this system and the ANFIS model is that apart from being a multiple-input system, it is also a multiple-output system.Through the CANFIS model, the calculation of the evapotranspiration is carried out, and it is compared with other methods.This article concludes that using the CANFIS model provides the best results.
Similarly, [86] proposes an ANFIS model for predicting evapotranspiration, which applies a firefly optimization algorithm (ANFIS-FA) inspired by the firefly's intermittent behavior.The ANFIS-FA model obtains better results than a conventional ANFIS model.Likewise, [87] compares fifteen prediction algorithms of evapotranspiration rates, concluding that ANFIS-FA provides the best performance.
Another common combination of technologies in irrigation systems is fuzzy logic with machine vision.One study [80] suggests a fuzzy logic system for controlling the amount of water used for irrigation.In addition, it uses the K-means cluster algorithm (machine vision) for detecting diseases in the crop.
Another study [88] implements a system combining fuzzy logic with machine vision for an irrigation and weeding system.The soil surface moisture distribution area data are obtained through machine vision, and the moisture sensor is used for measuring soil moisture.Both variables are used as inputs to a fuzzy system controlling the irrigation.A system also using machine vision in combination with a neuro-fuzzy classifier is implemented in [89] to irrigate a crop of lilies.
There are two main types of fuzzy logic systems: Mamdani and Takagi-Sugeno.They differ in terms of the fuzzy inference rules used, and their applications vary among the publications.For [77,84,85], the Takagi-Sugeno method is used, whereas the Mamdani method is used in [76,78,79,82,83].
Another study [80] carries out a comparison between both methods, in which the Takagi-Sugeno method obtains slightly better results.Other articles [81,83,86] do not specify which of the two methods is implemented in the fuzzy logic system.
In the publications, two main approaches for irrigation were identified.The first one considers soil factors, and the second one includes atmospheric factors.In the first one, the crop soil conditions, i.e., the humidity and temperature of the soil, among other variables, are crucial for deciding the timing of the irrigation [76].In the second one, the room temperature and relative humidity values, among others, are crucial for calculating the evapotranspiration value [86].These measurements allow the amount of water required by the crop to compensate for its water loss due to transpiration to be calculated.
Of the publications read in detail, some use the evapotranspiration method [84][85][86][87].All four publications have something in common; they calculate the interest variable through ANFIS models.In the other publications analyzed here, the irrigation decision combines the soil moisture reading with other variables, such as the relative humidity and luminosity.
The only publications with a different approach to those previously described are [78,88,89].In the first one, the irrigation is determined using the pH values of the irrigation system's drainage water, jointly with the electrical conductivity value.In [88,89], the water requirement for the plants is identified from the crop's images.
For future research, it could be of interest to go more in depth into each of the sentences analyzed here, as well as to review in more detail other types of agriculture, such as aquaponics and hydroponics, and to analyze the interactions of intelligent irrigation systems with different irrigation techniques.Fertigation techniques should also be considered in the future.

Conclusions
In this work, a systematic review was carried out using a modification of the PRISMA 2020 approach as the methodological basis, which has shown its usefulness in the literature for highlighting particular aspects of interest for a specific research work.
The systematic review conducted here shows that the literature on smart technologies regarding the control or modeling of irrigation systems has been growing in recent years.Therefore, this is considered a rising research niche to which we can continue to contribute from many points of view, as mentioned throughout the text.
Regarding the technological aspects of the analyzed works, it became evident that embedded systems are preferred in the implementation of smart irrigation system prototypes, which use technologies considered to be of interest for this work, such as IoT, ML, ANNs, and DL.
IoT is a core technology frequently used in irrigation system alternatives involving smart control.This is because this technology allows the representation of physical objects on the Internet and data transmission among devices is made simpler.Therefore, the data collection, monitoring, and remote control processes in the proposed irrigation systems are easier.
The irrigation systems using ML, ANNs, and DL can be used in many cases to implement innovative proposals, such as machine vision methods or technologies for predicting the behavior of the variables of interest, such as the humidity.Such innovative proposals are possible given that ML, ANNs, and DL allow the handling of a considerable amount of data.
On the other hand, fuzzy logic is the other type of technology that appears multiple times in the literature related to ML, ANNs, and DL.In these cases, usually a fuzzy logic system is responsible for irrigation control.An irrigation system using ML, ANNs, or DL is a very viable option if there is a large and robust amount of data.However, if there is a need to implement an irrigation system from scratch and you have practical knowledge of the subject, a fuzzy logic system is the best option.This approach also easily allows optimization, as shown in the various publications on ANFIS models.
The irrigation systems using artificial intelligence considered in this review are very different.Some of them use basic measures of soil moisture, while some others predict soil moisture in advance.Irrigation control systems based on indirect measurements, such as evapotranspiration, were also found.The thing that all of these systems have in common is that they show that smart irrigation is better than conventional irrigation in terms of efficiency and timeliness.This reveals that the research on smart irrigation systems is crucial to continue improving agricultural processes.
A particularly relevant element for the authors of this work is the possibility of applying what was reported in the literature to agricultural systems in urban environments.In this sense, the information gathered from the papers analyzed here implies that smart irrigation systems for rural agriculture, with the necessary modifications, are highly replicable for urban agriculture (especially for crops using soil as the substrate), since cities have all of the necessary resources, such as space, water, and electricity.However, these resources usually involve significant costs or are more limited than in rural agriculture.
It is interesting to note the relatively low occurrence of reports of smart irrigation technologies in urban environments compared to those from rural areas, despite the advantages that cities have for the implementation of smart irrigation systems.This is not surprising because most of the crops are rurally grown, but it draws attention to the potential for the development of this type of system in urban agriculture.

Figure 2 .
Figure 2. The numbers of publications per year.Obtained using VantagePoint.

Figure 2 .
Figure 2. The numbers of publications per year.Obtained using VantagePoint.

Figure 3 .
Figure 3.The numbers of publications per country.Obtained using VantagePoint.

Figure 4 .
Figure 4.A cloud of sentences.Obtained using VantagePoint.

Figure 3 .
Figure 3.The numbers of publications per country.Obtained using VantagePoint.

Figure 4 .
Figure 4.A cloud of sentences.Obtained using VantagePoint.

Figure 4 .
Figure 4.A cloud of sentences.Obtained using VantagePoint.
Agronomy 2023, 13, x FOR PEER REVIEW 10 of 26 dealing with joint topics are represented as nodes.The yellow dots within the nodes represent the numbers of common publications.Unconnected colored circles represent those publications that do not relate to topics other than their own.

Figure 5 .
Figure 5.A nodal relationship diagram of the correlations among technologies.Obtained using VantagePoint.

Figure 5 .
Figure 5.A nodal relationship diagram of the correlations among technologies.Obtained using VantagePoint.

Figure 6 .
Figure 6.The relations between the types of agriculture.Obtained using VantagePoint.

Figure 6 .
Figure 6.The relations between the types of agriculture.Obtained using VantagePoint.

Figure 7 .
Figure 7.A nodal diagram of urban agriculture technologies.Obtained using VantagePoint.

Figure 8 .
Figure 8.A pie chart of urban agriculture technologies.Obtained using VantagePoint.

Figure 7 .
Figure 7.A nodal diagram of urban agriculture technologies.Obtained using VantagePoint.

Figure 8 26 Figure 7 .
Figure8illustrates a pie chart showing the technologies that appear most frequently in the articles selected for the VantagePoint analysis.It can also be seen that fuzzy logic is the most mentioned technology in the articles, followed by IoT and ML.The proportion of articles discussing IoT highlights its importance as an applied and complementary technology to those technologies mentioned above.

Figure 8 .
Figure 8.A pie chart of urban agriculture technologies.Obtained using VantagePoint.Figure 8.A pie chart of urban agriculture technologies.Obtained using VantagePoint.

Figure 8 .
Figure 8.A pie chart of urban agriculture technologies.Obtained using VantagePoint.Figure 8.A pie chart of urban agriculture technologies.Obtained using VantagePoint.

Table 3 .
The third search equation.

Table 4 .
The highlights for urban agriculture.

Table 6 .
A list of publications by irrigation method.