Application of Artificial Intelligence and Machine Learning in Vertical Farming: A Comprehensive Review

1. Introduction

Global food production systems face a convergence of pressures that seriously threaten the long-term sustainability of agriculture. Soil fertility is declining owing to the combined effects of global warming, indiscriminate use of agrochemicals, recurrent natural disasters, and rapid urbanization [1,2,3]. Together, these factors have led to widespread soil degradation and a steady reduction in per capita arable land. Climate change is accelerating these problems, further undermining the viability of agricultural systems worldwide [4,5]. Anthropogenic carbon emissions, primarily from fossil fuel combustion, deforestation, and desertification, remain the main drivers of environmental degradation [6,7]. Simultaneously, water scarcity has become a critical bottleneck. Falling groundwater levels, inefficient irrigation practices, and widespread contamination are increasing pressure on watersheds and agricultural freshwater supplies [7]. These environmental challenges are unfolding alongside the rapid growth of the population. Projections suggest that the global population will reach 9.7–9.8 billion by 2050, requiring a sustainable increase in agricultural output of 50–60% to meet nutritional needs [8,9]. However, this expansion is severely limited by the ongoing loss of arable land. According to the Food and Agriculture Organization (FAO), per capita arable land is expected to fall below 0.20 ha by 2050, which is less than one-third of the 1970 level [8,10]. Alarmingly, arable land is disappearing at a rate of approximately 23 hectares per minute, highlighting the urgent need for better land and resource management to secure future food supplies.

Meanwhile, food production has become increasingly detached from urban consumers because of the demands of long-distance transport and distribution networks. This global food transportation system accounts for approximately 29% of total food-related greenhouse gas emissions [11]. Urban resilience, the ability of infrastructure systems to anticipate, absorb, and recover from disruptions, varies widely across regions. North America and Europe generally show the highest levels, whereas many African nations remain particularly vulnerable [12,13]. Emerging environmental stresses continue to weaken this resilience, which involves a system’s capacity to mitigate, adapt to or recover from major disturbances [14,15,16,17].

Vertical farming (VF) presents a compelling alternative to address these intertwined environmental and logistical challenges. By substantially reducing land and water use, shortening supply chains, and lowering carbon emissions, VF support the broader goals of ecological preservation [18,19,20,21]. At this pivotal moment, combining sustainable VF practices with advanced technologies, such as artificial intelligence (AI), offers genuine transformative potential [22]. Exploring the synergies between AI and VF could not only boost productivity but also help shift agriculture toward more ecologically sound and resource-efficient models [23,24]. This shift also holds promise for improving food safety and public health through better urban agricultural practices.

Urban agriculture encompasses various approaches, including greenhouse, container, rooftop, and warehouse farming [25]. Among these, VF has attracted considerable attention because of its ability to produce high-value crops with high productivity and a relatively low ecological footprint [26]. In VF, crops are grown in multiple vertically stacked layers, often in underutilized urban spaces or on rooftops under tightly controlled environmental conditions [27]. The modern concept traces back to Othmar Ruthner’s 1966 proposal for multi-story food production [28], but Dickson Despommier provided a comprehensive vision in 2009, proposing climate-controlled, multi-tiered buildings for horizontal or vertical crop growth [27,29,30]. By making efficient use of limited space and resources, VF offers a practical route to greater food security and environmental sustainability [31]. Its key benefits include year-round production of fresh produce, shorter transportation distances, reduced spoilage, and a significantly smaller carbon footprint than conventional supply chains [30].

The economic importance of VF is evident in its growing market share. Valued at approximately USD 5.1 billion in 2023, the sector is projected to reach USD 35 billion by 2032 [32]. Figure 1 shows these projections broken down by cultivation technology (hydroponics, aeroponics, and aquaponics), alongside a conceptual overview of the key components of a typical VF system, including soilless cultivation, nutrient delivery, automation, irrigation, lighting, CO 2 control, and environmental regulation. Regionally, East and Southeast Asia, along with North America (particularly China, Japan, Singapore, Republic of Korea, USA, and Canada), currently hold a dominant position, accounting for over 33% of the global market [31]. Europe and the Middle East have emerged as important players in recent years [32].

Figure 1. (a) Global VF market size projections from 2022 to 2032 segmented by cultivation technology (hydroponics, aeroponics, and aquaponics), indicating strong growth trends and technology adoption patterns. (b) Conceptual framework of a VF system showing key operational components, including structure, soilless cultivation, water–nutrient delivery, automation, irrigation, lighting, CO 2 control, and environmental regulation subsystems.

At the country level, Japan operates around 200 VF facilities producing crops such as lettuce, broccoli, tomatoes, and bell peppers, led by companies such as Spread Co., Ltd. (Kyoto) and Mirai Co., Ltd. (Gifu) [33]. China has more than 250 sites, with SANANBIO cultivating over 300 varieties of leafy greens, fruits, medicinal herbs, and flowers. Republic of Korea has built advanced infrastructure through firms such as N.Thing, NextOn, Farm 8, and ALGA Farmtech [12,33], including a prototype in Suwon that incorporates renewable energy for heating, cooling, and lighting [34]. In the United States, innovators such as 80 Acres Farms, AeroFarms, AmHydro, BrightFarms, Hort Americas, Plenty, and The Plant are at the forefront. Europe has witnessed accelerating adoption driven by consumer demand for fresh, local produce. The Netherlands leads with projects such as PlantLab, UrbanFarmers, and B-Four Agro, followed by initiatives such as Agricool in Paris and Infarm in Berlin [35]. Germany’s Food4Future (f4f) program stands out for integrating the production of algae and insects with traditional crops [36].

VF marks a significant shift in agronomic science, enabling crops to be grown in vertical systems within controlled indoor environments, such as warehouses, repurposed containers, or dedicated facilities. This approach helps address many drawbacks of conventional agriculture, including high land and water requirements, heavy reliance on agrochemicals, and substantial CO 2 emissions [37,38]. With the intensification of urbanization and the reduction in arable land, VF offers a sustainable method for the continuous production of food in proximity to urban centers. A core feature of VF is its dependence on Controlled Environment Agriculture (CEA), which allows precise management of temperature, humidity, CO 2 levels, and light to optimize plant growth [38,39,40,41,42,43]. While this level of oversight assures superior yields and uniformity, it concurrently introduces intricacies in operational oversight and data handling. Although this control yields higher productivity and uniformity, it also creates considerable operational and data management complexity. AI and machine learning (ML) have become essential tools [41].

AI and ML are reshaping numerous fields by extracting insights from large datasets, uncovering hidden patterns, and supporting autonomous decision-making with limited human input [41,44]. In VF, these technologies can enhance environmental control, crop monitoring, maintenance, and farming logistics. AI systems can process data from sensors and imaging devices to support real-time decisions that improve crop health, reduce resource use, and boost overall efficiency [44,45]. ML algorithms are particularly useful for the early detection of diseases, nutrient deficiencies, and pests, allowing timely interventions that limit yield loss. They also aid in yield prediction, harvest scheduling, and labor automation [41,43,46]. As big data analytics and edge computing become more accessible, AI and ML solutions are becoming increasingly scalable and cost-effective for farms of different sizes.

Nevertheless, the integration of AI and ML into VF remains in its early stages and faces several barriers. These include the need for large, high-quality training datasets, significant upfront investments, and a lack of standardized data interoperability platforms. Successful adoption also requires interdisciplinary expertise in agronomy and data science, which can be challenging for many operators. The present review critically evaluates the current landscape of AI and ML applications in VF, discussing the technologies used, their strengths and limitations, and the likely directions for future intelligent agricultural systems.

1.1. Significance of the Present Review

Although the existing literature has thoroughly examined the general applications and benefits of VF [6,18,19,32,40,47,48,49,50,51,52,53,54,55,56,57,58,59,60], there remains a notable gap in the comprehensive integration of modern AI and IoT technologies in the specific context of vertical farming. Many previous reviews are somewhat fragmented: some focus on ML and IoT in the broader context of precision agriculture but pay limited attention to the unique characteristics of VF [55], while others explore smart farming technologies without closely linking them to high-density vertical systems [53,61]. Isolated case studies, such as the application of Random Forest for energy optimization in a single hydroponic setup [62], rarely address issues of scalability, system integration, or commercial realities at larger scales. This scattered research landscape calls for a more focused and integrated review that specifically connects AI, ML, and IoT to VF.

Furthermore, questions persist regarding the long-term ecological and economic sustainability of VF. Its dependence on energy-intensive infrastructure, potential land use conflicts, and the need for renewable energy integration raise valid concerns regarding its overall environmental benefits [40,59,60]. Without careful planning and efficiency improvements, VF’s sustainability claims could be undermined. Sensor integration in hydroponic systems allows for the real-time monitoring and control of water levels, temperature, and humidity, whereas advanced algorithms enable responsive environmental adjustments [48,63]. Smart controls can reduce unnecessary interventions, save energy, and improve the responsiveness of the system. Some studies have reported notable savings, such as 20.4% in water and temperature management and 82.1% in lighting energy [24], but these findings are often context-specific and require broader validation across different settings.

This review seeks to bridge the knowledge gap by critically examining the integration of AI, ML, and IoT with VF. Unlike previous reviews, which generally addressed smart agriculture, the present study provides a comprehensive, VF-focused review of innovative systems designed to extend automation, enhance energy efficiency, reduce operating costs, and optimize environmentally friendly production practices.

The significant characteristics of this review paper are as follows:

• A systematic, PRISMA-guided methodology for literature selection and synthesis promotes transparency and reproducibility.

• A detailed analysis of IoT infrastructure and ML techniques tailored to VF, with attention to practical implementation, integration challenges, and real-world constraints, is presented.

• A comparative evaluation of AI/ML performance across different vertical farming setups, considering the suitability for various tasks, data conditions, and edge-computing environments.

• A critical assessment of commercial case studies, distinguishing verified results from company claims while evaluating economic viability, energy use, and operational risks.

• A structured discussion of key limitations, research gaps, and future directions is organized around four main pillars: data and interoperability, algorithmic advances, energy and life cycle assessment, and economic policy frameworks.

1.2. Outline of This Paper

The manuscript is structured thematically to guide readers from foundational ideas to advanced applications and real-world examples of the technology. Section 2 describes the review methodology. Section 3 introduces the key concepts of VF and controlled-environment agriculture. Section 4 provides a technical overview of IoT architectures and multilayer sensing, communication, and processing models in smart VF systems, highlighting the roles of sensor networks,