AI Roles in 4R Crop Pest Management—A Review

Abstract

Insect pests are a major threat to agricultural production, causing significant crop yield reductions annually. Integrated pest management (IPM) is well-studied, but its precise application in farmlands is still challenging due to variable weather, diverse insect behaviors, crop variability, and soil heterogeneity. Recent advancements in Artificial Intelligence (AI) have shown the potential to revolutionize pest management by implementing 4R pest stewardship: right pest identification, right method selection, right control timing, and right action taken. This review explores the roles of AI technologies within the 4R framework, highlighting AI models for accurate pest identification, computer vision systems for real-time monitoring, predictive analytics for optimizing control timing, and tools for selecting and applying pest control measures. Innovations in remote sensing, UAV surveillance, and IoT-enabled smart traps further strengthen pest monitoring and intervention strategies. By integrating AI into 4R pest management, this study underscores the potential of precision agriculture to develop sustainable, adaptive, and highly efficient pest control systems. Despite these advancements, challenges persist in data availability, model generalization, and economic feasibility for widespread adoption. The lack of interpretability in AI models also makes some agronomists hesitant to adopt these technologies. Future research should focus on scalable AI solutions, interdisciplinary collaborations, and real-world validation to enhance AI-driven pest management in field crops.

1. Introduction

With the global population expected to surpass 9.5 billion by 2050 [1], food demand is rising rapidly. Meeting this demand requires producing 50% more food on limited arable land [2] while overcoming the a/biotic stresses–caused crop yield losses [3,4]. Among the stresses, pest damage poses a significant threat to agri-food production, resulting in average worldwide crop yield losses of 40% [5], depending on the crop species, e.g., 26–29% in soybean, while 30–50% in potatoes [6,7]. Regional environmental conditions also influence pest-induced crop production losses. For example, the U.S. produces only about 58% of China’s wheat output, but its pest-induced yield losses are 1.32 times higher (Figure 1). The adverse impact of pests extends beyond crop losses, disrupting the delicate ecological balance of farming ecosystems. Pest outbreaks often harm beneficial organisms, leading to a further decline in biodiversity within agroecosystems [8]. Because of climate change, the outbreaks of existing pests are increasing, and they also potentially become the major ones in areas where they did not cause problems before [9]. Therefore, addressing pest-related threats is critical to achieving sustainable agriculture.

Integrated pest management (IPM) is a well-studied strategy for effective pest control in agricultural systems. It encompasses a diverse array of approaches, including physical and mechanical methods [11], biological control strategies [12], chemical interventions, genetic advancements [13], and innovative cultural practices [7,14]. However, the effective implementation of IPM on farms is a complex network, requiring the right pest identification, the right method, the right timing, and the right action—an approach we named as 4R pest management. This strategy is widely valued as an ideal framework because it requires a deep understanding of species-specific pest behaviors, life cycles, movement patterns, and their interactions with weather conditions such as temperature and precipitation [15]. These factors collectively determine the selection and application of tailored methods while considering the crop-specific economic threshold. This process demands extensive data-driven analysis, experimental insights, field experience, and predictive capabilities.

Artificial intelligence (AI) has gained significant attention in recent years, driven by its advancements, particularly the emergence of large language models such as ChatGPT (GPT-3 model) and DeepSeek (version V3) [16,17]. AI also has the potential to transform pest control by providing advanced tools to monitor pest behavior, predict outbreaks, and optimize management strategies. Specifically, AI-powered systems could utilize and integrate data from sensors, drones, and satellite imagery to detect pest populations with high precision [18]. By analyzing large datasets, AI algorithms are expected to precisely identify pest behaviors and predict pest outbreaks by linking pest activities to environmental conditions. Building on these advancements, automated equipment and technologies guided by AI models may have the capacity to precisely target pests—for example, optimizing the timing and location of pesticide applications based on pest density monitoring [19]. As agricultural landscapes grow more complex, machine learning (ML) supports applications like soil nutrient estimation [20], crop yield prediction [21], and nitrogen recommendation [22]. Therefore, AI possibly provides a scalable, adaptive solution to pest management, bridging the gap between traditional methods and modern challenges.

This narrative review aims toro synthesize recent advances in AI applications for pest management. Specifically, this study is guided by the central question based on the flowchart shown in Figure 2: how can AI technologies support the development of a 4R-based pest management framework? To explore this question, we selected articles based on their relevance to specific research questions, with emphasis on recent peer-reviewed studies that applied AI technologies to real-world pest detection, prediction, or decision-making scenarios.

2. 4R Pest Management and AI Roles

2.1. Rationale for the 4R Framework in Pest Management

In this study, we propose a “4R pest management” framework inspired by the well-established “4R nutrient stewardship” concept [23], which refers to the Right source, Right rate, Right timing, and Right place in nutrient management. Analogously, our pest management framework emphasizes the Right identification, Right method, Right timing, and Right action as a structured approach to enhance precision, sustainability, and effectiveness in pest control (Figure 3). This stewardship starts with accurate species identification, as misidentification can result in ineffective treatments or unintended ecological harm. After accurate identification, choosing the most suitable control method—whether cultural, physical, biological, chemical, or their combination—is crucial to align with the pest’s biology and the crop’s developmental stage. Timing is equally critical; applying control measures too early or too late can reduce effectiveness and increase costs. Finally, implementing the right action—defined as the precise application of control measures using appropriate dosage, tools, and delivery techniques—ensures the desired outcome while minimizing environmental impact. This step is distinct from Right timing, which refers to when the intervention should occur; Right action refers to how it should be executed. While not yet standard in pest science, this structure offers a conceptual tool to organize and evaluate AI technologies in pest management by aligning them with key decision points. This holistic approach aligns with IPM principles, optimizing pest control while preserving ecosystem health.

2.2. AI Roles in Pest Management

Accurate pest recognition and identification serve as the essential foundation for implementing precise 4R-based pest management strategies. Traditional methods rely heavily on manual inspections, expert knowledge, and visual or physical sampling, which offer valuable contextual insights and are accessible without high-tech infrastructure (Figure 4). However, these approaches can be time-consuming, labor-intensive, and subject to human error or bias, limiting scalability and real-time responsiveness. In contrast, AI-powered systems enhance efficiency by leveraging drones, sensors, and ML models for automated monitoring, pest identification, infestation assessment, and precision application of control strategies. These technologies enable rapid, data-driven decision-making and continuous monitoring, which significantly improve accuracy and responsiveness. This challenge underscores the need for automatic and accurate pest categorization systems.

In this study, we propose a foundational framework for integrating AI into the 4R pest management approach (Figure 5). Specifically, the system starts with data processing and feature engineering, followed by model training and evaluation using separate datasets for pest identification, outbreak prediction, and management decision-making. Real-time field data, such as images and environmental parameters, feed into this system, enabling automated decisions on whether an insect is harmful, its species, and whether intervention is needed based on economic thresholds. After treatment, feedback from post-management field conditions helps iteratively refine the models. This closed-loop system ensures adaptive and context-specific pest management, enhancing both efficiency and sustainability in modern agroecosystems. To ensure effective real-world pest management, comprehensive data and domain knowledge must be integrated into AI models and iteratively refined before deployment.

As illustrated in Figure 5, the proposed AI-driven system consists of two core modules: Model A, which focuses on real-time pest identification and outbreak prediction, and Model B, which guides decision-making for intervention. Right identification is addressed by Model A using deep learning (DL) to classify pest species and estimate densities based on image or acoustic data. The right method is supported by Model B, which incorporates multi-source data (e.g., pest type, severity, and environmental conditions) to recommend the most suitable control option (chemical, biological, or mechanical). The right timing is achieved by integrating environmental sensor and weather forecast data into temporal prediction models that anticipate pest emergence or outbreak thresholds. Right action is facilitated through the linkage of AI recommendations with field-deployable technologies, such as drones or smart traps, to execute precision treatments in real time. This modular, closed-loop system promotes continual learning, allowing AI models to adapt over time through pest-intervention feedback, thereby strengthening decision-making accuracy across all stages of 4R pest management.

2.3. 4R Pest Management

2.3.1. Right Pest Identification

Right pest identification refers to the precise recognition of pest species, their developmental stages, and population dynamics in a given agricultural setting. Accurate pest identification is the foundation of effective pest management, ensuring that control measures target the correct species. Misidentification is prone to lead to ineffective control, unnecessary pesticide use, and disruption of beneficial insects, such as pollinators. For instance, ladybugs are predators of aphids and other crop-damaging pests, and mistaking them for harmful species can disrupt natural pest control processes [24]. Accurate and effective pest identification requires a thorough understanding of pest behavior, activity patterns, and the surrounding agroecological environment because these factors determine how they interact with crops and respond to environmental conditions [7]. Generally, pest behaviors are closely influenced by crop traits, enabling pests to exploit host vulnerabilities while evading detection through camouflage, altered movement patterns, or feeding strategies [25]. Environmental factors such as temperature, humidity, and light also influence pest activity and distribution by modulating their life cycles and interactions with crops. These conditions, in turn, shape species-specific feeding patterns, movement strategies, and reproductive behaviors—key traits that aid in accurate pest identification. For instance, chewing pests such as potato beetles leave visible damage on plant leaves, stems, and fruits, making them easier to identify based on feeding traces [7]. In contrast, sap-feeding insects like aphids and whiteflies cause more subtle symptoms, such as leaf curling and chlorosis, requiring microscopic examination or molecular tools for accurate identification [26]. Some pests, like the rice planthopper, exhibit hiding behavior by clustering at the base of rice plants, which makes field detection challenging. Soil-dwelling larvae such as wireworms become active in early spring, targeting the roots of germinating seeds [27], whereas potato beetles emerge from the soil in spring and become active when air temperatures reach 10 °C [7]. These environment-specific and pest-unique behaviors undoubtedly complicate accurate pest identification.

Recent advances in AI present a promising opportunity to enhance pest identification through enhanced accuracy and efficiency using computer vision and ML. This is because AI-powered systems can efficiently analyze high-resolution images from online databases, accurately distinguishing harmful pests from beneficial species with impressive precision. Yang and Qiu [28] employed an enhanced YOLOv5 model for pest detection and achieved a mean average precision of 92.5% with details in

Table 1. AI technologies and their application in advancing 4R pest stewardship.

AI Technique	Tasks	Dataset	Crops	Pests	Outcome	References
YOLOv5s-pest (YOLOv5 + HSPPF, NCBAM, Recursive Gated Conv, and Soft-NMS)	Pest detection	IP16 (14 classes from IP102)	Rice, wheat, beet, alfalfa, corn, citrus, mango, and vine	16 pest types (e.g., alfalfa plant bug, and aphids)	Mean average precision of 92.5%	[28]
YOLOv5 with self-attention mechanism and multi-scale feature fusion	Pest detection	IP102 dataset (>75,000 images with 102 categories)	Rice	7 pest types (e.g., rice leaf roller, and pink rice borer)	Mean average precision of 79.8%	[29]
Custom CornerNet with DenseNet-100	Pest localization and classification	IP102 dataset	Rice, corn, wheat, beet, alfalfa, citrus, vitis, and mango	102 pest categories (insects in various life stages: egg, larva, pupa, and adult)	Achieved 68.74% classification accuracy and 57.23% mAP; outperformed other object detection models like YOLOv3, SSD, and Faster R-CNN in speed and accuracy	[31]
Improved YOLOv5s (ECMB-YOLOv5) - Backbone: MobileNetV3 - Attention: ECA - Neck: BiFPN - Loss: SIoU	Pest detection	IP102 dataset + photographs with a cellphone (total 2570 images)	Rice and wheat	7 pest types (e.g., rice leaf roller, and Mole cricket)	Mean average precision of > 95%	[32]
YOLOvs and Tiny-YOLOv3 neural network models	Pest location + pesticide spraying	Drone images	longan crops (orchard)	Tessaratoma papillosa (Drury)	Mean average precision of 93% and 89% with frames per second of 2.96 and 8.71 for YOLOv3 and Tiny-YOLO, respectively	[33]
BP Neural Network (ANN)	Predict droplet deposition and control UAV spray flow rate	Wind tunnel and field data; experimental data on UAV flight; environment, and structure; tillering stage rice simulation	Rice	General rice pests and diseases (not specified individually)	Droplet deposition prediction error < 20%; system achieved stable performance (R2= 0.997); variable spray met prescription values accurately in field tests	[34]
ResNet-50, Inception-v3, VGG-16, VGG-19, and Xception with SLIC superpixel segmentation and fine-tuning strategies	Classification of segmented pest images	INSection 5K13C: 5000 images of soybean pests in real field conditions	Soybean	12 pest categories (e.g., Spodoptera spp., Anticarsia gemmatalis, Nezara viridula adult/nymph, and Gastropoda) + 1 no-pest class	ResNet-50 with fine-tuning achieved highest accuracy of 93.82%. All DL models outperformed classical ML methods.	[35]
Deep CNN based on ResNet-18, modified to perform object detection with class probability maps	Detection, classification and sex differentiation of Drosophila suzukii	4753 labeled images from static traps and UAV-based images (249 traps total: 101 with SWD and 148 with bycatch)	Soft-skinned fruits (e.g., strawberries, raspberries, and cherries—SWD host crops)	Drosophila suzukii (spotted wing drosophila), male and female	AUC (static images): 0.603 (male), 0.506 (female), 0.669 (combined); AUC (UAV images): 0.284 (male), 0.086 (female), and 0.266 (combined); Demonstrated feasibility of UAV-based detection despite reduced image quality compared to static setup	[36]
Deep learning (BorerNet model with attention mechanism) using MFCC features	Classification and identification of wood-boring pests based on boring vibration signals under noisy environments	Custom dataset of vibration signals collected via self-developed piezoelectric sensors in field and soundproof settings; includes EAB, SCM, environmental noise, and simulated mixed signals	Ash trees (Fraxinus chinensis)	Emerald ash borer (Agrilus planipennis) and small carpenter moth (Streltzoviella insularis)	BorerNet achieved 96.67% accuracy and 0.95 F1-score, outperforming ResNet50, DenseNet, Inception V3, and VGG. Demonstrated strong robustness in noisy conditions and potential for real-world deployment	[37]

, demonstrating high accuracy in identifying target pests. Hu et al. [29] also improved YOLOv5 by incorporating a global contextual attention mechanism for superior feature extraction, replacing PANet with BiFPN for richer feature fusion, integrating a Swin Transformer to capture global context, and adding detection heads to broaden the detection scale. The optimized model achieved an 80% mean average precision in identifying seven rice adult pest species. Amrani et al. [30] used a YOLOv3 network–based framework and tested it on a dataset of 25,878 images, achieving an accuracy of 72%. Albattah et al. [31] developed a custom CornerNet model with DenseNet-100, achieving 68.74% classification accuracy and 57.2% mean average precision on the IP102 dataset across 102 pest species. By incorporating smartphone-captured images with IP102 dataset, Zhang et al. [32] employed an enhanced YOLOv5s model and achieved a mean average precision exceeding 95% in identifying seven distinct pest categories. More commonly, drone-mounted and stationary cameras are used to capture images, which are then analyzed using established DL models for pest identification, most of which deliver reliable and accurate results (Table 1).

AI systems can also incorporate environmental data, such as habitat and host plant characteristics, to enhance identification precision [38]. This capability is critical in dynamic agricultural landscapes where pest populations and beneficial insects coexist. AI-driven pest categorization is also scalable and adaptable, making it suitable for diverse farming systems. With AI integration, mobile applications enable farmers to capture pest images and receive real-time identification, supporting fast and informed decision-making. For example, GranoScan, a mobile app developed by Italian researchers, assists wheat farmers by providing real-time detection and identification of over 80 threats to wheat crops, including insect pests [39]. These systems can continuously update their databases to stay relevant as new pest species emerge. In addition, AI-powered acoustic sensors and DL models have been developed to detect hidden pests, such as wood-boring beetles and root-feeding grubs, by analyzing their sound patterns and acoustic signals [40,41]. This novel approach enhances the detection of subsurface or concealed infestations often missed by traditional scouting. It builds on the understanding that pest-induced plant responses, including the release of volatile stress signals, can attract natural enemies and indirectly indicate pest presence. The integration of AI with IoT devices further enhances its utility [42], as sensors deployed in the field can capture data on pest activity and transmit it to centralized AI models for analysis. By automating pest identification, AI not only improves the accuracy and efficiency of pest management practices but also supports sustainable agriculture by preserving beneficial insect populations.

AI-based systems have made remarkable strides in pest identification; however, their performance is still constrained by certain limitations, particularly the restricted capabilities of current data collection tools, which can hinder the quality and diversity of training data. For example, mimicry and camouflage allow pests such as the diamondback moth to blend with leaf surfaces, making traditional scouting equipment ineffective [43]. Additionally, seasonal and migratory behaviors significantly impact identification accuracy. The desert locust shifts between solitary and gregarious phases, with dramatic changes in coloration, morphology, and behavior, making identification challenging without genetic markers or remote sensing tools [44]. Another critical behavioral factor in pest identification is interaction with other species within the agroecosystem. Some pests rely on mutualistic relationships, such as aphids and ants, where ants protect aphids in exchange for honeydew, leading to indirect identification through monitoring ant activity [45]. Crop type and growth stage also determine the suitability of a field for certain pests, influencing the likelihood of infestation. Pest behavior and morphology, including mimicry and color variations, can further complicate identification, especially among closely related species. Therefore, understanding these behavioral intricacies is essential for developing advanced tools to improve detection accuracy and enhance pest management strategies.

2.3.2. Right Control Method

The right control method refers to selecting the most effective, sustainable, and economically viable strategy for managing pest populations while minimizing negative environmental and ecological impacts. Effective pest control relies on IPM principles, which combine multiple approaches such as biological, cultural, mechanical, and chemical control to reduce pest populations below economic thresholds. The correct method must consider pest species, crop type, infestation severity, and environmental factors to ensure maximum efficacy with minimal unintended consequences [46]. This is because certain control measures work best at specific developmental stages and under particular environmental conditions. Economic considerations, including the cost of control methods and their return on investment, also impact decision-making. This section provides a brief overview of the strengths and limitations of commonly used pest control strategies, including physical, biological, and chemical methods, followed by the AI roles.

Physical approaches rely on direct interventions to deter or eliminate pests through environmental barriers or mechanical means. Examples include reflective mulches to repel insects, nets to exclude larger pests [47,48], and machines like the pneumatic potato beetle killer developed by Almady and Khelifi [49]. Soil solarization and tillage also disrupt pest habitats. These methods are especially valuable in small-scale or resource-limited systems due to their low impact on soil and crops, though heavy equipment may cause soil compaction. Chemical methods are a key component of pest control, offering fast and effective suppression of pests using insecticides, fungicides, and herbicides. Common chemicals like organophosphates, carbamates, and glyphosate are widely used [50], especially in large-scale farming, due to their ease of application and broad-spectrum efficacy. Chemical controls are integral to IPM when used strategically. However, overuse has led to resistance in pests such as aphids and whiteflies, reduced effectiveness, and environmental risks [51]. These include harm to beneficial organisms, water contamination, and human health concerns, prompting stricter regulations and a push for safer alternatives [52,53]. Biological control uses natural enemies—such as predators, parasitoids, and pathogens—to manage pest populations while preserving beneficial organisms [54]. Examples include Trichogramma wasps targeting corn borer eggs and Metarhizium anisopliae fungi controlling termites [12,55]. This eco-friendly method supports biodiversity and offers long-term pest suppression with minimal chemical input. However, its success depends on environmental conditions, agent specificity, and labor-intensive implementation. Challenges like limited climate adaptability and potential ecological disruption underscore the need for improved scalability and reliability in biological pest control.

AI offers transformative potential to address the limitations of pest management methods, making them more efficient, scalable, and targeted. AI has emerged as a powerful tool to address the challenges associated with biological pest control, improving its efficacy, precision, and scalability. By integrating AI into biological pest management systems, researchers and farmers can optimize the deployment and monitoring of natural enemies, ensuring their success in diverse agricultural contexts. AI-powered systems can analyze complex datasets, including pest population dynamics, environmental conditions, and crop health indicators, to guide the timing and location of biological agent releases. For example, ML algorithms have been used to predict the emergence of pests based on weather patterns [56], enabling proactive deployment of natural predators like ladybugs or parasitic wasps. AI also plays a critical role in monitoring the performance of biological control agents. Advanced computer vision algorithms, coupled with high-resolution imaging from drones or field cameras, can track the activity and distribution of both pests and natural enemies in real time. For instance, Zhou et al. [57] used YOLOv4 and Faster R-CNN models to effectively and precisely detect two-spotted spider mites and their predatory mites on strawberries using smartphone images, achieving a detection accuracy of 93%. These advancements not only enhance the reliability of biological control in diverse agricultural systems but also reinforce its role as a key component of integrated pest management strategies. As AI technologies continue to evolve, their integration into biological pest control will undoubtedly expand, contributing to the development of resilient and environmentally sustainable food systems. Studies found that only 20–30% of sprayed pesticides are used by crops, while the rest is lost through runoff, leaching, evaporation, and drift, causing environmental pollution and reducing crop quality [58,59]. This underscores the need for precise pesticide application in both dosage and targeting. AI technologies enable the development of advanced pest monitoring systems that optimize pesticide application, ensuring chemicals are used only when and where they are needed. Specifically, AI-powered drones equipped with multispectral cameras can identify pest hotspots within fields by analyzing crop stress indicators such as discoloration or defoliation. By targeting pesticide sprays to these specific areas, farmers can significantly reduce chemical usage while maintaining effective pest control [60]. Moreover, environmental factors such as wind speed, temperature, and relative humidity significantly influence pesticide spray effectiveness in the field [61]. Deploying field sensors to monitor these conditions in real time enables integration with AI models to optimize and guide precise pesticide application. Through these innovations, AI addresses the challenges of chemical pesticide use by improving application precision, reducing environmental impacts, and combating resistance development.

2.3.3. Right Time Management

Right timing in pest management refers to the precise application of control measures at the most vulnerable stage of the pest’s life cycle, maximizing effectiveness while minimizing costs and environmental impact. Early detection and rapid response allow for interventions when pest populations are still manageable, preventing outbreaks that could escalate into severe infestations. A study analyzing data from 430 rice fields in China found that even a 15–20% incidence of delayed pest control significantly increased the overall frequency of pesticide applications across farming communities [62]. Similarly, in the case of fall armyworm, research from sub-Saharan Africa demonstrated that timely monitoring and early application of control measures greatly reduced maize yield loss compared to late interventions [63]. Similarly, precisely timing of control measures against the potato beetle can significantly reduce economic losses, as a single larva can consume around 40 cm² of potato foliage during its larval development, and an adult beetle can defoliate nearly 10 cm² per day [7]. These examples underscore that pest dynamics are time-sensitive, and failure to act promptly can result in irreversible economic and ecological consequences. A widely recommended approach is to initiate pest control when populations reach the first economic threshold, with timing adjusted based on environmental conditions, crop growth stage, and the specific pest species (Figure 3). Therefore, integrating real-time monitoring tools, predictive models, and responsive decision-making frameworks is essential to enhance the timeliness and effectiveness of pest management strategies.

AI-enhanced pest management allows interventions to be precisely timed at pests’ most vulnerable life stages, aligning with the IPM principle of applying controls at the right time for maximum impact. Advanced AI models (e.g., ML-based decision systems) integrate diverse data—including real-time weather conditions, pest and host phenology (life-cycle progression of both the pest and the crop), crop development stage, and historical outbreak patterns—to pinpoint the optimal window for action [64]. By forecasting pest population surges or key life-cycle events (such as peak egg hatch or larval emergence), these systems guide timely interventions that target the pest when it is most susceptible, thereby maximizing the efficacy of treatments while avoiding ill-timed or excessive applications [65]. For example, a commercial AI-driven platform combines daily trap counts with local meteorological data and multi-year pest trends to predict pest dynamics, achieving >80% accuracy in forecasting infestation peaks [66]. Such predictive timing not only improves pest control outcomes but also enables reduced pesticide use, lower costs, and minimal environmental impact by preventing needless sprays and focusing resources only when and where they are truly needed [65].

2.3.4. Right Action Taken

Right action application in pest management emphasizes the precise delivery of control measures using the most effective techniques, dosages, and equipment to maximize efficacy while minimizing environmental impact, resource waste, and non-target effects. This includes the accurate application of biological control agents, chemical pesticides, cultural practices, and mechanical interventions to ensure targeted pest suppression. Proper application is essential for reducing pesticide resistance, improving cost-effectiveness, and protecting beneficial organisms in agroecosystems. For example, traditional farming practices often involve the use of heavy machinery, such as tractors, in Western countries, and handheld sprayers for pesticide application in other regions. While these methods aim to enhance efficiency, they can inadvertently harm crop development, degrade soil quality, and pose significant health risks to humans [67]. Specifically, the deployment of heavy agricultural equipment can lead to soil compaction, which diminishes soil porosity and restricts root growth. Beyond environmental concerns, the application of pesticides, especially without adequate protective measures, exposes farmworkers and nearby communities to harmful chemicals. Such exposure has been linked to various health issues, including neurological disorders, respiratory problems, and an increased risk of certain cancers. Fortunately, modern farmers can now utilize specialized drones equipped with GPS technology to precisely spray pesticides [68], allowing accurate control over both the application rate and targeted locations based on the severity and distribution of pest infestations.

AI plays a pivotal role in implementing the “Right Action” in pest management by guiding highly targeted, optimized interventions that maximize pest control efficacy while minimizing collateral damage. AI-driven decision systems analyze real-time field data to pinpoint pest hotspots and recommend the optimal control method—whether targeted spraying, baiting, or trapping—along with the precise timing and dosage required. This enables the precise delivery of pest control measures using the most effective techniques and equipment. For example, AI-powered sprayers equipped with computer vision can distinguish crops from weeds or pest-infested plants and spray only the targets, reducing pesticide volumes dramatically (herbicide use has been cut by over 90% in trials of such smart sprayers) [69]. In orchards, intelligent sensor–controlled sprayers dynamically modulate spray output to match each tree’s canopy, avoiding over-application and drift while still achieving effective pest suppression [70]. Likewise, autonomous drones guided by AI can perform spot treatments or deliver micro-doses of pesticide to individual weed patches or insect hotspots, treating only where needed and leaving surrounding vegetation and beneficial insects unharmed [71]. By ensuring the right product is applied at the right rate, place, and time, these AI-enhanced pest management strategies bolster efficacy (through timely, accurate targeting) and simultaneously reduce environmental impact, resource waste, and non-target effects compared to traditional blanket applications.

3. Challenges in AI Utilization for Pest Management

1. Data challenges: High-quality, labeled datasets for AI in pest management are difficult to obtain due to imbalance, visual noise, and costly expert labeling [72]. Fragmented data and inconsistent standards further complicate model development. Solutions include stablishing standardized protocols, promoting data sharing, and utilizing semi-supervised or synthetic data. Current datasets like IP102 are useful but limited in species diversity.

2. Adoption barriers: Despite AI’s potential, adoption remains low among farmers due to infrastructure gaps (devices and connectivity), high costs, and limited awareness or training. Affordable IoT tools, user-friendly interfaces, multilingual support, and farmer education are needed to improve access and trust.

3. Limited AI coverage: AI applications mostly focus on pest identification, with limited progress in predicting outbreaks, guiding control strategies, or enabling real-time decision-making under field conditions.

4. Scalability issues: Most AI models are tested in small areas and struggle to scale due to spatial variability, connectivity limits, and high deployment costs. Environmental factors (e.g., light and weather) and technical constraints (e.g., battery and sensor alignment) affect accuracy. Future systems should include adaptive learning, multimodal inputs, and robust, site-flexible architectures.

4. Conclusions

The 4R framework offers a promising foundation for precision pest management, emphasizing accurate pest identification, appropriate control methods, optimal timing, and effective interventions. Its success depends on pest behavior, crop health, environmental conditions, and soil interactions. AI has shown strong potential in pest detection, particularly through image and acoustic sensing, with UAVs and IoT sensors enabling real-time data collection. While DL models like YOLOv5 perform well in trials, few have transitioned into scalable, user-friendly farm tools. Commercial platforms (e.g., Plantix, PEAT, and Taranis) are emerging, but adoption remains uneven. Public–private partnerships, open-access tools, and regionally adapted models are needed to support broader field deployment.

Emerging technologies are advancing AI-enabled 4R pest management. Hyperspectral imaging improves early pest detection (Right Identification), while edge computing enables real-time decisions (Right Timing and Action). Federated learning allows collaborative model training without compromising data privacy, promoting broader adoption. Successful implementation also depends on economic feasibility, particularly for smallholders. Though initial costs can be high, AI can reduce pesticide use and labor, offering long-term benefits. Adoption is influenced by usability, trust, and technical support. Scaling from trials to real-world farms remains challenging due to data and pest variability. Regulatory frameworks and data governance, especially for automated pesticide application, are still evolving. Despite promising accuracy, further work is needed to ensure field-level viability, affordability, and regulatory alignment.