Survey of Agent-based Intelligent Systems for Disease Detection in Onion Cultivation

The following abbreviations are used in this manuscript:

ABIS   Agent-Based Intelligent Systems AI Artificial Intelligence

IoT      Internet of Things

ML      Machine Learning

DL       Deep Learning

ABS    Agent-Based System

FL       Federated Learning

UAV   Unmanned Aerial Vehicle

SVM   Support Vector Machines RF Random Fores

CNNs Convolutional Neural Networks

OPBD Onion Purple Blotch Disease

H-CNN           Hybrid Convolutional Neural Network

FC       Fully Connected PCA Principal Component Analysis

MDTW-LSTM           Modified Dynamic Time Warping-Long Short-Term Memory

MAP   Mean Average Precision

CLIPS C Language Integrated Production System

CAMs Class Activation Maps

PLS     Partial Least Squares X

AI        Explainable Artificial Intelligence

Diseases such as crop disease, pests, and weeds are some of the challenges that affect the quantity and quality of agricultural production. These diseases are classified as biotic  (infectious) and abiotic (noninfectious). Infectious diseases (biotic) are usually caused by infection-causing agents such as bacteria, fungi, viruses, nematodes, and parasitic plants [1]. Non-infectious diseases (abiotic) are caused by poor farm management or unfavorable  environmental conditions that include high or low temperature, wind, moisture, drought or flood, soil compaction, prolonged and excessive rainfall, inadequate water management, deficiency or excess nutrients, and chemical injury caused by pesticides or salt [2]. Onions are among the most widely cultivated and consumed crops worldwide, playing a vital role in both subsistence farming and commercial agriculture [3]. However, onion cultivation faces significant challenges due to diseases and pests, which can lead to considerable yield losses and economic failures for farmers [4]. The common onion diseases presented in  Figure 1 are fungal, bacterial, viral, nematode, and other, as well as pests such as thrips, thrive under specific environmental conditions and often remain undetected until they have caused substantial damage. Traditional methods of detecting pests and diseases in onion crops rely on manual observation by farmers or agricultural experts. Although effective in some cases, these processes are slow, dependent on human perception, and likely to result in errors. [5]. In addition, limited access to agricultural experts in rural areas exacerbates the problem, leaving many farmers to rely on their experience, which may not always result in accurate diagnoses. Researchers are committed to supporting farmers and agricultural experts with plant-specific issues. Detecting and treating diseases become easier when visible to the naked eye, especially with sufficient information and regular crop monitoring. However, this becomes evident only in cases of severe disease or poor crop yields. Several innovations, particularly automated disease detection techniques, offer benefits to farmers [6]. [Figure 1].

 Figure 1: Taxonomy of Onion Diseases 

One of the causes of the low productivity ratio is the presence of several stressors such as diseases, fungi, insect pests, and abiotic stresses [7]. The occurrence of various diseases and insect pests greatly affects the quality and yield of onion crops. Onion diseases and pests include Anthracnose, Bulb rot seed, Damping off, Iris yellow spot virus, Stenophyllous blight, thrips, and others. Pests and diseases collectively cause 30-50% bulb yield losses. Controlling each of these diseases and thriving pests requires a specific set of management practices. The excessive use of pesticides and fungicides poses danger to the environment and also greatly affects humans health [8]. 

Advancements in AI, particularly in image processing, such as Machine Learning (ML) and Deep Learning (DL), have revealed great potential to automate and improve pest and disease detection [9]. Image processing techniques can accurately detect the visual symptoms of diseases, but these methods often overlook the environmental factors that influence disease outbreaks. For example, climate variables, such as temperature, humidity, and rainfall, create favorable conditions for disease propagation, thus making their consideration essential for a more comprehensive detection approach. Despite these advancements, existing systems for disease detection generally focus on a single data source such as plant images. This creates a gap in the precision and reliability of disease detection models, particularly for diseases influenced by complex interactions between visual symptoms and environmental factors. 

In this research, we focus on exploring the potential and impact of combining the images dataset with the environmental climate dataset in providing an efficient agent-based disease detection system. Given the above-mentioned objectives, this survey highlights key research studies, discusses current advancements, and examines the advantages of multi-agent coordination in disease monitoring and control. In addition, this paper identifies the challenges and limitations of agent-based model deployment, including data availability, computational complexity, and real world implementation barriers. 

This survey provides a comprehensive review of agent-based intelligent systems (ABIS) for disease detection in onion cultivation, exploring their integration with AI-driven models, ML, and the IoT. It classifies onion diseases into biotic and abiotic categories, analyzing detection techniques and ML models used in the literature. highlighting their impact on onion yield and the limitations of traditional detection methods. Also, the study evaluates the role of AI and IoT in onion disease monitoring, stressing the advantages of multi-agent coordination in precision onion cultivation. 

The following is the organization for the remaining part of the paper: Section 2 gives the review of related works and related or similar surveys in onion diseases, Section 3 describes the research methodology, Section 4 introduces the classifications of onion diseases, Section 5 traditional ways of detecting onion disease and their preventive management, Section 6 presents the Agent-Based System (ABS) for disease detection, challenges and future directions are highlighted in Section 7 and finally, Section 8 concludes the work by summarizing the salient contributions of the paper. 

Related Works:   

In recent years, research on agent-based intelligent systems for disease detection in agriculture has gained significant attention [10]. Various studies have explored the integration of AI, ML, and multi-agent systems to improve early detection and diagnosis of plant diseases. These approaches improve precision agriculture by providing automated real-time analysis, automated diagnosis, and guidance for farmers decision-making [11]. The presence of pests and diseases significantly threatens onion farming, reducing both the quality and quantity of yields. Various factors, such as plant infections, environmental stressors, poor irrigation, and soil salinity, contribute to these challenges [12]. In particular, soil salinization, whether occurring naturally or as a result of human activities can negatively impact plant health. As a result, detecting and predicting plant diseases has become a crucial focus in agricultural research. Implementing effective identification and remediation strategies can facilitate productivity by enabling timely interventions and enhance overall crop management. 

Onion diseases can be broadly classified into biotic and abiotic categories, each presenting unique challenges to crop health and productivity. Living organisms such as fungi, bacteria, viruses, and nematodes are responsible for biotic diseases. These pathogens invade onion plants, leading to visible symptoms that can severely impact yield if not controlled. Abiotic diseases result from non-living environmental factors, often linked to poor management practices or climatic conditions. These stressors do not involve pathogens but can make plants more susceptible to biotic infections.  

In the literature, there are a range of use cases in the management of crop diseases (scouting), including disease detection and prediction, disease classification, weed monitoring, and pest detection, monitoring, and classification [13]. The emergence of high-quality hyperspectral cameras and image processing algorithms libraries facilitates disease management in these solutions.  

An approach of particular interest is the federated learning (FL)-based method using  unmanned aerial vehicles (UAV) imaging for disease identification and classification presented in [14]. In this paper, four different farm locations are tested using UAVs to detect pest occurrences. The proposed pest classification solution effectively classified the nine detected pests in Kaggle pest datasets using FedAvg. In research conducted in [15], another FL framework was developed for the detection of rural weeds using hyperspectral pasture images captured from three independent sites. 

Related Surveys: 

This section reviews key surveys related to agent-based intelligent system disease detection in agriculture, with a focus on how these technologies have been applied to onion cultivation and similar crops. The findings of these surveys help identify gaps in current research, offering a foundation to advance more efficient and targeted solutions in disease management. 

A study by [16] explores the application of expert systems in the diagnosis of agricultural diseases, emphasizing their potential to improve crop productivity and management. Given that 61% of India’s agricultural land relies solely on rainfall, crop failures due to delayed rains often cause significant financial stress for farmers. The review highlights the way expert systems, which have been successfully applied in various scientific and business fields, can assist farmers in identifying suitable crops, efficiently managing the cultivation, and improving productivity. The authors particularly focus on the design of an expert system for diagnosing diseases in paddy crops, demonstrating the practical benefits such technologies can offer in real-world agricultural settings.

A comprehensive review by [17] highlights the role of AI, particularly ML and DL, in enhancing disease detection and management in crop cultivation. Focusing on potato diseases such as late blight and early blight, the study emphasizes on how conventional detection methods are often slow, labor intensive, and error-prone. Through the analysis of 72 key studies, the authors found that image-processing techniques, combined with climate data, provide more accurate and timely disease forecasts. Commonly used algorithms include support vector machines (SVM), random forest (RF), convolutional neural networks (CNN), and mobile networks, with precision rates ranging from 64.3% to 100%. 

AI has become a crucial technology for managing plant diseases and pest infestations, which severely impact global food security. Recent advances, particularly in image recognition and disease modeling, have improved the early detection and precise diagnosis of biotic stresses, helping to reduce crop losses and improve agricultural productivity. This review provides a detailed analysis of these developments and their practical applications in modern farming practices [18]. 

Survey of Onion Diseases:  

Onion production is frequently threatened by various diseases that affect crop yield, quality, and overall agricultural productivity. These diseases arise from both biotic factors, such as bacteria, viruses, fungi, and nematodes, and abiotic stressors, including environmental conditions, nutrient deficiencies, and soil imbalances. Identifying and managing these diseases is crucial to ensure sustainable onion production and minimizing economic losses. 

According to [19] onion is a highly perishable vegetable in India, valued as a flavoring agent. Despite its ability to be stored for 8–10 months under proper conditions, around 35–40% of onions are lost due to storage diseases, with fungal bulb rot contributing to 15–30% of the losses. Various fungal pathogens, including Penicillium spp., Rhizopus spp., Aspergillus spp., and Fusarium spp. affect onions during post-harvest storage, with Aspergillus niger being the most aggressive. Conventional fungicides are insufficient and pose health risks due to residual toxicity. Therefore, alternative strategies, such as plant-derived compounds and biofungicides, especially nanobiofungicides, present a promising solution for managing fungal diseases and reducing onion storage losses. 

A study by [20], shows that onion production is significantly affected by fungal pathogens, particularly species of the Botrytis genus, which cause three major diseases: neck rot, flower blight and leaf blight. Botrytis squamosa leads to leaf blight, causing severe yield losses, while neck rot, a post harvest disease, is linked to B. aclada, B. allii, and B. byssoidea, making early detection challenging. In seed production, flower blight is caused by these species along with B. cinerea, reducing seed yield and quality. This review discusses the historical impact of Botrytis diseases on onions and explores recent advances in understanding their biology, ecology, and host interactions. Gaining insights into the genetic, biochemical, and physiological aspects of Botrytis infections is crucial for developing resistant onion cultivars. 

In another review conducted by [21], onion is a vital vegetable crop cultivated world-wide, including Ethiopia for its daily use and economic benefits. Initially introduced from Sudan, its production has expanded, with smallholder farmers playing a key role. However, productivity remains below the global average due to a lack of poor agronomic practices, improved cultivars and limited attention to onion farming. Farmers often rely on low-quality planting materials from local markets, semi-formal seed producers, or untested imported seeds, which are prone to diseases and pests. The absence of an authorized body for producing and distributing high-quality onion seeds remains a major challenge. Despite this, onion cultivation is increasing due to its economic profitability, with ongoing efforts to improve productivity through better nutrient management, irrigation scheduling, optimal plant density, and pest control. This paper reviews the key challenges facing onion production in Ethiopia. 

A review presented by [22] focuses on how diseases and insect pests significantly impact onion production in India, with their prevalence and distribution increasing over the past decade, causing annual yield losses of 10–15% of the onion produced. While several pests and diseases have been reported, a recent survey identified purple blotch (Alternaria porri), onion thrips (Thrips tabaci), and Stemphylium blight (Stemphylium vesicarium) as the most critical at the national level. Additionally, Colletotrichum blight (Colletotrichum gloeosporioides) was found to be important economically but was primarily localized in Maharashtra state. 

We classified onion diseases into Biotic, Abiotic and General categories as reviewed in the subsections of 2.2 and illustrated in [figure 2]. 

Biotic (infectious):  

The study of [23] presented an image-based field monitoring system for automatic disease detection in onion cultivation using deep neural networks (DNNs) trained with weakly supervised learning. The system comprises a motor system, PTZ camera, wireless transceiver, and an image logging module for periodic image capture in onion fields. The DL model classifies six onion disease symptoms and localizes affected areas using Class Activation Maps (CAMs), with an optimal threshold of 60% of the maximum CAM value. The performance of the system was evaluated using the Mean Average Precision (MAP) metric at IoU 0.5, achieving accuracy between 74.1% and 87.2%, demonstrating real-time disease detection capabilities. The study emphasizes the potential of AI-driven automated monitoring systems to optimize precision agriculture and disease management in onion cultivation. 

Another study presented by [24] focuses on automatic grading and disease identification in agricultural products using image processing techniques to enhance efficiency in agricultural trade. The proposed method employs Kapur’s thresholding-based segmentation to detect diseased regions and k-means clustering for feature extraction. The Modified Dynamic Time Warping - Long Short-Term Memory (MDTW-LSTM) model is used for classification, outperforming traditional LSTM and DTW algorithms. The experimental results demonstrate a high accuracy of 97.35%, highlighting the potential of AI-driven automation in agricultural disease detection and product grading. 

A mobile app serving as a diagnostic tool can provide quick, accurate, and inexpensive results through image processing techniques as a way to improve agricultural practices and make disease management more accessible and efficient for farmers across the world. This capstone project used a total of 3000 labeled images to train the sequential model using Google Colaboratory. Purple blotch and leaf blight were the primary leaf diseases in onions involved in this project. Proper dataset distribution is an essential requirement in model development in AI; it ensures that any model developed is properly configured and tested [25].  

According to research conducted by [26] accurate and timely evaluation of Onion Purple Blotch Disease (OPBD) severity significantly helps in managing crop outputs and minimizing economic losses in onion cultivation. Traditional techniques of disease assessment, based on visual inspection, are prone to human error and time-consuming, further making the case for automated solutions. CNNs models, particularly the VGG16 architecture, have demonstrated significant potential for automating the grading of plant disease severity, offering an improvement over the manual method. The CNN-VGG16 hybrid model developed in the study achieved an impressive overall accuracy of 93.5%, surpassing many modern models in disease detection and severity grading. The research shows that AI-based agricultural solutions, such as the CNN-VGG16 model, have the potential to revolutionize precision agriculture by automating disease detection and classification processes.  

The integration of AI (ML and DL) technologies into precision agriculture promises to offer more efficient, scalable, and objective alternatives to traditional disease detection and severity grading methods. Despite the success of CNN in plant disease detection, there is a gap in research on using the VGG16 model specifically for multi-level severity grading of OPBD, indicating opportunities for further study. The proposed study aims to fill this gap by customizing and optimizing the VGG16 model for multi-level severity grading of OPBD, contributing to the advancement of AI-driven solutions in agriculture [26]. 

Another research carried out by [27] presented the development of an expert system designed to assist farmers and specialists in diagnosing onion plant diseases and providing appropriate recommendations. The system employs production rules to capture knowledge and ensure effective consultation. It was developed using C Language Integrated Production System (CLIPS) with a Delphi language interface for implementation. The expert system demonstrated high accuracy in diagnosing tested onion disease cases, ensuring reliable and precise disease identification [27].  

OPBD is marked by striking dark-purplish streaks that are formed on onion leaf surface. These lesions lower the rigidity of the layers, which means that they impair the effectiveness of photosynthesis, inhibiting bulb formation. This depends on several factors such as weather, onions grown, and the rate of fungal infection. 

OPBD is an important disease that can be prevented if it is properly understood. If the severity and spread of this disease are known, proper control measures will be established. Traditionally, grading has been based on manual expert evaluations involving subjectivity and effort requirements. Many of such subjective assessments result in an inability to make accurate interventions since they always have disparities that are not suitable for an assessment of the subject matter [28]. 

Allium cepa is the onions’ scientific name. Its edible bulbs are highly sought after since they are nutritious and provide a bold and distinctive flavor to dishes. Onions are usually grown every year for various purposes. Moreover, we can grow them from seeds or seedlings for a faster harvest. In contrast, when we grow onions from their seeds, they take two years to flower. After picking onions, the bulbs can be immediately. They can be used in processed food or stored for future use. Onion is also known for its nutritional benefits and unique flavor profiles, which are crucial components in cuisines worldwide. In 2018, farmers grew almost 100 million tons of onions worldwide on more than 5 million hectares of land. Onion is an integral part of many cuisines around the world and is a major ingredient in many pharmaceutical industries. A shortage in the global supply of onions would mean a disruption of many supply chains, which can have tangible economic consequences. 

However, several leaf diseases are a common obstacle to the effective development of these crops. Leaf spots, blight, and mildew are only some of the symptoms of these diseases, which are caused by a wide variety of destructive fungi, bacteria, and viruses. 

Onion leaf diseases, including Purple Blotch, Downy Mildew, and Neck Rot, have been associated with some notable pathogens such as Alternaria porri, Peronospora destructor, and Botrytis allii. If these infections are not found and treated quickly and efficiently, they can wipe out whole harvests [29]. The authors also explained that onions do not last long in Bangladesh because they are often grown improperly and dried incorrectly. This results in onions of poor quality with too much moisture. Farmers lose about 2.4% of their total onion crops due to a lack of knowledge about onion planting and drying. Therefore, it is vital to fully understand the origin, pathogenesis, and spread of these diseases to develop and implement effective disease management strategies. 

Although traditional approaches, such as the application of fungicides and the maintenance of sanitation in cultivation areas, are commonly practiced, there is an increasing inclination towards innovative solutions. This includes exploring genetic resistance and biological control methods to combat these diseases sustainably. The severity and incidence of these diseases are also heavily influenced by environmental conditions. Factors such as temperature, plant density, and humidity can significantly dictate the onset and progression of these diseases. This nexus between environmental factors and leaf disease manifestation adds complexity to disease management processes in different geographical regions [29] [Table 1].

Table 1: Biotic (Infectious) 

Abiotic (non-infectious): 

The authors in [30] examine the impact of climate conditions and crop management practices on onions and garlic growth and yield, which are key field vegetable crops. Using text mining techniques such as word cloud and semantic networks, the research analyzes literature from multiple databases to identify trends in crop modeling and yield prediction. Some findings highlight that temperature, precipitation, solar radiation, and humidity significantly influence crop productivity. Additionally, sowing time, seed treatment, irrigation intervals, and fertilization are key management factors that affect yield. The study emphasizes the need for integrated crop models that incorporate both environmental and agronomic factors to improve the accuracy of yield predictions. 

Another research introduced by [31] highlighted that cultivation of Allium crops, particularly onion and garlic, faces significant challenges due to climate change and environmental stresses. Factors such as increased temperatures, genetic limitations, and breeding complexities impact their production and productivity. Onion breeding is hindered by its large genome size, biennial life cycle, and high cross-pollination, while garlic’s asexual nature restricts genetic improvements. Addressing these challenges requires a systematic research approach, including the development of climate-resilient varieties, genomic resources, and advanced breeding techniques. Integrating omics tools and conventional breeding methods is essential to ensure a stable supply of Allium crops in response to global demand [31] [Table 2].

Table 2: ABiotic (Non-Infectious)

The research of [32] introduced a novel hierarchical CNN-based model for plant disease classification, which aims to improve computational efficiency and accuracy by addressing nuances often overlooked by conventional systems. The proposed hierarchical CNN model achieved an impressive 99.7% accuracy on the validation set, outperforming existing models and demonstrating its ability to handle challenges such as overlapping disease symptoms. Studies utilizing CNNs, such as those employing AlexNet, GoogLeNet, and ResNet, have demonstrated high precision in plant disease identification, with ResNet achieving 97.28% accuracy for tomato leaf disease detection. 

The use of pre-trained models, such as VGG-19 with ensemble methods, has also been explored, with results showing impressive accuracy of 98.6% for plant disease classification. The methodology involves using benchmark datasets, such as the PlantVillage and PlantDoc datasets, which contain diverse plant species and disease conditions, providing a robust foundation for training and evaluating DL models in plant disease detection. Plant diseases are a significant challenge to global agricultural productivity, with nearly 40% of global food production being compromised due to these diseases, leading to severe socio-economic consequences. 

The precise identification and classification of plant diseases is critical for effective disease management, but this task is complicated by the variety of symptoms that can be influenced by factors such as plant species, growth phase, and environmental conditions. Traditional feature-based plant disease classification methods, which involve manual feature extraction such as color histograms and texture patterns, often require domain-specific expertise and may fail when applied to new diseases or diverse image datasets [Tables 3-5]. 

DL models, particularly CNNs, have shown great promise in automating plant disease classification by autonomously identifying image features, addressing many of the limitations of traditional methods [32]. Accurate classification and identification of plant diseases is important for maintaining crop health and optimizing crop production. DL techniques have shown great promise in this field, and this study presents a Hybrid-CNN (H-CNN) model for classification of plant disease. The suggested model architecture combines multiple Convolutional (Conv), and Fully Connected (FC) layers, which are normally trained on publicly available datasets of plant leaves images. The model employs visual and spectral features of plant leaves for classification efficiency. The visual features are collected using a pre-trained VGG-16 network, while the spectral features are extracted using Principal Component Analysis (PCA). The hyper-parameters for the CNN layers are optimized using a grid search algorithm [33]. 

 Figure 2: Analysis of the Existing Research

Table 3: General 

Table 4: Comparing of Proposed Solution in Related Work

Table 5: Biotic Onion Diseases (infectious)

From the review of onion disease detection, several contributions have been made in image-based disease detection, DL applications, and automated field monitoring systems. However, significant research gaps remain. Existing studies primarily focus on analyzing plant images to detect visible symptoms of diseases, often overlooking the crucial role of environmental factors such as humidity, temperature, and rainfall, which significantly influence the onset and spread of plant diseases. Additionally, most researchers focus solely on biotic diseases, leading to a scarcity of research on abiotic diseases or studies that address both biotic and abiotic factors. This limitation reduces the overall effectiveness and applicability of the reviewed works. To address these challenges, there is a need for a more robust and comprehensive study that integrates both image processing and environmental data while simultaneously considering both biotic and abiotic diseases.

This section presents the methodology used to investigate related work, including the keywords utilized, research objectives and questions, as well as the criteria for selection and quality assessment. The process was carried out in three main stages: planning, reviewing and reporting. During the planning stage, research objectives and  questions were established. In the reviewing stage, articles were selected according to established selection and evaluation criteria. Finally, the reporting stage involved presenting the research findings. 

Research Objectives : 

The main objective of this research is to examines and analyses the existing agent-based intelligent systems used for disease detection in onion cultivation. In addition, evaluating the effectiveness of agent-based models in identifying and diagnosing onion diseases. To achieve this objective, we provide a comprehensive survey of the existing agent-based intelligent systems used for disease detection in onion cultivation and analyze previous studies and their applications. 

Research Questions: 

1. What are the existing agent-based intelligent systems used for disease detection in onion cultivation?
2. How effective are agent-based models in diagnosing onion diseases compared to other methods?
3. What are the main advantages and challenges of using agent-based systems for disease detection? 

1. Articles collection procedure: 

The research primarily sourced articles are from reputable databases such as MDPI, IEEE Xplore, Springer, Elsevier (ScienceDirect), ACM and others published between 2020 and 2024. 

       b.Onion Diseases Classification: 

Diseases such as crop disease, pests, and weeds influences the production capacity and quality of agricultural goods. In onion farming, these diseases also affect the end yield of onion farmers if left untreated (not managed efficiently). These diseases vary from biotic (infectious) and abiotic (non-infectious). Infectious diseases (biotic) are usually caused by infection causal agents such as bacteria, fungi, viruses, nematodes and parasitic plants The non-infectious diseases (abiotic) are caused by inadequate farm management or unfavorable environmental conditions which include moisture, low or high temperature, drought or flood, wind, soil compaction, continuous and heavy rain, deficiency or excess of nutrients, improper water management and chemical injury caused by pesticides or salt [Figure 3]. 

       c.Biotic Onion Diseases (infectious): 

Biotic (infectious) diseases are onion diseases usually caused by infection causal agents  such as bacteria, fungi, viruses, nematodes, parasitic plants and the rest as presented in Table 5. By identifying these diseases accurately, appropriate control measures such as crop rotation, resistant varieties, or fungicides can be applied to manage them effectively. The common biotic onion diseases as presented in Figure 3 are Fungal, Bacterial, viral, Nimatode and Others [Table 6]. 

Table 6: Abiotic Onion Diseases (noninfectious)         

 Figure 3: Onion Biotic Diseases

Abiotic Onion Diseases(non-infectious):  

Abiotic onion diseases are disorders in onion crops caused by non-living (abiotic) factors rather than pathogens like fungi, bacteria, or viruses. These factors typically result from environmental conditions, cultural practices, or nutrient imbalances. Below are common abiotic onion diseases or disorders as presented in Table 6.

Traditional method of disease detection:  

Traditional disease detection in onions primarily involves manual field inspections  and visual observations of symptoms [34]. These methods typically rely on farmers and agricultural experts to identify visual cues of disease, such as leaf discoloration, wilting, and spots. Conventional approaches include: 

• Field Surveys: Manual inspection of crops for visual symptoms of disease.
• Laboratory Testing: Plant tissue samples are taken to laboratories for microbiological analysis to identify pathogens.
• Expert Consultation: Experienced agricultural workers or plant pathologists diagnose diseases based on symptoms observed in the field. 

Prevention and Management:  

• Conduct soil tests to ensure proper nutrient balance and pH.
• Use irrigation systems to provide a consistent water supply.
• Select onion varieties suitable for local growing conditions.
• Avoid overuse of chemicals and follow proper application techniques.
• Protect onions from extreme weather with shade nets or mulches. 

Agent-Based Intelligent Systems in Agriculture: 

Agent-based systems (ABS) are computational models that simulate the actions and interactions of autonomous agents within an environment to study complex phenomena [35]. These agents are individual entities with specific objectives, behaviors, and decision-making capabilities. In the context of agriculture, ABS models are designed to replicate the  behavior of plants, pests, diseases, or environmental factors to enhance decision-making  and optimize crop management. 

Principles of Agent-Based Systems: 

• Autonomy: Agents in the system operate independently, making decisions based on predefined rules and interactions with other agents or the environment
• Interaction: Agents can interact with each other and the environment, influencing their own behavior or the system’s overall behavior.
• Adaptation: Agents can adapt to changing conditions or inputs, making ABS highly flexible and dynamic
• Decentralization: Agent-based systems are typically decentralized, meaning no central controller is directing the agents; they make decisions based on local information and rules. 

How Agent-Based Models Work in Disease Detection: 

In agriculture, agent-based models are employed to simulate the dynamics of disease spread, pest infestations, and other environmental factors affecting crops. For disease detection, agents representing plants and pathogens interact within a simulated environment where factors such as temperature, humidity, and soil conditions can influence the growth and spread of diseases [36]. 

• Disease Propagation: ABS can model how diseases like Onion Downy Mildew or Purple Blotch spread from infected plants to healthy ones, simulating real-time disease dynamics.
• Early Detection: By simulating the interaction of plants and pathogens under various conditions, ABS can help predict potential disease outbreaks and identify early signs of infection before they become visible to the human eye.
• Optimizing Intervention: ABS can also recommend intervention strategies, such as when and where to apply pesticides or fungicides, based on the simulated progression of the disease. 

Challenges and Future Directions: 

This section presents the current challenges of Agent-based systems (ABS) in agriculture and highlights the directions for the future.  

Technical Challenges in Agent-Based Systems:  

Despite the promising capabilities of agent-based intelligent systems in disease detection, several technical challenges hinder their efficiency and widespread adoption [10]. One key issue is computational complexity, as agent-based models require significant processing power to analyze large datasets and perform real-time decision-making. Additionally, scalability issues arise when attempting to expand the system to larger onion farms with diverse environmental conditions, necessitating adaptive and flexible models. Another concern is inter-agent communication, where ensuring seamless interaction between autonomous agents for data sharing and decision-making remains a critical challenge [37]. Without effective coordination, the overall system performance may be compromised. 

Real-World Implementation Challenges in Onion Farming: 

The implementation of agent-based disease detection systems in onion agriculture encounters several challenges. A primary obstacle is the farmers’ insufficient technical proficiency, as many traditional cultivators may be untrained in utilizing advanced AI technologies. Moreover, financial and infrastructural limitations hinder small-scale farmers from acquiring and sustaining such intelligent systems, particularly in areas lacking high-speed internet and dependable power. Additionally, the environmental variability associated with onion cultivation presents a formidable issue, necessitating adaptive models that can effectively generalize across diverse agricultural contexts [38]. 

Potential Improvements in Agent-Based Disease Detection:  

To improve agent-based disease detection systems, various enhancements can be investigated. According to [39] a noteworthy strategy involves utilizing hybrid models that merge agent-based methodologies with optimization techniques such as Genetic  Algorithms or Particle Swarm Optimization to refine decision-making processes. Furthermore, real-time monitoring can be optimized through the integration of high-speed data processing methods, facilitating prompt detection and response. An additional crucial development is the implementation of Explainable AI (XAI), which enables farmers to comprehend the rationale behind specific disease predictions, thereby fostering trust and applicability.  

Integration with Deep Learning and IoT for Better Accuracy: 

The integration of deep learning and IoT technologies presents significant opportunities for improving agent-based disease detection in onion farming. Deep learning-based image recognition, particularly through CNNs, can enhance disease detection accuracy by analyzing leaf images more effectively [40]. Meanwhile, IoT-enabled sensors can be deployed to collect real-time environmental data such as soil moisture, temperature, and humidity, providing valuable insights for disease prediction. Furthermore, cloud-based systems can be utilized for efficient data storage and analysis, enabling farmers to access diagnostic results remotely and make informed decisions. 

Future Research Directions: 

Future research should focus on developing adaptive agent-based models that can dynamically adjust to different environmental conditions and farming practices. Additionally, efforts should be made to enhance affordable and user-friendly implementations of agent-based systems, making them accessible to smallholder farmers. Lastly, investigating synergistic approaches that combine AI, IoT, and blockchain technology can further improve the reliability, transparency, and efficiency of disease detection systems in onion cultivation.

Onion cultivation plays a vital role in global agriculture, but it faces significant challenges due to diseases that reduce yield and quality. These onion diseases can be biotic  (infectious) and abiotic (non-infectious). This survey focus on exploring the agent-based disease detection solutions found in the literature with the view of analyzing their impact on detection of different types on diseases affecting onion. We provide a taxonomy of onion diseases by analyzing the available solution presented in each categories focusing on the ML models used, dataset types and result accuracy. We highlights the challenges and limitations of agent-based model deployment, including data availability, computational complexity, and real-world implementation barriers. Finally, we proposes the potential future directions for the future research.  

Author Contributions: Conceptualization, Mubashir Haruna; methodology, Anas Tukur Balarabe; validation, Muhammad Saidu Aliero; formal analysis, Bashir Yauri Aliyu; investigation, Dalhatu Muhammed; resources, Maria Trocan; writing original draft preparation, Mubashir Haruna and Bashir Yauri Aliyu; writing  review and editing, Dalhatu Muhammed.

This research received no external funding. 

The authors declare no conflicts of interest. 

1. Udutalapally V, Mohanty SP, Pallagani V, Khandelwal V (2020) scrop: A novel device for sustainable automatic disease prediction, crop selection, and irrigation in internet-of-agro-things for smart agriculture. IEEE Sensors Journal, 21: 17525–17538.
2. Chatterjee PS, Ray NK, Mohanty SP (2021) LiveCare: An IoT-Based Healthcare Framework for Livestock in Smart Agriculture. IEEE Transactions on Consumer Electronics, 67: 257–265.
3. Miassi Y, Dossa F, Banzou K (2018) Onion (Allium cepa) production in urban and peri-urban areas: financial performance and importance of this activity for market gardeners in Southern Benin. Current Investigations in Agriculture and Current Research, 3: 2.
4. Yeshiwas Y, Alemayehu M, Adgo E (2023) The rise and fall of onion production; its multiple constraints on pre-harvest and post-harvest management issues along the supply chain in northwest Ethiopia. Heliyon.
5. Vänninen I (2022) Advances in insect pest and disease monitoring and forecasting in horticulture.
6. Nishitha P, Sarika T, Navya S, Peneti S, Rajeshwari D, et al. (2024) Smart Agriculture: CNN-Based Systems for Early Detection and Diagnosis of Plant Diseases and Pests. In Proceedings of the Second International Conference on Intelligent Cyber Physical Systems and Internet of Things (ICoICI). IEEE, pp. 1427–1431.
7. Gowtham HG, Hema P, Murali M, Shilpa N, Nataraj K, et al. (2024) Fungal endophytes as mitigators against biotic and abiotic stresses in crop plants. Journal of Fungi, 10: 116.
8. Banerjee S, Sarda K, Khandelwal S, Gajarushi A, Gawande S, et al. (2024) IoT-based Sensing System for Thrips Pest and Disease Management in Onion Crop. 2024 IEEE Applied Sensing Conference (APSCON). 1–4.
9. Saleem MH, Potgieter J, Arif KM (2021) Automation in agriculture by machine and deep learning techniques: A review of recent developments. Precision Agriculture 22: 2053–2091.
10. Khemakhem F, Ellouzi H, Ltifi H, Ayed MB (2020) Agent-based intelligent decision support systems: a systematic review. IEEE Transactions on Cognitive and Developmental Systems, 14: 20–34.
11. Mohyuddin G, Khan MA, Haseeb A, Mahpara S, Waseem M, et al. (2024) Evaluation of Machine Learning approaches for precision farming in Smart Agriculture System-A comprehensive Review. IEEE Access.
12. Wood J (2023) The Effects of Biotic Stress on Onion Physiology. PhD thesis, New Mexico State University.
13. Muhammed D, Ahvar E, Ahvar S, Trocan M, Montpetit MJ, et al. (2024) Artificial Intelligence of Things (AIoT) for smart agriculture: A review of architectures, technologies and solutions. Journal of Network and Computer Applications, 228: 103905.
14. Khan FS, Khan S, Mohd MNH, Waseem A, Khan MNA, et al. (2022) Federated learning-based UAVs for the diagnosis of Plant Diseases. In Proceedings of the 2022 International Conference on Engineering and Emerging Technologies (ICEET). IEEE, pp. 1–6.
15. Patros P, Ooi M, Huang V, Mayo M, Anderson C, et al. (2022) Rural ai: Serverless-powered federated learning for remote applications. IEEE Internet Computing.
16. Deepthi MB, Sreekantha D (2017) Application of expert systems for agricultural crop disease diagnoses—A review. In Proceedings of the International conference on inventive communication and computational technologies (ICICCT). IEEE, 222–229.
17. Kaur A, Randhawa GS, Abbas F, Ali M, Esau TJ, et al. (2024) Artificial Intelligence Driven Smart Farming for Accurate Detection of Potato Diseases: A Systematic Review. IEEE Access.
18. Paul B, Paul M, Rub A (2024) Advancements in AI-Based Pest and Disease Detection in Agriculture: A Comprehensive Review of Image Recognition and Disease Modelling.
19. Hassan M, Yousuf V, Shah T, Bhat N, Bhat Z, et al. (2020) Stemphylium blight of onion: a review. Agricultural Reviews, 41: 51–58.
20. Steentjes MBF, Scholten OE, van Kan JA (2021) Peeling the onion: Towards a better understanding of Botrytis diseases of onion. Phytopathology, 111: 464–473.
21. Etana MB, Aga MC, Fufa BO (2019) Major onion (Allium cepa L.) production challenges in Ethiopia: A review. Journal of Biology, Agriculture and Healthcare, 9: 42–47.
22. Gupta R, Srivastava K, Pandey U (1993) Diseases and insect pests of onion in India. In Proceedings of the International Symposium on Alliums for the Tropics 358, 265–270.
23. Kim WS, Lee DH, Kim YJ (2020) Machine vision-based automatic disease symptom detection of onion downy mildew. Computers and Electronics in Agriculture, 168: 105099.
24. Chaturvedi A, Nadgaundi SK, Raja MK, Dubey VR, Singhal A, et al. (2024) Intelligent Agents System for Vegetable Plant Disease Detection Using MDTW-LSTM Model. IEEE, 1–6.
25. Panaligan N, Talaid N (2024) Spring Onion Disease Detection and Treatment Recommendations. Isabela State University Linker: Journal of Engineering, Computing and Technology, 1: 1–11.  
26. Govindharaj I, Thapliyal N, Aeri M, Kukreja V, Sharma R (2024) Onion Purple Blotch Disease Severity Grading: Leveraging a CNN-VGG16 Hybrid Model for Multi-Level Assessment. In Proceedings of the International Conference on Innovations and Challenges in Emerging Technologies (ICICET). IEEE, 1–5.
27. Ali A, Randa Q (2016) Onion RBS for Disorders Diagnosis and Treatment. IJAPR, 8 :1-8
28. Ojha NK, Thapliyal N, Aeri M, Kukreja V, Sharma R (2024) CNN-VGG16 Hybrid Model for Onion Purple Blotch Disease Severity Multi-Level Grading. In Proceedings of the 2024 5th International Conference for Emerging Technology (INCET). 1–5.
29. Samia IJ, Jahan M, Rony MAH, Fatema K (2024) An efficient and high-accuracy based automated onion leaf disease diagnosis approach using mask R-CNN framework. International Conference on Computing Communication and Networking Technologies (ICCCNT). 1–7.
30. Kim JH, Kim DJ, Seo BH, Kim KS (2021) Analysis of Literatures Related to Crop Growth and Yield of Onion and Garlic Using Text-mining Approaches for Develop Productivity Prediction Models. Korean Journal of Agricultural and Forest Meteorology, 23: 374–390.
31. Khar A, Singh H, Verma P (2022) Mitigating abiotic stresses in Allium under changing climatic scenario. In Genomic designing for abiotic stress resistant vegetable crops; Springer 253–278.
32. Vardhan M, Sharma S (2024) Enhancing Plant Pathology with CNNs: A Hierarchical Approach for Accurate Disease Identification. In Proceedings of the Proceedings of the 13th International Conference on Software and Computer Applications, 159–164.
33. Sharma S, Vardhan M (2023) Hyperparameter Tuned Hybrid Convolutional Neural Network (H-CNN) for Accurate Plant Disease Classification. International Conference on Communication, Circuits, and Systems (IC3S). 1–6.  
34. Singh V, Sharma N, Singh S (2020) A review of imaging techniques for plant disease detection. Artificial Intelligence in Agriculture, 4: 229–242.
35. An L, Grimm V, Sullivan A, Turner Ii B, Malleson N, et al. (2021) Challenges, tasks, and opportunities in modeling agent-based complex systems. Ecological Modelling, 457: 109685.
36. Singh BK, Delgado-Baquerizo M, Egidi E, Guirado E, Leach JE, et al. (2023) Climate change impacts on plant pathogens, food security and paths forward. Nature Reviews Microbiology, 21: 640–656.
37. Firoozi AA, Firoozi AA (2024) Intelligent Multi-Agent Systems for Advanced Geotechnical Monitoring. Artificial Intelligence Annual Volume.
38. Dhal SB, Kar D (2024) Transforming Agricultural Productivity with AI-Driven Forecasting: Innovations in Food Security and Supply Chain Optimization. MDPI Forecasting, 6: 925-951
39. Mazzetto S (2024) Interdisciplinary Perspectives on Agent-Based Modeling in the Architecture, Engineering, and Construction Industry: A Comprehensive Review. Buildings, 14: 3480.  
40. Tugrul B, Elfatimi E, Eryigit R (2022) Convolutional neural networks in detection of plant leaf diseases: A review. Agriculture, 12: 1192.