An Integrated IoT and AI Framework for Real‑Time Crop

Monitoring, Adaptive Soil Analysis and Intelligent Yield Prediction

Padmaja Pulicherla

, P. M. Priyanka

, Valli Madhavi Koti

, B. Sushma

K. Shanmugapriya

and B. Poornima

Department of CSE(AIML), Hyderabad Institute of Technology and Management, Hyderabad, Telangana, India

Department of Computer Science and Engineering, Ravindra College of Engineering for Women, Kurnool‑518002, Andhra

Pradesh, India

Department of Computer Science, GIET Degree College, East Godavari, Andhra Pradesh, India

Department of Information Technology, MLR Institute of Technology, Hyderabad, Telangana, India

Department of Computer Science and Engineering, Nandha Engineering College, Erode‑638052, Tamil Nadu, India

Department of Computer Science and Engineering, Mahatma Gandhi Institute of Technology, Hyderabad, Telangana,

India

Keywords: IoT, AI, Crop Monitoring, Soil Analysis, Yield Prediction.

Abstract: With the emergence of the Internet of Things (IoT) and Artificial Intelligence (AI), smart agriculture becomes

more dynamic, and provides new possibilities in crop monitoring and predicting the yield. To this end, in this

paper, we introduce a unified architecture to combine Realtime IoT-enabled soil and crop health monitoring

with intelligent AI-based analytics for intelligent decision-making in agriculture. Unlike existing approaches,

the proposed model utilizes edge-competing, time-series learning, and explainable AI to give dynamic

inherent insight into soil health, crop growth trends, and estimate productivity. The system is also intended to

be effective in low-connectivity rural areas, employing sensor fusion, weather data, and GPS-linked

environmental profiling to offer recommendations at the local level. With the built-in calibration modules, the

architecture guarantees the accuracy, scalability and sustainability in such a way that the farmers can improve

the productivity while saving resources.

1 INTRODUCTION

As the backbone of many economies, agriculture is

more and more utilizing smart technologies to address

contemporary issues associated with food security,

resource management and climate change.

Conventional farming practices are manual

observation based and general decisions are taken

and it's improper that leads to inefficiency and

unpredictability. The development of smart

agriculture, based on the combination of Internet of

Things (IoT) and Artificial Intelligence (AI), has

created opportunities to convert the agricultural field

into a data-oriented and responsive sector. IoT-based

devices are used for a constant monitoring of the

conditions of the crops and soil in real-time, and the

AI’s algorithms work on large amount of data to

provide analysis, identify anomalies and predict

results with great accuracy. In this paper, we propose

a holistic, adaptive framework for smart agriculture

that integrates these tools and empowers informed

decision-making, supportive automation of

monitoring, and improved yield prediction. Through

concentrating on online adaptability, multi-source

data fusion, and explana-ble intelligence, the

framework is designed to fulfill a wide range of needs

from farmers in different environmental scenarios for

productivity, sustainability and adoption.

2 PROBLEM STATEMENT

Current farming practices, as well as the urgent

requirement of sustainable and efficient agriculture

systems, traditional farming practice has long way to

go due to the challenges related with real-time

monitoring, precision decision making, and forecast

of accurate yield. Current models have not generally

succeeded in integrating in-situ data coming from

ground level sensors with intelligent tools, leading to

Pulicherla, P., Priyanka, P. M., Koti, V. M., Sushma, B., Shanmugapriya, K. and Poornima, B.

An Integrated IoT and AI Framework for Real-Time Crop Monitoring, Adaptive Soil Analysis and Intelligent Yield Prediction.

DOI: 10.5220/0013866300004919

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 1st International Conference on Research and Development in Information, Communication, and Computing Technologies (ICRDICCT‘25 2025) - Volume 1, pages

389-395

ISBN: 978-989-758-777-1

389

ununiform estimations of crop health and unefficient

use of resources. Furthermore, since the single-

solution-approach is not scalable and not amenable to

alternative climatic or soil conditions, it diminishes

its applicability in rural settings with limited or no

connectivity. A smart, holistic and real-time

platform for merging IOT sensing with AI analytics

to deliver dynamic insights into crop growth, soil

health and yield outcomes, enabling farmers to make

intelligent and localized decisions upon agriculture, is

thus urgently required.

3 LITERATURE SURVEY

The combination of IoT and AI for agriculture has

attracted a great deal of interest, with applications

targeting improvement of different farming related

tasks such as monitoring, prediction, and

automation. Banerjee et al. (2025): examined the

utilization of digital twins in precision agriculture and

emphasized the non-existence of real-time validation

and affordability for a massive deployment. Bassine

et al. (2023) highlighted that machine learning and

remote sensing techniques were readily applied to

yield forecasting, however there was a dearth of

focus on comparative performance benchmarking.

Fuentes-Peñailillo et al. (2024) modelled a satellite-

based soil crop interaction but did not integrate real-

time sensor data for dynamic updates. Garg et al.

(2021) developed a multimodal precision agriculture

system with IoT and machine learning, however,

scalability in field conditions was not considered.

Ikram (2022) developed intelligent decision system

for crop yield, but it depended on the continuous

internet connectivity, that prevented it for rural

deployment.

Kim et al. (2025) described IoT and AI use cases

in resource-limited settings but did not consider

environmental sensing like soil testing. Kumar and

Sharma (2025) provided monitoring tools for real-

time but didn't used the historical yield power of the

recognition. Li et al. den Toom et al. (2025)

confirmed a promising IoT-AI combination, but their

model yielded suboptimal performance when testing

with mixed-crop datasets. AR and AI based

agricultural monitoring is proposed by Mishra et al.

(2025) without mobile edge deployment support.

Patel et al. (2023) presented an autonomous robot for

crop monitoring that was not integrated with nutrient

sensing and multispectral imaging, though.

Rao and Mehta (2025) concentrated on joint crop

recommendation, and they did not provide for long-

term sensor calibration strategies. Sharma and Verma

(2024) proposed a run-time sensing platform, but it

does not handle sensor heterogeneity. The attempt of

Singh and Kumar (2025) was only review based

study; they did not verify their inferences empirically.

Smith et al. (2025) provided a complete review on AI

and IoT in smart farming without deep technical

indicators. Real time crop prediction based on soil

sensor was suggested by thakur and patel (2024)

where their model did not allow multi calibration of

crop.

Verma and Gupta (2025) developed an AI and

IoT-enabled smart agriculture, but it failed to

emphasize the sustainable resource management and

pesticide recommendation. Wang et al. (2025)

focused on productivity enhancement through AI and

didn’t provide a full deployment-ready model.

Wilberforce and Mwebaze (2025) came up with a

theory of IoT framework (Committee on Internet of

Things Framework for Agriculture Technologies 6-

17) for agriculture that was not validated in a real

field situation. Yadav & Singh, (2024) had worked on

IoT based crop yield prediction and they did not

infuse advanced forecasting models. Zhang et al.

(2025) employed XAI for smart farming with no

seasonal adaptation in their model. Finally, Zhou and

Li (2025) explored the AI component of agriculture

practice but without any feedback mechanism for

improving soil quality.

Together; these studies demonstrate the trend of

integration of smart technologies in the agriculture;

however; they also claimed the existence of gaps in

the areas of integration, flexibility, and real-time

decision support. To overcome these limitations, the

objectives of the proposed work will be to develop a

seamless and deployable smart agriculture

infrastructure that binds real-time sensing with

adaptive analytics for the optimum growing

environment.

4 METHODOLOGY

The smart agriculture system presented consists of

an integrated Internet of Things-enabled sensing,

edge computing, AI-based analytics, and cloud

synchronization solution that is specifically designed

to serve as a combined, scalable architecture for real-

time crop/soil monitoring and precise yield

prediction. This approach starts with the installation

of low power, low cost IoT sensors in the agricultural

field that continuously acquire information about it

like the soil moisture, temperature, soil pH, ambient

weather conditions and the health status of the crops.

These

sensors are installation in specific locations

ICRDICCT‘25 2025 - INTERNATIONAL CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION,

COMMUNICATION, AND COMPUTING TECHNOLOGIES

390

Table 1: Sensor specifications and deployment parameters.

Sensor Type

Parameter

Measured

Measurement Range Accuracy

Deployment

Depth/Height

Soil Moisture

Volumetric Water

Content

0–100% ±2% 5 cm below surface

pH Sensor

Soil

Acidit

/Alkalinit

pH 3–10 ±0.1 10 cm below surface

Temperature

Sensor

Soil and Air

Temperature

-40°C to 85°C ±0.5°C 2 m above surface

Humidity Sensor

Ambient

Humidity

0–100% RH ±2% RH 2 m above surface

Light Sensor Light Intensity 0–100,000 Lux ±5% 2 m above surface

according to geospatial map analysis and soil

sampling profiles, in order to cover large areas and

obtain representative data.

Sensor data is transmitted through a local gateway

system, which supports edge computing capabilities

for preliminary data preprocessing. This allows the

system to operate efficiently even in regions with

limited internet connectivity, enabling real-time

filtering, noise reduction, and early-stage anomaly

detection. By performing data normalization and

cleaning at the edge, the framework significantly

reduces bandwidth usage and computational burden

on the central server.

Table 1 shows the Sensor

Specifications and Deployment Parameters.

Figure. 1: Field-wise heatmap of sensor deployment and

soil moisture variability.

Once preprocessed, the data is synchronized with

a cloud-based platform where AI models are

deployed. A hybrid learning approach is used that

combines supervised and time-series deep learning

models to analyze environmental conditions and

predict crop yield. Convolutional Neural Networks

(CNNs) are used for image-based leaf and crop

condition analysis when drone or mobile-based image

inputs are available, while Long Short-Term Memory

(LSTM) models handle temporal prediction tasks

related to soil degradation, water stress, and

forecasted yield. Additionally, explainable AI

techniques such as SHAP (SHapley Additive

exPlanations) are incorporated to interpret the

influence of each input variable on the output

predictions, enabling transparency and trust in the

system’s decisions.

Figure. 1 shows the Field-wise

Heatmap of Sensor Deployment and Soil Moisture

Variability.

In addition to current weather forecast, it also

considers historical agricultural datasets, which is

useful for enhancing the reliability of the yield

prediction model. Environmental behaviours,

including precipitation outlooks, temperature

differences and wind, are monitored and combined

with the data from sensors to develop adaptable

farming plans. Using this information, the model

generates personalized advice on irrigation times,

application of agrochemicals, optimal time of

harvesting and fertilization.

Figure 2 shows the

Workflow of the Proposed IoT-AI Agriculture

Framework.

An Integrated IoT and AI Framework for Real-Time Crop Monitoring, Adaptive Soil Analysis and Intelligent Yield Prediction

391

Figure 2: Workflow of the proposed IoT-AI agriculture

framework.

Table 2: Edge Vs Cloud Processing Performance

Comparison.

Metric

Edge

Processin

Cloud

Processin

Avera

e Latenc

(

)

120 220

Power Consumption

(W)

3.5 7.8

Data Transmission

Load (MB)

30 75

Uptime Availability

(

)

98.2 96.7

Offline Functionality Supported

Not

orte

To ensure continuous learning and adaptability,

the framework includes a feedback loop where

farmers can validate or reject the AI's

recommendations, and these interactions are recorded

and used to refine the model. This human-in-the-loop

design ensures that the model evolves with local

practices, crop types, and seasonal variations.

Table

2 shows the Edge vs Cloud Processing Performance

Comparison.

All user interactions, sensor logs, and prediction

outcomes are securely stored in a distributed cloud

database with role-based access, supporting remote

access via mobile or web applications. A simple

dashboard visualizes real-time sensor data, AI

predictions, and recommended actions, enabling

users with varying digital literacy levels to make

informed decisions.

Overall, the proposed methodology offers a

holistic, modular, and intelligent system that

leverages the strengths of IoT, AI, and edge-cloud

synergy to support modern agriculture. By closing the

gap between data acquisition and actionable insight,

this framework empowers farmers to optimize

productivity, preserve resources, and ensure food

security in the face of environmental uncertainties.

5 RESULT AND DISCUSSION

The proposed IoT and AI-based smart agriculture

framework was tested in a real-time field setting with

a variety of crops and patches of soils. The findings

showed that the system could effectively administer,

interpret and forecast agricultural parameters, thus

greatly improving farmer?s decision-making support.

The thorough examination with respect to the sensor

accuracy, model efficiency, and system adaptability

demonstrate the overall efficiency and robustness of

the proactive 5G-SDM reconfiguration scheme.

The IoT sensors deployed in the field were able

to gather continuous data of soil moisture, pH,

temperature, and humidity successfully. The sensors

compared well to laboratory standard instruments,

with variance of less than ±2 %, and were therefore

considered appropriate for field applications.

Moreover, the sensor calibration module integrated in

the framework facilitates automatic setting according

to environment feedback, thus keeping the data

collection in long-term linearity. The edge gateway

system successfully executed well under limited

bandwidth, with a data compression ratio of 3:1 that

preserved sensor reading fidelity. This solution made

sure to send and process only necessary data with

minimal consumption of network resources.

Table 3

shows the AI Model Evaluation Metrics for Yield

Prediction.

Table 3: AI model evaluation metrics for yield prediction.

Model Type

MAE

(%)

RMSE

(%)

MAPE

(%)

R²

Scor

Linear

ression

8.3 9.2 13.7 0.82

Decision

Tree

7.1 8.7 12.1 0.85

LSTM

(

Pro

osed

)

4.4 6.3 6.5 0.91

ICRDICCT‘25 2025 - INTERNATIONAL CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION,

COMMUNICATION, AND COMPUTING TECHNOLOGIES

392

In terms of computation, the edge processing cut

decision-making delay by up to about 45 percent

versus cloud-only models. It was immensely useful

for providing timely alerts for irrigation and disease

identification in areas which were not well connected.

Another study combined image-based crop health

analysis with CNN to achieve average detection

accuracy of 94.3% on five crop varieties. The models

were able to accurately detect signs of leaf spot,

blight, and pest infestations in drone imagery and

smartphone photos uploaded by farmers, helping then

take early action to prevent extensive crop damage.

Figure 3 shows the Yield Prediction Accuracy Across

Different AI Model.

Figure 3: Yield prediction accuracy across different AI

models.

Table 4: CNN model performance for disease classification.

Crop Type Accuracy (%) Precision (%) Recall (%) F1-Score (%)

Tomato 95.4 94.1 96.2 95.1

Wheat 93.7 92.5 94.6 93.5

Maize 92.3 90.8 93.1 91.9

Potato 91.5 89.7 92.4 91.0

Overall Avg. 94.3 91.8 94.1 93.6

The yield prediction module is based on LSTM

model and is trained on a dataset that combines real-

time sensor data, meteorological forecast and

historical productivity of crops. Error rate was run

against traditional linear regression and decision tree

models that resulted in rates of >12% whereas the

model performed at a mean absolute percentage error

(MAPE) of 6.5%. This evidence a high degree of

generalization to temporal trends and context and

seasonality specific predictions. Further, the model

was robust to response to unanticipated changes in

weather that have not been included in the model

input including late monsoons and heatwaves.

Table

4 shows the CNN Model Performance for Disease

Classification.

Figure 4: CNN classification performance for crop disease

detection.

Explain ability was another important strong point

of the framework. SHAP-based interpretations helped

farmers see visual explanations for why a

recommendation was made, including why irrigation

was postponed, or why one fertilizer was

recommended. This built user trust and promoted

farmer interaction with the AI system. The feedback

loop is another indispensable factor in the success of

this study, as more than 82% of farmer feedback

matched system recommendation, and the remaining

18% helped to further calibrate the model.

Figure 4

shows the CNN Classification Performance for Crop

Disease Detection.

Table 5: Resource optimization through AI-based

recommendations.

Resource

Type

Traditional

Usage

Smart

System

Usage

Reduction

(%)

Water

(L/month)

48,000 37,400 22%

Fertilizer

(kg)

520 440 15.4%

Pesticides

(L)

112 93 17%

Labor

Hours

180 128 28.8%

Yield

Increase

– +17% –

An Integrated IoT and AI Framework for Real-Time Crop Monitoring, Adaptive Soil Analysis and Intelligent Yield Prediction

393

Figure 5: Traditional vs AI-guided resource usage.

A mobile application was designed as the user

interface in providing the accessibility to the system.

Novice users too could navigate the dashboard and

seek important information. The app showed real-

time data visualisations, alert signals, and tailored dos

and don'ts for farmers. Language localization and

voice support were included to meet regional

requirements and help increase user satisfaction.

Table 5 shows the Resource Optimization Through

AI-Based Recommendations.

Field trials showed, on average, 17% better yield

compared to replicates that receive traditional

practice independent of technological augmentation.

61—in North Carolina with improved irrigation

scheduling using real-time soil moisture information.

Category ATRB Volume 49 2006 was AI-driven

precision application of fertilizers and pesticides also

saved the farmer 15% in costs and was better for the

environment.

Figure 5 shows the Traditional vs AI-

Guided Resource Usage.

The system demonstrated the scalability

dimension, and it was successfully tested in various

plots that covered wheat, tomato, and maize varieties.

The modularity of this framework facilitated the

transference to other environments by replacing crop

models and sensors locations. In addition, the live

alerts for insect infestations and water stress reduced

crop loss, the bane of traditional agriculture.

In conclusion, the developed framework

outperformed other approaches to combining IoT

and AI in real life farming applications. Its low

connectivity requirements, interpretability through

explainable AI, and ability to be adapted to new

plants and new climates highlight this tool as a

potentially game-changing technology for precision

agriculture. The findings validate the benefits of

intelligent farming technologies in reality, and pave

the way for its larger-scale adoption within rural and

semi-urban farming communities. This work has

important implications not only for technological

progress in agriculture, but also resonates with global

agendas for sustainable agriculture, natural resource

preservation, and food security.

6 CONCLUSIONS

The progress and realization of an IoT and AI-based

smart agriculture have shown promising prospects for

the transition of traditional agriculture system to a

data-oriented, learning, and smart system. Utilizing

real-time data from sensors and AI models, the

system allows for accurate crop stats, soil health

analysis and yield prediction in harsh rural areas with

low connectivity. The combination of edge

computing, time-series forecasting and explainable

AI means that this intelligence can be timely, trust-

worthy and transparent, so making recommendations

that can be acted upon. Experimental results

demonstrate the system's potential to increase crop

yield, resource utilization, and decision-making

capable of promoting scalable and sustainable

agriculture. This study provides a practical insight

into the future of precision farming when farmers use

smart technologies to adapt their farming operations

ahead of environmental stresses and optimize both the

productivity and sustainability of production.

REFERENCES

Banerjee, S., Mukherjee, A., & Kamboj, S. (2025).

Precision agriculture revolution: Integrating digital

twins and advanced crop recommendation for optimal

yield. arXiv. https://arxiv.org/abs/2502.04054arXiv

Bassine, F. Z., Epule, T. E., Kechchour, A., & Chehbouni,

A. (2023). Recent applications of machine learning,

remote sensing, and IoT approaches in yield prediction:

A critical review. arXiv. https://arxiv.org/abs/2306.0

4566arXiv

Fuentes-Peñailillo, F., et al. (2024). Soil and crop

interaction analysis for yield prediction with satellite

data. Science of The Total Environment. https://www

.sciencedirect.com/science/article/pii/S030147972501

0710ScienceDirect

Garg, S., Pundir, P., Jindal, H., Saini, H., & Garg, S. (2021).

Towards a multimodal system for precision agriculture

using IoT and machine learning. arXiv.

https://arxiv.org/abs/2107.04895arXiv

Ikram, A. (2022). Crop yield maximization using an IoT-

based smart decision system. Hindawi. https://onlineli

brary.wiley.com/doi/10.1155/2022/2022923Wiley

Online Library

Kim, J., et al. (2025). IoT and AI for smart agriculture in

resource-constrained environments. Journal of

ICRDICCT‘25 2025 - INTERNATIONAL CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION,

COMMUNICATION, AND COMPUTING TECHNOLOGIES

394

Ambient

Intelligence and Humanized Computing. https://link.s

pringer.com/article/10.1007/s43926-025-00119-3

SpringerLink

Kumar, R., & Sharma, A. (2025). IoT-driven smart

agricultural technology for real-time soil and crop

monitoring. Smart Agricultural Technology. https://w

ww.sciencedirect.com/science/article/pii/S277237552

5000802ScienceDirect

Li, Y., et al. (2025). Integrating artificial intelligence and

Internet of Things (IoT) for precision agriculture.

Computers and Electronics in Agriculture. https://ww

w.sciencedirect.com/science/article/pii/S26663511240

00147ScienceDirect

Mishra, P., & Singh, V. (2025). Smart agriculture

monitoring system using AR and AI. International

Research Journal of Modernization in Engineering

Technology and Science. https://www.irjmets.com/up

loadedfiles/paper//issue_3_march_2025/70350/final/fi

n_irjmets1743075253.pdfIRJMETS+1IRJMETS+1

Patel, M., Gupta, N., & Rao, K. (2023). Autonomous

agricultural robots for crop monitoring and yield

prediction. IEEE International Conference on Robotics

and Automation (ICRA). https://www.irjmets.com/upl

oadedfiles/paper//issue_2_february_2025/67983/final/

fin_irjmets1740288949.pdfIRJMETS

Rao, S., & Mehta, D. (2025). Integrated IoT approaches for

crop recommendation and yield prediction. Sensors.

https://www.mdpi.com/2624-831X/5/4/28MDPI

Sharma, L., & Verma, K. (2024). Smart IoT-driven

precision agriculture: Land mapping, crop monitoring,

and yield prediction. Sensors. https://pmc.ncbi.nlm.ni

h.gov/articles/PMC11918389/PMC

Singh, A., & Kumar, P. (2025). IoT-cloud based intelligent

system for crop yield prediction: A review.

ResearchGate https://www.researchgate.net/publicatio

n/390427177_IoTCloud_based_Intelligent_System_fo

r_Crop_Yield_Prediction_A_ReviewResearchGate

Smith, J., et al. (2025). Smart farming using AI, IoT, and

remote sensing. Spectroscopy Online. https://www.sp

ectroscopyonline.com/view/smart-farming-using-ai-

iot-and-remote-sensingSpectroscopy Online

Thakur, R., & Patel, S. (2024). An IoT-enabled real-time

crop prediction system using soil sensors. Journal of

Sensor and Actuator Networks. https://www.mdpi.co

m/2673-4117/5/4/130MDPI

Verma, S., & Gupta, R. (2025). Smart agriculture crop yield

growth using AI and IoT. International Research

Journal of Modernization in Engineering Technology

and Science. https://www.irjmets.com/uploadedfiles/p

aper//issue_2_february_2025/67983/final/fin_irjmets1

740288949.pdf

Wang, H., et al. (2025). Artificial intelligence in

agriculture: Advancing crop productivity and

sustainability. Computers and Electronics in

Agriculture. https://www.sciencedirect.com/science/a

rticle/pii/S2666154325001334ScienceDirect

Wilberforce, N., & Mwebaze, J. (2025). A framework for

IoT-enabled smart agriculture. arXiv.

https://arxiv.org/abs/2501.17875arXiv

Yadav, R., & Singh, M. (2024). Predicting crop yields using

IoT. BIO Web of Conferences. https://www.bio-

conferences.org/articles/bioconf/pdf/2024/01/bioconf_

msnbas2024_05009.pdfBio Conferences

Zhang, L., et al. (2025). Next-gen agriculture: Integrating

AI and XAI for precision crop yield prediction.

Frontiers in Plant Science. https://www.frontiersin.org

/journals/plant- science/articles/10.3389/fpls.2024.145

1607/fullFrontiers

Zhou, Q., & Li, M. (2025). How AI empowers crop

monitoring and yield prediction. Agricultural

Engineering. https://hsg-ame.com/ame-blog/f/how-ai-

empowers-crop-monitoring-and-yield-predictionHSG-

AME Certified Laboratories

An Integrated IoT and AI Framework for Real-Time Crop Monitoring, Adaptive Soil Analysis and Intelligent Yield Prediction

395