Muhammad Jawad Bashir

Rationale, Aims and ObjectivesModern smart farming systems increasingly rely on Artificial Intelligence (AI) to automate monitoring, climate control and crop management. However, agricultural AI systems often face two major challenges: limited high-quality farming data for diverse scenarios and lack of transparency in AI-driven decisions. This research aimed to develop an explainable AI framework capable of generating realistic synthetic farming data while also improving trust in automated decision-support systems. The study introduced a novel Conditional Time-Transformer model that generates context-aware greenhouse data using environmental conditions such as weather, time patterns and operational settings, while integrating explainable AI methods to assess how AI systems make decisions. Research FindingsThe proposed AI framework improved the quality and realism of synthetic greenhouse data compared to existing generative AI approaches. The model achieved stronger alignment with real-world greenhouse conditions and improved downstream decision-making performance for greenhouse control tasks. Importantly, the research demonstrated that explainable AI techniques can help evaluate whether AI systems trained on synthetic data make decisions in ways that remain consistent with real farming behaviour. The findings showed that combining conditional generative AI with explainability can support more transparent, reliable and trustworthy smart farming systems while reducing dependence on expensive long-term agricultural data collection. Policy ImplicationsThe findings highlight the need for governance frameworks that promote transparency and accountability in AI-driven agricultural technologies. Policymakers should encourage the use of explainable AI validation methods before deploying AI systems in high-impact farming operations. The research also supports public investment in trustworthy AI infrastructure for agriculture, particularly for climate-smart and resource-efficient farming systems. Synthetic data generation may help overcome agricultural data shortages while preserving operational privacy and reducing deployment costs. However, regulatory guidelines should require ongoing validation of AI reasoning and decision consistency to ensure safe and reliable adoption within real-world agricultural environments. Industry RecommendationsAgricultural technology providers and greenhouse operators should integrate explainable AI tools into smart farming platforms to improve transparency and farmer trust. Companies deploying AI-based greenhouse automation should validate AI decisions using both performance metrics and explainability assessments before operational use. The use of synthetic data generation can help organisations train AI systems under rare or extreme environmental conditions without waiting for long-term field data collection. Industry stakeholders should also prioritise human-in-the-loop monitoring for AI-assisted climate-control decisions, particularly in safety-critical situations where incorrect predictions could affect crop productivity, resource usage or operational stability.

Smart farming systems offer significant potential to address food security challenges through data-driven and sustainable decision support. However, the deployment of AI-driven solutions is constrained by limited labelled data and insufficient transparency, hindering trust and adoption. Generative AI has been used to mitigate data scarcity, but challenges remain in generating context-aware multivariate time-series under diverse conditions and in ensuring transparency when synthetic data is used for downstream modelling. To address these issues, we propose a Conditional Time-Transformer (CTT), a conditional generative model that enables controlled synthesis via explicit conditioning on external contextual variables. The model is integrated into a decision-support workflow and evaluated using a Train-on-Synthetic, Test-on-Real protocol. Explainable AI techniques are incorporated to analyse global and local feature-importance shifts in downstream models trained on synthetic data. Results show improved fidelity (MMD 0.19, DTW 20.60, FID 0.23) and enhanced classification performance over baselines, supporting robust and interpretable smart farming systems.

In the evolving landscape of traffic management and vehicle surveillance, efficient license plate detection and recognition are indispensable. Historically, many methodologies have tackled this challenge, but consistent real-time accuracy, especially in diverse environments, remains elusive. This study examines the performance of YOLOv8 variants on License Plate Recognition (LPR) and Character Recognition tasks, crucial for advancing Intelligent Transportation Systems. Two distinct datasets were employed for training and evaluation, yielding notable findings. The YOLOv8 Nano variant demonstrated a precision of 0.964 and mAP50 of 0.918 on the LPR task, while the YOLOv8 Small variant exhibited a precision of 0.92 and mAP50 of 0.91 on the Character Recognition task. A custom method for character sequencing was introduced, effectively sequencing the detected characters based on their x-axis positions. An optimized pipeline, utilizing YOLOv8 Nano for LPR and YOLOv8 Small for Character Recognition, is proposed. This configuration not only maintains computational efficiency but also ensures high accuracy, establishing a robust foundation for future real-world deployments on edge devices within Intelligent Transportation Systems. This effort marks a significant stride towards the development of smarter and more efficient urban infrastructures.

The integration of the Internet of Robotic Things (IoRT) in smart greenhouses has revolutionised precision agriculture by enabling efficient and autonomous environmental control. However, existing time series forecasting models in such setups often operate as black boxes, lacking mechanisms for explainable decision-making, which is a critical limitation when trust, transparency, and regulatory compliance are paramount in smart farming practices. This study leverages the Temporal Fusion Transformer (TFT) model to automate actuator settings for optimal greenhouse management. To enhance interpretability and trust in the model decision-making process, both local and global explanation techniques were employed using model-inherent interpretation, local interpretable model-agnostic explanations (LIME), and SHapley additive explanations (SHAP). These explainability methods provide information on how different sensor readings, such as temperature, humidity, CO2 levels, light, and outer climate, contribute to actuator control decisions in an automated greenhouse. The trained TFT model achieved a test accuracy of 95% on a class-imbalanced dataset for actuator control settings in an automated greenhouse environment. The results demonstrate the varying influence of each sensor on real-time greenhouse adjustments, ensuring transparency and enabling adaptive fine-tuning for improved crop yield and resource efficiency.

The rapid evolution of Artificial intelligence in healthcare has opened avenues for enhancing various processes, including medical billing and transcription. This paper introduces an innovative approach by integrating AI with Locally Linear Embedding (LLE) to revolutionize the handling of high-dimensional medical data. This AI-enhanced LLE model is specifically tailored to improve the accuracy and efficiency of medical billing systems and transcription services. By automating these processes, the model aims to reduce human error and streamline operations, thereby facilitating faster and more accurate patient care documentation and financial transactions. This paper provides a comprehensive mathematical model of AI-enhanced LLE, demonstrating its application in real-world healthcare scenarios through a series of experiments. The results indicate a significant improvement in data processing accuracy and operational efficiency. This study not only underscores the potential of AI-enhanced LLE in medical data analysis but also sets a foundation for future research into broader healthcare applications.

Mung bean (Vigna radiata), or Green gram, is a widely cultivated legume from the Fabaceae family, vulnerable to both biotic and abiotic stresses. Two major diseases, Mung bean Yellow Mosaic Virus (MYMV) and Cercospora Leaf Spot (CLS), significantly hinder its yield. Timely detection and accurate classification are therefore critical for effective management. This study introduces an integrated precision agriculture approach combining remote sensing and deep learning for Mung bean health monitoring. Three curated datasets were employed: a 15 day temporal multispectral dataset from drone surveys, the MYMV Segmentation dataset (MYMV S), and the Mung bean Disease Classification dataset (MDC), which facilitates the classification of CLS, MYMV, and healthy leaves. Advanced models were tested, including YOLOv8 and Mask R CNN for segmentation, and DEIT and MobileNetV2 for classification. The YOLOv8 Nano variant excelled with an AP@50 above 77 percent, 94.9 percent accuracy, and 82.13 percent mIOU, demonstrating suitability for real time edge deployment. For classification, DEIT achieved 99 percent accuracy. In addition, the Normalized Difference Vegetation Index (NDVI) proved to be the most reliable indicator for overall crop health. The findings offer scalable solutions for disease detection, mapping, and targeted pesticide use, advancing sustainable agriculture. More broadly, they underscore the transformative role of computer vision and remote sensing in enhancing global agricultural practices and strengthening food security.

The integration of the Internet of Robotic Things (IoRT) in smart greenhouses has revolutionised precision agriculture by enabling efficient and autonomous environmental control. However, existing time series forecasting models in such setups often operate as black boxes, lacking mechanisms for explainable decision-making, which is a critical limitation when trust, transparency, and regulatory compliance are paramount in smart farming practices. This study leverages the Temporal Fusion Transformer (TFT) model to automate actuator settings for optimal greenhouse management. To enhance interpretability and trust in the model decision-making process, both local and global explanation techniques were employed using model-inherent interpretation, local interpretable model-agnostic explanations (LIME), and SHapley additive explanations (SHAP). These explainability methods provide information on how different sensor readings, such as temperature, humidity, CO2levels, light, and outer climate, contribute to actuator control decisions in an automated greenhouse. The trained TFT model achieved a test accuracy of 95% on a class-imbalanced dataset for actuator control settings in an automated greenhouse environment. The results demonstrate the varying influence of each sensor on real-time greenhouse adjustments, ensuring transparency and enabling adaptive fine-tuning for improved crop yield and resource efficiency.

Agriculture is vital to economic growth, contributing 4% to global GDP and over 25% in some developing countries. Most farming practices are outdated, necessitating modernization for improved efficiency. Advances in deep learning, multi- and hyperspectral imagery (MHSI), UAVs, and agri-bots have revolutionized precision agriculture (PA). Computer vision (CV) techniques, enhanced by MHSI, have automated tasks like crop classification, disease monitoring, and biomass estimation. UAVs assist in field scouting, disease detection, and precision spraying, while agri-bots with IoT sensors facilitate real-time data-driven actions such as fruit picking and weed control. This chapter reviews the latest developments in CV, MHSI, UAVs, and agri-bots, examining current methods, challenges, datasets, and future applications in precision agriculture.

The rapid evolution of Artificial intelligence in healthcare has opened avenues for enhancing various processes, including medical billing and transcription. This paper introduces an innovative approach by integrating AI with Locally Linear Embedding (LLE) to revolutionize the handling of high-dimensional medical data. This AI-enhanced LLE model is specifically tailored to improve the accuracy and efficiency of medical billing systems and transcription services. By automating these processes, the model aims to reduce human error and streamline operations, thereby facilitating faster and more accurate patient care documentation and financial transactions. This paper provides a comprehensive mathematical model of AI-enhanced LLE, demonstrating its application in real-world healthcare scenarios through a series of experiments. The results indicate a significant improvement in data processing accuracy and operational efficiency. This study not only underscores the potential of AI-enhanced LLE in medical data analysis but also sets a foundation for future research into broader healthcare applications.

In the evolving landscape of traffic management and vehicle surveillance, efficient license plate detection and recognition are indispensable. Historically, many methodologies have tackled this challenge, but consistent real-time accuracy, especially in diverse environments, remains elusive. This study examines the performance of YOLOv8 variants on License Plate Recognition (LPR) and Character Recognition tasks, crucial for advancing Intelligent Transportation Systems. Two distinct datasets were employed for training and evaluation, yielding notable findings. The YOLOv8 Nano variant demonstrated a precision of 0.964 and mAP50 of 0.918 on the LPR task, while the YOLOv8 Small variant exhibited a precision of 0.92 and mAP50 of 0.91 on the Character Recognition task. A custom method for character sequencing was introduced, effectively sequencing the detected characters based on their x-axis positions. An optimized pipeline, utilizing YOLOv8 Nano for LPR and YOLOv8 Small for Character Recognition, is proposed. This configuration not only maintains computational efficiency but also ensures high accuracy, establishing a robust foundation for future real-world deployments on edge devices within Intelligent Transportation Systems. This effort marks a significant stride towards the development of smarter and more efficient urban infrastructures.

Transitioning to green energy transport systems, notably electric vehicles, is crucial to both combat climate change and enhance urban air quality in developing nations. Urban air quality is pivotal, given its impact on health, necessitating accurate pollutant forecasting and emission reduction strategies to ensure overall well-being. This study forecasts the influence of green energy transport systems on the air quality in Lahore and Islamabad, Pakistan, while noting the projected surge in electric vehicle adoption from less than 1% to 10% within three years. Predicting the impact of this change involves analyzing data before, during, and after the COVID-19 pandemic. The lockdown led to minimal fossil fuel vehicle usage, resembling a green energy transportation scenario. The novelty of this work is twofold. Firstly, remote sensing data from the Sentinel-5P satellite were utilized to predict air quality index (AQI) trends before, during, and after COVID-19. Secondly, deep learning models, including long short-term memory (LSTM) and bidirectional LSTM, and machine learning models, including decision tree and random forest regression, were utilized to forecast the levels of NO2, SO2, and CO in the atmosphere. Our results demonstrate that implementing green energy transportation systems in urban centers of developing countries can enhance air quality by approximately 98%. Notably, the bidirectional LSTM model outperformed others in predicting NO2 and SO2 concentrations, while the LSTM model excelled in forecasting CO concentration. These results offer valuable insights into predicting air pollution levels and guiding green energy policies to mitigate the adverse health effects of air pollution.