Nagalingam Rajeswaran

As smart sensing applications grow rapidly, the IoT edge architectures need to support low latency, make decisions with little memory, power, and communication resources while preserving the privacy of the data. But traditional cloud-based IoT solutions come with transmission delay, increased energy consumption, and privacy issues because of the constant transfer of raw data. In this paper, we propose a privacy-preserving TinyML-driven IoT edge architecture that enables real-time smart sensing and autonomous AI-based control. The proposed framework includes on-device sensor pre-processing, lightweight TinyML inference, adaptive model selection, encrypted feature-level communication, trust-aware decision validation, and local control execution. Raw data streams from the sensors are also processed locally, and only compact encrypted features or a summary of the decisions are sent if necessary to minimize privacy exposure. The experimental evaluation reveals that the proposed architecture has an accuracy of 97.4%, an F1 score of 96.9%, and a secure-event detection rate of 98.1% and reduces the inference latency by 14.8%, the energy consumption by 4.1 mJ per inference and the amount of data transmitted by 74.5% compared to conventional edge-cloud processing. Results show that the proposed architecture is a scalable, privacy-aware, and energy-efficient solution for real-time autonomous IoT control of smart environments with limited resources.

The growing deployment of Internet of Things (IoT) monitoring systems has resulted in demands for low-latency, secure, and energy-efficient intelligence on embedded chips. But most cloud-based and edge-assisted solutions are prone to high communication latency, lack adaptability, consume more energy, and lack decision security under resource-limited conditions. This paper introduces an Edge-Intelligent IoT Framework for Real-Time Adaptive Monitoring and Trust-Aware Secure Decision Validation with Resource-Aware Artificial Intelligence and Machine Learning (AI/ML) on embedded chips. Unlike conventional TinyML or Edge AI deployments that use a fixed inference model, the proposed framework introduces a validation-calibrated adaptive inference mechanism that jointly considers chip resources, input complexity, and sensor trust before accepting an embedded decision. The main scientific contribution is the unified coupling of resource-aware model selection with trust-aware decision validation for low-power embedded IoT inference. The framework dynamically selects the inference path and validates sensor trust before decision acceptance. Through experimentation, the proposed framework is demonstrated with 97.2% accuracy, 96.4% F1-score, and 98.1% AUROC, and 40.4% lower inference latency (31.2 ms to 18.6 ms) and 39.6% lower energy (9.6 mJ to 5.8 mJ) compared with traditional TinyML deployment. These results were obtained using the MHEALTH wearable IoT dataset with a leakage-safe 70:15:15 split and were statistically validated across five independent runs. The findings demonstrate a promising resource-aware TinyML-style embedded inference pipeline for wearable IoT monitoring, with improved latency-energy efficiency and trust-aware decision validation under the evaluated settings.

This paper introduces TempNet, a groundbreaking multi-scale attention-based deep learning architecture that achieves unprecedented accuracy in operational renewable energy forecasting. The proposed hybrid model integrates multi-scale convolutional neural networks with different kernel sizes of {1, 3, 5, 7} for temporal pattern extraction, bidirectional LSTM layers with 128 units each, and sophisticated attention mechanisms to address the fundamental challenges of renewable energy variability. In a comprehensive evaluation on 18 months of high-resolution renewable energy data (44,166 observations), TempNet achieved exceptional performance with near-perfect forecasting metrics (RMSE of 0.44, MAE of 0.1, R2 of 0.99), showing significant improvements compared to baseline models. The proposed architecture incorporates several critical innovations, including a robust preprocessing pipeline with IQR-based outlier detection affecting 4.2% of observations, cyclical encoding for temporal features, and advanced data augmentation strategies that increased training samples by 25%. Cross-validation analysis across temporal folds demonstrates consistent performance with a coefficient of variation below 3% for all metrics, while bootstrap validation confirms statistical significance with 95% confidence intervals. Comparing TempNet with previous methods shows its superiority. The model maintains exceptional accuracy across diverse operational scenarios, including extreme weather events where maximum absolute error remained below 2.3% of mean generation capacity, significantly outperforming the 8-12% typically reported in the literature. TempNet's computational efficiency enables real-time deployment in smart grid environments, with inference times suitable for operational forecasting horizons from 1 to 24 hours. These breakthrough results can help achieve higher renewable penetration levels, reduce reserve requirements, and improve grid operations, supporting the global transition toward sustainable energy infrastructure.

Combining Additive Manufacturing (AM) and Subtractive Manufacturing (SM) technologies holds significant potential for transforming industrial production. However, integrating these two approaches remains challenging due to factors such as process compatibility, material loss, and issues in the design and fabrication processes. This research addresses these challenges using deep learning-based Artificial Intelligence (AI) frameworks to enhance hybrid manufacturing systems. Utilizing Convolutional Neural Networks (CNNs) and Reinforcement Learning (RL) models, the study proposes a feedforward intelligent control model that adapts tool path generation, material usage, and defect recognition in real time. Experimental results on benchmark manufacturing datasets demonstrate that the proposed method achieves a 23% reduction in material wastage and a 15% improvement in efficiency compared with existing hybrid methods. Moreover, defect detection accuracy increased to 98.7%, validating the effectiveness of AI-generated quality assurance tools. Production schedules were also reduced by 12% through efficient design-for-manufacturing integration. These observations highlight that deep learning is pivotal in reconciling additive and subtractive techniques, opening new possibilities for modern, sustainable, accurate, and efficient manufacturing processes.

In recent years, deep learning methods have dramatically improved medical image analysis, though earlier models faced difficulties in capturing intricate spatial and contextual details. These challenges highlighted the necessity for more powerful and flexible models. In this study, we introduce an innovative method for the multi-classification of breast cancer histopathological images utilizing Bidirectional Recurrent Neural Networks (BRNN). The BRNN structure consists of four unique elements: the backbone branch for transfer learning, the Gated Recurrent Unit (GRU), the residual collaborative branch, and the feature fusion module. Specifically, the transfer learning aspect exploits a Convolutional Neural Network (CNN) based on the ResNet50 architecture to draw image features from the ImageNet dataset, enhanced with an attention mechanism for improved feature representation. Additionally, the residual branch identifies specific pathological features using Floating Point Operations (FLOP). This cooperative method ensures thorough extraction of breast cancer classification features. The BRNN model was tested on the BreaKHis breast cancer dataset, comprising 7,909 microscopic images across 8 various classes from 82 patients. Our model achieved an average classification accuracy of 97.25 percent, exceeding current leading techniques. The BRNN model, refined using the Adagrad optimization algorithm, efficiently integrates the learned features from both branches. This tool provides oncologists with a substantial enhancement in diagnosing and planning treatment for breast cancer.

This paper discusses the design and evaluation of a four-element Multiple-Input, Multiple-Output (MIMO) antenna array with Circular Polarization (CP), tailored explicitly for sub-6 GHz 5G wireless systems. The proposed antenna utilizes an elliptical ring slot with a variable width in the ground plane and an asymmetric feed line layout to achieve orthogonal CP radiation in both the forward and backward directions. Three interconnected strip lines are incorporated into the ground plane for enhanced performance, increasing isolation and ensuring effective linkage between the antenna components and their respective grounds. The antenna exhibits strong isolation, a low Envelope Correlation Coefficient (ECC), and minimal Channel Capacity Loss (CCL), making it ideal for MIMO applications. The analysis confirms the antenna's stable CP gain and operational efficiency throughout the sub-6 GHz band, meeting the rigorous standards of contemporary 5G networks.

Emotional Intelligence (EI) constitutes a vital element of human communication, and its integration into text-based dialogues has gained great significance in the modern digital era. The present paper proposes an innovative method for modeling sentiment and emotion within text-based conversations using the ChatGPT language model. The advancements in sentiment and emotion recognition are centered on the role of EI in text-based conversational models. The study underscores the significance of diverse datasets, including Interactive Emotional Dyadic Motion Capture (IEMOCAP), MELD, EMORYNLP, and DAILYDIALOG, for training and evaluating emotion detection algorithms. IEMOCAP and MELD offer detailed emotional annotations, EMORYNLP emphasizes sensitive dialogue scenarios, and DAILYDIALOG encompasses a wide range of everyday interactions, providing distinct advantages for capturing emotional subtleties. The proficiency of different emotion categorization models, including ChatGPT and models with four levels of detail, is demonstrated through their capacity to understand and respond to emotions aptly. The crucial role of conversational AI with sophisticated EI in fostering empathy and context-sensitive interactions is emphasized.

A smart battery electric vehicle control framework is proposed in this paper. The specific controller empowers ceaseless observation and management of the battery's state with the scope of extending the vehicle's driving range under varying temperature and driving pattern conditions. The proposed method utilizes an incorporated scheme for dealing with a crossover energy stockpiling framework to expand a battery's lifespan while further ensuring its smooth activity.

In this article, we proposed an integrated application of automatic moisture and Irrigation control using Real-Time Clock (RTC) and Light-Dependent Resistor (LDR). The main idea of this paper is predicated on using as little energy as possible. In this work, we use the DS1307 real-time clock module to automatically switch on or off motor dependent on the time of day. The sensors are used for monitoring the soil conditions in agriculture field. Programming controls the timing of when the device is active. The most significant benefit of the proposed design is the reduction of risk associated with potential crashes and save water. Motors on the farm are often switched on at dry time and left OFF on wet conditions till necessity. Their operation is entirely automatic. Because of the protection of the farmers from the electrical shock during the rainy season, this proposed design intends to automate the operation of soil monitoring and irrigation in order to reduce water, power consumption and advance technological progress in agriculture farms and support in smart India. A significant amount of electricity is lost in the typical lighting system; however, this may be prevented with automatic control employing LDR.