The growing burden of anxiety disorders highlights the urgent need for scalable and non-invasive systems for mental health monitoring. Electroencephalography (EEG)-based Brain-Computer Interfaces (BCIs) offer a promising solution by capturing neural oscillations linked to anxiety. However, conventional detection methods oversimplify the brain's graphstructured connectivity and are computationally intensive, limiting their feasibility for real-time, edge-based deployment. To address these limitations, we propose two edge-optimized Graph Convolutional Network frameworks (GatedGCN and GAT) for anxiety classification using EEG signals. By modeling multichannel EEG data as dynamic graphs, the system captures spatial-temporal brain dynamics critical for detecting anxietyrelated patterns. The architecture incorporates adaptive graph construction, hierarchical spatio-temporal convolutions, and quantization-aware training to enable a reduced-size inference model with minimal accuracy loss. Our approach achieves realtime, resource-efficient performance on low-power edge devices that enables continuous, private, and accessible anxiety monitoring to pave the way for practical mental health interventions in wearable and mobile healthcare settings.

Low-power wide-area networks (LPWAN) have substantially improved the Internet of Things (IoT). LoRaWAN is a potential technology for IoT applications because it uses low-power, long-distance communication and offers excellent availability with low energy consumption. LoRaWAN power consumption can be reduced using the pure Aloha protocol at the MAC level. Optimizing orthogonal transmission parameters is still a major difficulty for enhancing network performance, even though they reduce packet loss and prevent collisions, especially in dynamic and heterogeneous networks. However, the challenge of random channel selection in LoRaWAN communication often leads to inefficient resource utilization and degraded network performance. This paper proposes a novel game-theoretic approach for optimal channel selection in LoRaWAN networks. Our method leverages real-time Received Signal Strength Indicator (RSSI) data and a non-cooperative game theory model to dynamically select channels, thereby improving throughput and reducing packet loss. Through extensive simulations and a realworld testbed, we demonstrate that our proposed mechanism outperforms existing approaches such as the Online Decision algorithm and MFMSF. Specifically, it achieves up to$22 \backslash \%$improvement in throughput,$15 \backslash \%$higher packet delivery ratio, and$18 \backslash \%$reduction in latency, while consuming up to$25 \backslash \%$less energy under heavy and dynamic traffic conditions. This work offers a significant advancement in enhancing the scalability and reliability of LoRaWAN networks, paving the way for more efficient IoT communications.

The growing integration of digital technologies in agriculture is reshaping livestock farming by enabling automation, data-driven decision-making, and real-time monitoring. Among critical tasks, the timely and accurate prediction of feed and water intake is essential for ensuring animal health, early disease detection, and sustainable resource management. Traditional manual methods are labor-intensive, error-prone, and lack responsiveness to dynamic conditions, while existing smart solutions often rely heavily on cloud infrastructure that introduces latency, increasing operational costs, and limiting usability in rural or resource-constrained areas. This paper presents a lightweight, edge-compatible machine learning (ML) model for real-time prediction of livestock feed and water intake using multimodal sensor data. The proposed approach emphasizes optimizing the model size to minimize dependence on cloud infrastructure to make it suitable for deployment in connectivity-limited farm environments. The use of systematic feature selection and correlation-based modeling enhances the interpretability and accuracy of the ML models, including Random Forest, CatBoost, XGBoost, and Neural Networks. Experimental results validate the performance of the proposed solution for feed and water intake prediction that enables timely and autonomous interventions and promotes operational efficiency, sustainability, and scalability in modern farming practices.

The IEEE 802.15.4 standard is one of the widely adopted networking specifications for realizing different applications of the Internet of Things (IoT), One of its Medium Access Control (MAC) protocols, the Deterministic Synchronous Multi-channel Extension (DSME), enhances stringent QoS by allocating DSME-Guaranteed Time Slots (GTSs) between pairs of devices. However, the standard does not specify a mechanism for scheduling these DSME-GTSs, presenting numerous research opportunities in this area. In this paper, we propose a novel Dynamic Hierarchical Slot-Channel Allocation with Recursive Weighting (DyHSARW) scheme aimed at improving the scheduling of DSME-GTS in large-scale IEEE 802.15.4-based IoT networks. The proposed approach dynamically allocates non-overlapping time slots (using the HCF technique) across multiple channels (based on a device's association order), optimizing resource utilization while adapting to the hierarchical structure of the network. Specifically, the HCF condition checks if the transmission weights of a child-parent pair and the previous time slot's value are compatible, i.e., if their HCF is equal to 1. This condition indicates that the parameters are co-prime, which minimizes the likelihood of collision in the time slot allocation process. This method seeks to overcome the limitations of existing GTS scheduling algorithms, which face challenges in efficient resource allocation, increased latency, and higher energy consumption under dynamic traffic conditions.

Advancements in deep neural network accelerator architectures have resulted in the proliferation of Convolutional Neural Network (CNN) applications in computer vision. These energy-efficient accelerators provide considerable performance improvement and incur low area overhead. Due to restricted power and area constraints, this enables the accelerators to be ideally suited for several real-time mobile and edge device applications However, the major challenge is deploying the continuously growing, complex deep CNN s on embedded devices that are constrained in nature in terms of area and power. In addition, these applications generate an enormous amount of data, and accommodating them in an embedded device is also challenging. To alleviate the computational complexity of CNNs on embedded devices, this work proposed an algorithm that introduces a design trade-off between accurate computation and approximate computation. The approximate computing method reduces the computation workload and enhances the speed as well as the power efficiency of error-resilient applications. We present an approximate Vedic multiplier architecture optimized for area, power, and delay. An accelerator architecture has also been developed that computes the CNN inference model in multiple channels parallelly without any penalty on the network's accuracy or the hardware cost. It is shown in the experimental analysis that the proposed computing architecture enables the embedded devices to work with low power and area. The proposed accelerator achieves a 5.3 % improvement in computing throughput over the state-of-the-art accelerators.

As the Internet of Things (IoT) is poised to become a global phenomenon, it is imperative to schedule the transmissions of IoT devices effectively and in a fair way. Leveraging Long Range (LoRa) technology, we can achieve transmissions that consume minimal power while covering vast distances, aligning with the requirements of IoT devices. However, the proximity of multiple devices within the same area often leads to packet interference and collisions. To address this, our study introduces a pioneering scheduling method utilizing a constellation of Low Earth Orbit (LEO) satellites to manage and streamline the transmission of data from End Devices (EDs). This method employs two LEO satellites: the first satellite assigns the sequence for EDs to dispatch their packets, and the second collects these packets in the predetermined sequence before forwarding them to the LoRa Network Server (LNS). For urgent (URG) communications, EDs can alert the first satellite, which then coordinates with the LNS to schedule these priority transmissions. The LNS generates a schedule that is relayed to the second satellite, informing EDs with URG packets of their specific transmission times and channels. This scheduling approach is designed to optimize channel usage effectively while accommodating the transmission of urgent data.

LoRaWAN (Long Range Wide Area Network) is a low-power, wide-area wireless communication protocol designed specifically for the Internet of Things (IoT) and machine-to-machine (M2M) applications that enable long-range, bidirectional communication between low-power devices. Lo-RaWAN employs Adaptive Data Rate (ADR) technology to dynamically adjust the data rate for each device based on its signal quality and distance from the gateway. ADR enables improved network performance, extends device battery life, and simplifies network management, making LoRaWAN suitable for various IoT deployments. However, the end devices' inefficient utilization of radio resources (e.g., spreading factor and transmission power) significantly degrades network performance, device battery life, and adaptability to changing network conditions. Machine Learning (ML) algorithms analyze and optimize the real-time network conditions to enhance network performance. This work aims to develop an ML-based approach that adaptively selects the most suitable Spreading Factor (SF) for end devices (ED). Two independent ML algorithms such as K-means and Reinforcement Learning (RL) have been applied to EDs and Gateways, respectively, to dynamically allocate SF for both static and mobile EDs. Through simulations, the performance of the proposed mechanism is analyzed in terms of packet success rate, convergence time, energy consumption, latency, and throughput.

Enabling smart technologies in agriculture has led to the improvement of crop productivity. In India, agriculture is a primary occupation, and 70% of the population is dependent on it. Plant diseases cause significant losses in an agriculture-oriented economy. Timely monitoring of plant health and detecting plant disease is a laborious process. Automated monitoring and detection techniques hold great promise for identifying plant condition and providing useful information to facilitate effective agricultural management measures. Deep learning (DL) algorithms improve the detection accuracy in many computer vision applications of smart and precision agriculture. This paper presents a plant disease detection and classification method using YOLOv3 (You Only Look Once) model to design an Internet-of-Things (IoT) device. An approximate computing technique has been adopted that minimizes the computational complexity of DL algorithms to deploy on any embedded devices efficiently. The proposed model achieves an average of 96.92% of classification accuracy while detecting plant disease for three different classes.

The emergence of applications in vehicles requires computational capability which poses a major challenge in mo-bile edge computing. In this paper, we have proposed the token-based predictive offloading scheme with the aim of providing the optimal cost of offloading and reducing the average delay. Specifically, the proposed scheme uses a token-based approach for offloading the data from vehicles to MEC(Mobile Edge Computing) server. We have designed a separate table for MEC servers to show the status of consumed tokens and available tokens and another table for vehicles to store the status for another vehicle for vehicle to vehicle (V 2 V) communication. Dedicated Short Range Service (DSRC) technology is used to facilitate the communication between vehicle to vehicle and vehicle to infrastructure(e.g.toll gate). Extensive simulations are conducted in highway scenarios and the results demonstrate the superiority of this offloading scheme. The proposed scheme achieves low delay performance and decreased computation cost over other competing schemes in typical urban and highway scenarios.

With a plethora of Internet of Things (IoT) applications for smart cities, encompassing and supporting several enabling technologies for real-time performance, an enormous amount of network packets faces the challenge of timely delivery. The IEEE 802.15.4 standard is one of the most popular and extensively adopted networking specifications for implementing different IoT applications and catering to several application-specific Quality of Service (QoS) requirements. Deterministic Synchronous Multi-channel Extension (DSME) is one of the Medium Access Control (MAC) protocols of IEEE 802.15.4 standard that facilitates stringent QoS through the allocation of DSME-Guaranteed Time Slots (GTSs) between a pair of devices. Interestingly, the standard does not define any mechanism for scheduling the DSME-GTSs, thereby opening several research opportunities. In this paper, we propose a Distributed Delay Aware Scheduling (DDAS) mechanism to increase the efficiency of the DSME MAC by using priority-based guaranteed time slots scheduling. DDAS assigns priority to the devices according to the criticality of time and number of associated devices, i.e., it identifies various flow deadlines and assigns GTS slots accordingly. The DDAS scheme aims to satisfy and adhere to various delay deadlines in the data flows of an IoT application. The proposed scheduling mechanism is shown to outperform other closely related schemes in terms of latency as well as energy consumption.

Convolutional Neural Network (CNN) is the state-of-the-art learning technique for image understanding in several artificial vision systems. Extensive uses of memory storage and bandwidth with high computation capacity boost up the performance of a CNN. Convolution is the fundamental operation in CNN models and hence a large number of multiplier-accumulator (MAC) unit is required to compute the convolution operations adequately. MAC processing incurs extreme computational complexity by consuming significant amounts of time, energy and area. These limitations in computation have led to the exploration of different architecture of MAC to meet the demand of CNN processing. To increase the overall speed of a CNN model, an architecture called SAC (shift-accumulator) is proposed that reduces the number of overall convolution computations. Further, the SAC architecture is especially designed for convolution operations used in image sharpening, edge detection, blurring etc. The proposed SAC architecture facilitates faster convolution computations and reduces the overhead in terms of area and power than the conventional MAC architecture. Although the accuracy is slightly reduced, it does not significantly affect the efficacy of computational unit used in CNN processing.