International Journal of Reconfigurable and Embedded Systems (IJRES)
Not a member yet
454 research outputs found
Sort by
Building a photonic neural network based on multi-operand multimode interference ring resonators
Get PDFPhotonic neural networks (PNNs) offer significant potential for enhancing deep learning networks, providing high-speed processing and low energy consumption. In this paper, we present a novel PNN architecture that employs nonlinear optical neurons using multi-operand 4×4 multimode interference (MMI) multi-operand ring resonators (MORRs) to efficiently perform vector dot-product calculations. This design is integrated into a photonic convolutional neural network (PCNN) with two convolutional layers and one fully connected layer. Simulation experiments, conducted using Lumerical and Ansys tools, demonstrated that the model achieved a high test accuracy of 98.26% on the MNIST dataset, with test losses stabilizing at approximately 0.04%. The proposed model was evaluated, demonstrating high computation speed, improved accuracy, low signal loss, and scalability. These findings highlight the model’s potential for advancing deep learning applications with more efficient hardware implementations
Enhancing intrusion detection systems with hybrid HHO-WOA optimization and gradient boosting machine classifier
Get PDFIn this paper, we propose a hybrid intrusion detection system (IDS) that leverages Harris Hawks optimization (HHO) and whale optimization algorithm (WOA) for feature selection to enhance the detection of cyberattacks. The hybrid approach reduces the dimensionality of the NSL KDD dataset, allowing the IDS to operate more efficiently. The reduced feature set is then classified using logistic regression (LR) and gradient boosting machine (GBM) classifiers. Performance evaluation demonstrates that the GBM-HHO/WOA combination outperforms the LR-HHO/WOA approach, achieving an accuracy of 97.68%. These results indicate that integrating HHO and WOA significantly improves the IDS's ability to identify intrusions while maintaining high computational efficiency. This research highlights the potential of advanced optimization techniques to strengthen network security against evolving threats
Pipelined reconfigurable architecture for 5G software-defined radio systems
Get PDFThe filters are used to allow a specific band of frequencies. In a wireless communication, the filter is used to select the frequency of operation with a narrow or broad band. As the generations increase the amount of data handled increases drastically. 5G data rate can be significantly deliver up to 20 Gigabits per second while 4G communication data rate is handled in the order of 100 Megabits per second. Now the challenge becomes processing data at such a speed with low power and low area specifications. The filters that can configure themselves as per the data received are reconfigurable filters so that the bandwidth is saved. Also, when the pipelining is introduced, the reconfigurable filter improves the performance of the design. This paper details about the pipelined reconfigurable finite impulse response (RFIR) filter with the simplest algorithm with auto updating capability. The design is modelled in Verilog hardware description language (HDL) language, synthesized for Cyclone III field-programmable gate array (FPGA). The results prove that the proposed filter increases only slightly with respect to delay and power dissipation with a trade off in area and maximum possible clock frequency
A study of IoT based real-time monitoring of photovoltaic power plant
Get PDFGlobal electricity demand has increased in the last few years. This need is growing all the time as energy consumption increases using conventional energy, which will soon be phased out. So, we had to look at alternative energies, namely renewable energies. The largest and most efficient of these is solar energy, and to make the most of this energy with the greatest efficiency, the performance of these solar panels needs to be directly monitored. This study presents an independent monitoring system based on the internet of things (IoT) to measure essential factors (terminal voltage, load current, energy consumption, humidity, temperature, and light intensity). These values are realistic and accurate, based on the sensors used to measure the aforementioned factors and then using the Node MCU ESP8266 to transmit the analyzed data to the circuit. The Thingspeak platform was then employed to display, analyze, and store these results in real time
Optimizing resource allocation in job shop production systems with seasonal demand patterns
Get PDFJob shop production systems that encounter seasonal demand patterns in the manufacturing industry are the subject of this article's exploration of the complex challenges of resource allocation. A nuanced understanding of each product's unique production processes, resource requirements, and lead times is necessary for the inherent complexity of job shop production, which characterized by diverse product lines. Resource reallocation becomes more complicated due to seasonal demand patterns, which require manufacturers to seamlessly transition resources between products and adjust strategies dynamically throughout the year. This article explores potential optimization techniques by drawing on insights from related studies on reliability monitoring and Petri nets. Strategically managing resource allocation is highlighted due to its significant impact on a company's competitiveness, adaptability to market changes, and overall financial performance. In the paper, there is a proposed architecture for resource allocation that combines data-driven insights, workforce planning, inventory management, machine allocation, lean principles, and technology integration. Effective strategies for reallocating resources are highlighted through the presentation of case studies and best practices, which include accurate demand forecasting and flexible workforce planning. The final section of the article emphasizes the holistic approach required to navigate the complexities of seasonal demand patterns and achieve sustained competitiveness and customer satisfaction
Performance comparison of indoor navigation and obstacle avoidance methods for low-cost implementation in wheelchairs
Get PDFWheelchairs are a huge support for the movement of people who have disabilities. The wheelchairs that were traditionally moved using manual effort have given way to powered and smart wheelchairs with various controlling methods. When powered wheelchairs are used indoors, navigation and avoiding obstacles become challenging and tricky for a disabled user. To address these challenges there have been implementations of expensive and high-end systems to make the wheelchair move autonomously but as a result such a wheelchair is not economically viable for many users. Thus, there is a need for an alternative low cost method for users to be able to navigate and move in an indoor environment. The paper reviews low-cost methods for implementing indoor navigation systems, weighing their performances to validate if these methods can be used as a viable alternative to the high-cost systems for autonomous navigation in an indoor environment
Development of internet of vehicles and recurrent neural network enabled intelligent transportation system for smart cities
Get PDFThe number of deaths has increased as a direct result of the increased frequency of traffic accidents, congestion, and other risk factors. Developing countries have prioritised the development of intelligent transport systems in order to reduce pollution, traffic congestion, and wasted time. This article describes an intelligent transport system that leverages the internet of vehicles (IoV) and deep learning to forecast traffic congestion. Data is acquired using a car’s global positioning system (GPS), road and vehicle sensors, traffic cameras, and traffic speed, density, and flow. All acquired data is stored in one location on a cloud server. The cloud server also stores historical traffic, road, and vehicle data. Using particle swarm optimisation, features are improved. The optimised dataset is used to train and test recurrent neural networks (RNNs), support vector machines (SVMs), and multi layer perceptrons (MLPs). A deep learning algorithm can predict traffic congestion and make recommendations to drivers on how fast to travel and which route to take. The experimental effort employs the performance measurement system (PeMS) traffic dataset. RNN has achieved accuracy of 95.1%
A fast half-subtractor using 8T static random access memory for in-memory computation
Get PDFThe existing system for computation completely incorporates Von-Neumann architecture which has limitations with respect to its memory, parallelism and power constraints. This has affected the efficiency of the computing system. Novel architectural solutions are required to meet the growing demands for improved computational efficiency and power management in very large scale integration (VLSI) systems. To deal with the large-scale data, computation in memory (CIM) has been introduced. The paper presents the half subtractor circuit and the In-memory computation co-design using eight transistors static random access memory (SRAM) cell whose read circuitry is transmission gate based. The proposed half-subtractor with the CIM is implementation is carried out in 180 nm complementary metal– oxide–semiconductor (CMOS) technology. The sensing scheme used is the latch-based sense amplifier along with the 8T SRAM cell. The proposed SRAM with transmission-gate based read circuitry along with latch-based sense amplifier reduces the delay and power consumed during the read operation significantly and a bit reduction during the write operation. The static noise margin (SNM) for read operation has been increased by 9% in the transmission gate-based SRAM as compared to conventional 8T SRAM. The delay of the proposed design has been reduced by 53% during the read operation and 4.43% during the write operation. The power consumed has been reduced by 3% and 8.6% during read and write operations, respectively
Development and evaluation of robotic exoskeleton arm for enhanced human load carrying efficiency
Get PDFIn recent years, there has been a significant amount of research dedicated to the development of robotic exoskeleton systems. These technologies have been widely explored for their potential in virtual reality, human power enhancement, robotic rehabilitation, human power assist, and haptic interface applications. This research focuses on creating an exoskeleton arm that can assist individuals in carrying heavy objects. The exoskeleton arm is initially designed using Fusion 360, with the identification and calculation of important components such as the exoskeleton structure, motors serving as joints, an electromyography (EMG) sensor, and an Arduino UNO microcontroller. The research involves various aspects of mechanical design, electronic components, and programming. The effectiveness of the developed exoskeleton arm is then tested through experiments involving several individuals lifting a 2.5 kg and 5.0 kg load. The results of the experiments demonstrate that the force generated by the muscles is reduced when using the exoskeleton arm, compared to using a supporting system. Individuals' performance dropped by 36.06% to 50.44% when using an exoskeleton to lift 2.5 kg. This emphasises its effect on muscle activation and efficiency following physical activity. A 10.14% to 23.25% decline in a 5.0 kg lift shows nuanced impacts, emphasising the need for personalised modifications
Chirp-pulsed eddy current testing for crack detection in low-carbon steel
Get PDFThis paper introduces a signal processing feature for chirp-pulsed eddy current testing (C-PECT) to improve crack detection in low-carbon steel, a common material in maritime structures. While C-PECT is an established technique, inspecting ferromagnetic materials is challenging due to significant background noise from lift-off variations and material permeability. The novelty of this work lies in the proposal of a frequency-domain integration feature designed to suppress this noise. The method utilizes a chirp-pulse-excited probe with a Hall sensor to measure the magnetic field response. By integrating the signal's magnitude spectrum, the frequency feature effectively flattens the background and enhances the signal-to-noise ratio. Experimental validation on a low-carbon steel specimen with artificial cracks demonstrates the feature's superior performance in providing clear, high-contrast crack indications compared to a conventional time-domain analysis. The results indicate that this approach offers a simple, computationally efficient, and robust solution for the qualitative detection and localization of cracks, enhancing structural integrity assessments in noisy industrial environments