International Journal of Embedded Systems (15 papers in press)

Regular Issues

• A survey on latency and power consumption estimation for embedded systems  
  by Nejra Beganovic, Mattias O'Nils 
  Abstract: Performance evaluation of Internet of Things (IoT) platforms becomes inevitable as the number of IoT devices is constantly increasing. Discussing from the aspect of their interdependences, it is of utmost importance to provide an efficient framework for the analysis of causal relation between consumed power, processing latency, data size reduction, and algorithm computational complexity of embedded systems. As embedded devices, operating often on limited and unreliable energy sources such as batteries or other energy harvesters, are the devices with the highest need for optimal power use, the main focus of this contribution is to review energy consumption modelling approaches and their relation to a latency modelling framework. Such analysis is necessary to provide the basis for efficient system design from early design stage and to guarantee the fulfillment of all system requirements. Accordingly, the paper points out not only existing challenges but also the possibilities for improvements with respect to power/energy savings.
  Keywords: power estimation; energy consumption; internet of things; embedded systems;.
  
  
• A physically unclonable function architecture with multiple responses on FPGA  
  by M.A. Muneeb, S. Nalesh, S. Kala 
  Abstract: Physically Unclonable Functions (PUFs) are hardware security devices that are commonly used for authentication purposes. The unique aspect of PUFs lies in the fact that they are based on the nano-scale differences in the material used, leading to different responses each time. This randomness and uniqueness enhances the security of PUFs, making them a popular choice. There are two main types of PUFs: Arbiter PUF and Ring Oscillator PUF. Recent advancements have revealed that incorporating additional hardware in Arbiter PUFs can further improve security. However, the challenge lies in protecting the PUF when it generates a single response. To overcome this, we propose an 8-bit response PUF architecture, which eliminates this vulnerability and improves the security of PUFs. The proposed architecture has been implemented on Xilinx Spartan-7 FPGA and the results have been evaluated.
  Keywords: hardware security; FPGA; PUF; multiple response.
  
  
• A hybrid meta-heuristic algorithm with fuzzy clustering method for IoT smart electronic applications  
  by Liejiang Huanh, Sichao Chen, Dilong Shen, Yuanjun Pan, Jixing Yang, Yuanchao Hu 
  Abstract: Nowadays, decision support systems and recommendation systems are emerging methodologies in the internet of things (IoT) environments for optimising smart electronic services such as smart tourist, smart logistic, smart transportation and smart home services. This paper proposes a smart recommendation system in the smart tourism application. Then, a content-based filtering method is proposed for improving search-based attributes in clustering. In addition, fuzzy C-means clustering is used to cluster the existing data in terms of the users’ requests and recommend the optimised choice. Also, bat optimisation algorithm (BOA), which is a well-known meta-heuristic algorithm is provided to improve the accuracy of the clustering to 98% better than other state-of-the-art case studies. In addition, precision and recall are evaluated for predicting decision-making aspects in smart tourist applications. The results achieved are compared with that of the similar research studies and is superiority is shown.
  Keywords: smart recommendation system; fuzzy logic; intelligent recommendation; internet of things; IoT; data mining; bat optimisation algorithm; BOA.
  
  
• Edge computing-based internet of medical things for health care using deep learning  
  by Himabindu Sathyaveti, C. Gomathy 
  Abstract: Edge cloud computing (ECC) servers for the Internet of Medical Things (IoMT) have made medical systems smarter by providing patients with access to cutting-edge technologies that enable choosing medical facilities. Modern medical solutions have gained popularity because of their economic and ethical benefits. To improve the quality of service (QoS), more bandwidth (BW) and edge computing (EC) are essential. Connecting the IoMT to high-processing devices like these healthcare service devices often demand low network latencies. This study reviews the most important, currently existing state-of-the-art methods, including Edge-based rehabilitation systems (ERS), IoT-based infectious disease management (IoT-IDM), and ML usage in healthcare IoT applications. Finally, the performances of the IoMT-influenced edge computing-based healthcare applications are evaluated for quality-of-service metrics such as latency, energy consumption, bandwidth utilization rate, computational cost, memory consumption, and offloaded efficiency.
  Keywords: edge computing; internet of medical things; healthcare services; deep learning.
  
  
• A new reinforcement learning approach for improving energy trading management for smart microgrids in the internet of things  
  by Qiuyu LU, Haibo LI, Jianping ZHENG, Jianru QIN, Yinguo YANG, Li LI, Keteng JIANG 
  Abstract: To face the challenge of climate change, it is necessary to change the usage of natural energy resources like oil, natural gas and coal to the use of renewable resources like the wind and sun energy. Providing a tool for efficient network management with using various distributed generation and storage services is a critical issue for smart microgrids. Owing to the nature of this type of power grid, reinforcement learning is an online-wide framework for solving scattering problems. Therefore, while considering economic dispatch as a reference method of distributed power grid, its management is modelled locally. In this paper, a distributed algorithm is defined for generalised consumption balancing. The simulation shows that this algorithm has the potential to maintain the stability of the safe and efficient operation of the entire care network, taking into account the local stability. Also, the cost of energy usage is reduced in the transmission and distribution system.
  Keywords: reinforcement learning approach; energy trading management; smart microgrids; internet of things; generalised consumption balancing; distributed algorithm.
  DOI: 10.1504/IJES.2023.10056514
   
  
• Automatic detection of contextual defects based on machine learning  
  by Cangming Liang, J. Liu, Jintao Feng, Anhong Xiao, Hui Zeng, Qujin Wu, Tonglan Yu 
  Abstract: In recent years, automatic detection technology has been widely used in software defect detection, which significantly reduces the cost of manual inspection. Machine learning is one of the common automatic detection technologies. By extracting defect features and using supervised classification algorithms to automatically identify possible software doubtful defects, the key is to design an automatic verification model with high accuracy. This paper focuses on the contextual defects related to control flow, and uses the dynamic checking method in the static testing process. By using control flow to represent the context information of code, the node features and basic path features in control flow graph are proposed, and a new defect automatic verification model based on SVM is designed. The experimental results show that the proposed model has higher recall and F1 than the existing methods, and significantly improves the efficiency of automatic identification of suspected defects.
  Keywords: automatic defect detection; machine learning; support vector machine; SVM; contextual d.
  DOI: 10.1504/IJES.2023.10056887
   
  
• AR2PNet: an adversarially robust re-weighting prototypical network for few-shot learning  
  by Sirui Li, Li Guo, Xianmin Wang, Songcao Hou, Zhicong Qiu, Yutong Xie, Haiyan Liang 
  Abstract: Robust re-weighting prototypical networks (RRPNet) model is a promising method to improve the robustness of prototypical networks (ProtoNet). However, the performance of RRPNet is limited when the examples are scare and the noise is trivial. In this paper we propose a novel re-weighting prototypical networks framework for few-shot learning based on AT, called AR2PNet, to enhance the performance of RRPNet. Specifically, instead of directly calculating the similarity between the naive representations of the examples, we calculate such similarity between prototype representations, which is conductive to reducing the computation cost as well as enhancing the model prediction accuracy. Meanwhile, to encourage the model to resist adversarial examples, we formulate the loss function as a minimax problem inspired by the conception of AT. We conduct experiments on CIFAR-FS and MiniImageNet dataset, and the experimental results demonstrate the effectiveness of the propose method.
  Keywords: deep learning; few-shot learning; prototypical network; adversarial training; AT.
  DOI: 10.1504/IJES.2023.10057290
   
  
• Real life implementation of an energy-efficient adaptive advance encryption design on FPGA  
  by Neeraj Bisht, Bishwajeet Pandey, Sandeep Budhani 
  Abstract: Advanced encryption standards (AES) is a mainstream algorithm regularly employed by numerous applications for encryption and decryption purposes. A significant disadvantage of the AES algorithm is its high-power consumption. In this research, experimental results are used to compare the on-chip energy consumption and junction power needs of AES algorithms. Five unique FPGAs and four distinct frequencies are used in these tests. Based on the findings, it was found that all FPGAs performed optimally at a frequency of 1.6 GHz. Compared to the worst performing FPGA Artix-7, Kintex-7 Low Voltage used 21.34% less on-chip power during encryption and 20.5% less during decryption. This work validates the considerable improvement in power efficiency by comparing the proposed architecture’s on-chip energy consumption figures to those of other existing models. It is suggested to use a 1.60 GHz Kintex-7 Low Voltage processor to run the AES encryption and decryption algorithms.
  Keywords: green computing; advanced encryption standards; field programmable gate array; on-chip energy usage; junction temperature.
  DOI: 10.1504/IJES.2023.10057602
   
  
• Improving real-time performance of Jailhouse on embedded systems via bank and cache partitioning  
  by Hubin Yang, Liu Yang, Fengyun Li, Yucong Chen, Rui Zhou, Qingguo Zhou 
  Abstract: As multi-core architectures become increasingly prevalent in embedded systems, the use of a lightweight hypervisor to integrate multiple tasks on a single hardware platform has become increasingly popular. Although this approach can significantly reduce costs while improving system performance, it can cause hardware shared resource contention issues. We propose a virtualisation-based mechanism that mitigates shared memory and cache competition among virtual machines through the implementation of DRAM bank and cache partitioning techniques. The results demonstrate that applying our approach led to an average and maximum latency reduction of 74.49% and 73.59%, respectively, when compared with Jailhouse without partitioning using Cyclictest. Therefore, it is evident that our proposed mechanism effectively improves Jailhouse's real-time performance on multi-core embedded systems.
  Keywords: DRAM bank partitioning; cache partitioning; real-time; Jailhouse hypervisor; isolation.
  DOI: 10.1504/IJES.2023.10058436
   
  
• Research on malicious traffic detection based on image recognition  
  by Wei Li, Yuliang Chen, Lixin Zhao, Yazhou Luo, Xin Liu 
  Abstract: With the rapid development of the internet, information security problems caused by malicious traffic are becoming more and more serious. Malicious traffic invades the target system, interferes with the regular operation of the target internet device, steals user privacy, and destroys network availability. Therefore, this paper proposes a malicious traffic detection method based on image recognition technology, which is used to detect network traffic data, mine malicious traffic, provide early warning for users, and avoid network security threats. Based on the feature extraction of the text information of the network traffic data, the method converts the string data of the network traffic into picture data containing feature information, and combines the convolutional neural network (CNN) to realise the analysis of the attack vector detection on network traffic. Experimental results show that, compared with traditional machine learning methods, this method has a more efficient and accurate identification ability for malicious traffic attacks.
  Keywords: web attack; malicious traffic detection; image recognition; convolutional neural network; CNN.
  DOI: 10.1504/IJES.2023.10058970
   
  

Special Issue on: ICSAI 2021 Artificial Intelligence Techniques and Applications in Distributed Real-Time Embedded Systems

• Imbalanced COVID-19 dataset classification with bidirectional sampling based on sample correlation  
  by Mansheng Xiao, Mingkai Fan, Guocai Zuo 
  Abstract: Aiming at the problem that the classification hyperplane is inclined toward the positive class when the CNN model directly classifies the imbalanced dataset, resulting in a high misclassification rate, a bidirectional sampling method based on sample correlation is proposed. Firstly, the sampling ratio is designed according to the number of the two types of sample, and then, considering the influence of the positional correlation between the samples, the methods of under-sampling negative samples and oversampling of positive samples are proposed. Therefore, the balance of the number of positive and negative samples is achieved. Finally, after sampling the imbalanced dataset of Kaggle images, the deep learning model SSD is used to train and identify the COVID-19 samples. The experimental comparison results show that the method proposed in this paper can improve the evaluation indices such as F+-measure and G-means by more than 5% in the identification of COVID-19.
  Keywords: bidirectional sampling; sample correlation; FCM; SSD; COVID-19.
  
  
• Reliability enhancement algorithm based on budget level in cloud-edge environments  
  by Longxin Zhang, Dantong Liu, Minghui Ai, Runti Tan, Zhihao Zeng 
  Abstract: Low energy consumption and high reliability are two core performance metrics for task scheduling on heterogeneous cloud systems. Dynamic voltage and frequency scaling is an efficient technique that reduces energy consumption by dynamically scaling the processors supply voltage/frequency. However, frequent changes in processor frequency can lead to a dramatic increase in transient faults, which can affect system reliability. Hence, a novel budget level (BL) energy pre-allocation strategy is designed and a reliability maximisation algorithm based on BL (RMBL) is proposed in this study. The RMBL algorithm includes three stages, namely, the establishment of a task priority queue, the pre-allocation of task energy consumption constraints, and the determination of the optimal virtual machine and frequency combination. Based on two real-world applications, namely, Laplace and Gaussian elimination, experimental results show that RMBL can achieve better reliability while satisfying energy consumption constraints.
  Keywords: cloud-edge environment; dynamic voltage and frequency scaling; energy consumption; reliability.
  
  
• A rock classification system based on an embedded platform  
  by Xiangyuan Zhu, Jie Yang, Weiyang Zhi, Haifeng Lu 
  Abstract: Rock is a major material for the crust and upper mantle formation of the Earth. In Earth Sciences, rock image classification is an essential and critical task in the geological survey. Owing to the scarcity of samples and unaffordability of rock classification systems, an embedded system was built to collect and identify rock images. The Raspberry Pi3B+ was applied as the micro controller unit and the Sony IMX219 image sensor was selected to shoot rock images. The new well-annotated dataset contains seven types of fresh rocks with 7,976 images. Based on the new dataset, a new rock classification model based on the ConvNeXt algorithm was proposed. To ensemble the local and global features of the rock images, a feature fusion strategy named Super-Image was designed. Compared with the prevalent models including VGG16, ResNet50, MobileNet_V3, GoogleNet, and DenseNet121, our enhanced ConvNeXt method achieved the macro-average F1 and accuracy of 99.61% and 99.63%, respectively.
  Keywords: rock image classification; embedded platform; ConvNeXt; super-image; dataset enhancement.
  
  
• Embedded elbow vein blood collection robot system based on artificial intelligence technology  
  by Huanwen Wang, Xiaoya Zhang, Hao Chen 
  Abstract: With the development and advancement of health care, venous blood collection is becoming an essential tool for the initial screening of patients for disease. In current clinical practice, it is mainly done manually by experienced physicians or nurses. To effectively reduce the repetitive work of health care workers, we designed an embedded elbow vein recognition system Artificial Intelligence technology (AI)-based. Firstly, a prototype of vein vascular identification based on Near Infrared (NIR) spectroscopy was completed by combining the analysis of vein vascular acquisition data with NIR spectroscopy. Secondly, a deep learning-based image classification algorithm is implemented to classify elbow cover images. Furthermore, the semantic segmentation algorithm CNN-based achieves intelligent extraction of the region of interest for the elbow vein vessels. Finally, we validate the system in a real environment, and the results show that the system achieves high recognition accuracy for cubital veins through effective training and testing strategies.
  Keywords: convolutional neural network; elbow vein vessel identification; medical robot; intelligent embedded systems; venipuncture.
  
  

Special Issue on: 5G Embedded and Cyber-physical Data for Next-Generation Intelligent Clusters of Distributed Networks

• FCGE: fuzzy clustering-based genetic evolutionary algorithm for 5G wireless sensor re-localization  
  by Vishal Patil, Ninad More 
  Abstract: Traffic analysis is required for successful packet transmission from source to destination after receiving network acknowledgement. Sensor nodes must send position information to nearby nodes, reducing accuracy and performance. However, precise node location determination improves network performance. This paper proposes a Fuzzy Clustering-based Genetic Evolutionary (FCGE) algorithm for effective re-localisation in heavy traffic. It chooses the best node for re-localisation, reducing network traffic, communication costs, and localization inaccuracy. A proper intelligent re-localisation method is used within a predetermined time frame to avoid localisation errors. To determine the optimal retrigger time, the genetic algorithm uses Time Bound Re-localization (TBR). It is more accurate even with many node migrations, localisation errors are reduced. Fuzzy k-means improves network data collection by creating disjoint clusters and global time synchronisation for mobile agents. The proposed FCGE algorithm reduced RMSPE by 13.26%, 9.65%, and 6.78% for 5, 10, and 20 anchor nodes, respectively.
  Keywords: Monte-Carlo localisation; time-bound re-localisation; mobile agent; wireless sensor network; intelligent re-localisation.