检索结果-内蒙古大学图书馆

AnchorCapsule: A Datastream-Serving Post-Processor for Object Detection in embedded Vision SoC

IEEE transactions ON CIRCUITS AND systems II-EXPRESS BRIEFS 2024年第2期71卷 842-846页

作者： Peng, Yi Yang, Yongkui Wang, Zheng Chen, Chao Chinese Acad Sci Shenzhen Inst Adv Technol Shenzhen 518055 Peoples R China Univ Chinese Acad Sci Beijing 101408 Peoples R China

We present AnchorCapsule, a post -processor for object detection to relieve the workload of CPUs with minimal cost in latency. The capsule takes advantage of a 'datastreamserving' architecture paradigm, where computing logic is built to cater for the throughput of DRAM datastreams and simultaneously minimize the usage of on -chip buffers. The implementation results show that AnchorCapsule has a tiny area of 0.038mm(2) but achieves 1 order of magnitude faster than Intel Xeon and 2 orders of magnitude faster than ARM A53, resulting in an endto -end system -level latency reduction of 46.7% and 16.7% for Yolo-v3-tiny and Yolo-v3 networks, respectively. The precision of AnchorCapsule has been validated on 1,500 samples from three prestige datasets, giving a promising result of 98%+ accuracy in the bounding box (BBox) coordinates, 99%+ in BBox sizes and 100% in object types. Compared with state-of-the-art, AnchorCapsule can filter 9,408 candidate BBoxes in a single run, which is 3.5x in the processing capacity of best-known published work.

关键词： Object detection post-processing datastream-serving architecture

来源：评论

学校读者我要写书评

暂无评论

Pragmatic Memory-System Support for Intermittent computing Using Emerging Nonvolatile Memory

引用

IEEE transactions ON COMPUTER-AIDED DESIGN OF INTEGRATED CIRCUITS AND systems 2023年第1期42卷 95-108页

作者： Sliper, Sivert T. Wang, William Nikoleris, Nikos Weddell, Alex S. Savanth, Anand Prabhat, Pranay Merrett, Geoff V. Univ Southampton Sch Elect & Comp Sci Southampton SO17 1BJ England Arm Ltd Arm Res Cambridge CB1 9NJ England

Intermittent computing (IC) is a key enabler for the vision of a trillion Internet of Things devices. By harvesting energy from the environment and leveraging nonvolatile memory (NVM) to retain computational progress across power cycles, IC enables untethered and battery-free devices to perform computation whenever ambient energy is available. The backbone of state retention is NVM, and recent advances in energy-efficient NVM have the potential to expand the application domain of IC significantly. Utilizing emerging NVM at the level of bit cells, researchers have proposed nonvolatile processors. However, these do not leverage hardware-software co-design, which can be used to overcome hardware limitations and to provide support for application-level constraints such as atomicity. In this article, we propose MEMIC, a memory architecture tailored for IC devices with byte-addressable NVM. A core focus of MEMIC is to combine volatile and NVM in such a way that the operations of IC are as efficient as possible, while also maximizing computational performance per joule. MEMIC uses volatile memory for energy efficiency and NVM for data retention. To avoid double-buffered checkpoints and costly roll backs when code needs to be reexecuted, MEMIC is designed to track and minimize writes to NVM during failure-atomic sections. Our evaluation shows that MEMIC's instruction cache reduces workload completion time under intermittent operation by 41%-70% and its data cache provides a further reduction of 13%-39%.

关键词： embedded systems hardware-software codesign intermittent computing (IC) low-power design

来源：评论

学校读者我要写书评

暂无评论

UAV-Mounted Cloudlet systems for Emergency Response in Industrial Areas

引用

IEEE transactions ON INDUSTRIAL INFORMATICS 2022年第11期18卷 8007-8016页

作者： Coutinho, Rodolfo W. L. Boukerche, Azzedine Concordia Univ Elect & Comp Engn Dept Montreal PQ H3G 1M8 Canada Univ Ottawa EECS Ottawa ON K1N 6N5 Canada

The advancements in Internet of Things (IoT) and embedded systems, 5G networks and next-generation wireless systems, and embedded and distributed artificial intelligence are empowering Industry 4.0. In this new industrial era, smart autonomous and connected systems will improve efficiency, reduce cost and pollution, and increase productivity and safety in industrial 4.0 based applications. Nevertheless, the new generation technologies mentioned above have not been explored to improve safety in multiplant and industrial zones. In this article, we shed light on the design of unmanned aerial vehicles (UAVs)-assisted systems for emergence response in multiplant and industrial zones. We discuss the important contributions of UAV-mounted cloudlet systems to support SAR missions by providing computation, communication, and storage resources to first responders, surveillance of the affected area, real-time monitoring of SAR members and victims, and caching nodes to improve content delivery among SAR teams. Besides, we design an envisioned UAV-cloudlet reference system for emergency response in industrial areas. Moreover, we highlight the current challenges that are being addressed in the literature aimed at making possible the efficient use of the designed UAV-mounted cloudlet systems for emergency response. Finally, we point out some future research directions that require further investigation.

关键词： Cloud computing Fourth Industrial Revolution Task analysis Servers Real-time systems Emergency services Inspection Drones emergency services industry applications multi-access edge computing

来源：评论

学校读者我要写书评

暂无评论

PArtNNer: Platform-Agnostic Adaptive Edge-Cloud DNN Partitioning for Minimizing End-to-End Latency

引用

ACM transactions on embedded computing systems 2024年第1期23卷 1-38页

作者： Ghosh, Soumendu Kumar Raha, Arnab Raghunathan, Vijay Raghunathan, Anand Purdue Univ Elmore Family Sch Elect & Comp Engn 610 Purdue Mall W Lafayette IN 47907 USA Intel Corp Adv Architecture Res Team NPU IP CGAI CCG Mission Coll Blvd Santa Clara CA 95054 USA

The last decade has seen the emergence of Deep Neural Networks (DNNs) as the de facto algorithm for various computer vision applications. In intelligent edge devices, sensor data streams acquired by the device are processed by a DNN application running on either the edge device itself or in the cloud. However, "edge-only" and "cloud-only" execution of State-of-the-Art DNNs may not meet an application's latency requirements due to the limited compute, memory, and energy resources in edge devices, dynamically varying bandwidth of edge-cloud connectivity networks, and temporal variations in the computational load of cloud servers. This work investigates distributed (partitioned) inference across edge devices (mobile/end device) and cloud servers to minimize end-to-end DNN inference latency. We study the impact of temporally varying operating conditions and the underlying compute and communication architecture on the decision of whether to run the inference solely on the edge, entirely in the cloud, or by partitioning the DNN model execution among the two. Leveraging the insights gained from this study and the wide variation in the capabilities of various edge platforms that run DNN inference, we propose PArtNNer, a platform-agnostic adaptive DNN partitioning algorithm that finds the optimal partitioning point in DNNs to minimize inference latency. PArtNNer can adapt to dynamic variations in communication bandwidth and cloud server load without requiring pre-characterization of underlying platforms. Experimental results for six image classification and object detection DNNs on a set of five commercial off-the-shelf compute platforms and three communication standards indicate that PArtNNer results in 10.2x and 3.2x (on average) and up to 21.1x and 6.7x improvements in end-to-end inference latency compared to execution of the DNN entirely on the edge device or entirely on a cloud server, respectively. Compared to pre-characterization-based partitioning approaches, PArtNNe

关键词： Deep learning edge inference collaborative AI real-time computing

来源：评论

学校读者我要写书评

暂无评论

DELICIOUS: Deadline-Aware Approximate computing in Cache-Conscious Multicore

引用

IEEE transactions ON PARALLEL AND DISTRIBUTED systems 2023年第2期34卷 718-733页

作者： Saha, Sangeet Chakraborty, Shounak Agarwal, Sukarn Gangopadhyay, Rahul Sjalander, Magnus McDonald-Maier, Klaus Univ Huddersfield Dept Comp Sci Huddersfield HD1 3DH W Yorkshire England Univ Essex Embedded & Intelligent Syst Lab Colchester CO4 3SQ Essex England Norwegian Univ Sci & Technol Dept Comp Sci N-7491 Trondheim Norway Univ Edinburgh Sch Informat Edinburgh EH8 9YL Midlothian Scotland Moscow Inst Phys & Technol Dolgoprudnyi 141700 Russia Univ Essex Sch Comp Sci & Elect Engn Colchester CO4 3SQ Essex England

Enhancing result-accuracy in approximate computing (AC) based real-time systems, without violating power constraints of the underlying hardware, is a challenging problem. Execution of such AC real-time applications can be split into two parts: (i) the mandatory part, execution of which provides a result of acceptable quality, followed by (ii) the optional part, that can be executed partially or fully to refine the initially obtained result in order to increase the result-accuracy, without violating the time-constraint. This article introduces DELICIOUS, a novel hybrid offline-online scheduling strategy for AC real-time dependent tasks. By employing an efficient heuristic algorithm, DELICIOUS first generates a schedule for a task-set with an objective to maximize the results-accuracy, while respecting system-wide constraints. During execution, DELICIOUS then introduces a prudential cache resizing that reduces temperature of the adjacent cores, by generating thermal buffers at the turned off cache ways. DELICIOUS further trades off this thermal benefits by enhancing the processing speed of the cores for a stipulated duration, called V/F Spiking, without violating the power budget of the core, to shorten the execution length of the tasks. This reduced runtime is exploited either to enhance result-accuracy by dynamically adjusting the optional part, or to reduce temperature by enabling sleep mode at the cores. While surpassing the prior art, DELICIOUS offers 80% result-accuracy with its scheduling strategy, which is further enhanced by 8.3% in online, while reducing runtime peak temperature by 5.8? on average, as shown by benchmark based evaluation on a 4-core based multicore.

关键词： Real-time systems approximate computing thermal management dead block caches resizing TDP

来源：评论

学校读者我要写书评

暂无评论

Sustainable Deep Learning at Grid Edge for Real-Time High Impedance Fault Detection

IEEE TRANSACTIONS ON SUSTAINABLE COMPUTING

引用

IEEE transactions ON SUSTAINABLE computing 2022年第2期7卷 346-357页

作者： Sirojan, Tharmakulasingam Lu, Shibo Phung, B. T. Zhang, Daming Ambikairajah, Eliathamby Univ New South Wales Sch Elect Engn & Telecommun Sydney NSW 2052 Australia

High impedance faults (HIFs) on overhead power lines are known to cause fires. They are difficult to detect using conventional protection relays because the fault current is insufficient to cause tripping. The delay in detecting HIFs can result in severe bushfires and energy losses;hence a high throughput, low latency detection scheme needs to be developed for HIF detection. Moreover, the complexities associated with HIF detection demands signal processing techniques combined with artificial intelligence to achieve higher detection accuracy. This paper proposes a sustainable deep learning-based approach in an edge device, that can be mounted on top of a power pole to detect HIFs in real-time. Data acquisition, feature extraction, and deep learning based fault identification are performed in an embedded edge node to achieve higher throughput, reduced latency as well as offload the network traffic. Furthermore, optimization techniques such as hardware parallelism and pipelining are adapted to achieve real-time fault identification on edge devices while ensuring the efficient usage of its limited resources. Real-time implementation of the proposed system is validated through laboratory experiments and the results demonstrate the suitability of edge computing to detect HIFs in terms of reduced detection latency (115.2 ms) and higher detection accuracy (98.67 percent).

关键词： Real-time systems Image edge detection Edge computing Fires Cloud computing Fault detection Artificial intelligence Deep learning edge computing embedded artificial intelligence real-time fault detection sensor data analytics signal processing sustainable computing

来源：评论

学校读者我要写书评

暂无评论

Compute-in-Memory-Based Neural Network Accelerators for Safety-Critical systems: Worst-Case Scenarios and Protections

引用

IEEE transactions ON COMPUTER-AIDED DESIGN OF INTEGRATED CIRCUITS AND systems 2024年第8期43卷 2452-2464页

作者： Yan, Zheyu Hu, Xiaobo Sharon Shi, Yiyu Univ Notre Dame Dept Comp Sci & Engn Notre Dame IN 46556 USA

Emerging nonvolatile memory (NVM)-based computing-in-memory (CiM) architectures show substantial promise in accelerating deep neural networks (DNNs) due to their exceptional energy efficiency. However, NVM devices are prone to device variations. Consequently, the actual DNN weights mapped to NVM devices can differ considerably from their targeted values, inducing significant performance degradation. Many existing solutions aim to optimize average performance amidst device variations, which is a suitable strategy for general-purpose conditions. However, the worst-case performance that is crucial for safety-critical applications is largely overlooked in current research. In this study, we define the problem of pinpointing the worst-case performance of CiM DNN accelerators affected by device variations. Additionally, we introduce a strategy to identify a specific pattern of the device value deviations in the complex, high-dimensional value deviation space, responsible for this worst-case outcome. Our findings reveal that even subtle device variations can precipitate a dramatic decline in DNN accuracy, posing risks for CiM-based platforms in supporting safety-critical applications. Notably, we observe that prevailing techniques to bolster average DNN performance in CiM accelerators fall short in enhancing worst-case scenarios. In light of this issue, we propose a novel worst-case-aware training technique named A-TRICE that efficiently combines adversarial training and noise-injection training with right-censored Gaussian noise to improve the DNN accuracy in the worst-case scenarios. Our experimental results demonstrate that A-TRICE improves the worst-case accuracy under device variations by up to 33%.

关键词： Performance evaluation Artificial neural networks Training Robustness Perturbation methods Nonvolatile memory Design automation embedded systems hardware/software co-design memory noise analysis

来源：评论

学校读者我要写书评

暂无评论

A Design Flow Based on Docker and Kubernetes for ROS-based Robotic Software Applications

引用

ACM transactions on embedded computing systems 2024年第5期23卷 1-24页

作者： Lumpp, Francesco Panato, Marco Bombieri, Nicola Fummi, Franco Univ Verona Dept Comp Sci Verona Italy

Human-centered robotic applications are becoming pervasive in the context of robotics and smart manufacturing, and such a pervasiveness is even more expected with the shift to Industry 5.0. The always increasing level of autonomy of modern robotic platforms requires the integration of software applications from different domains to implement artificial intelligence, cognition, and human-robot/robot-robot interaction. Developing and (re)configuring such a multi-domain software to meet functional constraints is a challenging task. Even more challenging is customizing the software to satisfy non-functional requirements such as real-time, reliability, and energy efficiency. In this context, the concept of Edge-Cloud continuum is gaining consensus as a solution to address functional and non-functional constraints in a seamless way. Containerization and orchestration are becoming a standard practice, as they allow for better information flow among different network levels as well as increased modularity in the use of multi-domain software components. Nevertheless, the adoption of such a practice along the design flow, from simulation to the deployment of complex robotic applications by addressing the de facto development standards (e.g., ROS- Robotic Operating System) is still an open problem. We present a design methodology based on Docker and Kubernetes that enables containerization and orchestration of ROS-based robotic SW applications for heterogeneous and hierarchical HW architectures. The methodology aims at (i) integrating and verifying multi-domain components since early in the design flow, (ii) mapping software tasks to containers to minimize the performance and memory footprint overhead, (iii) clustering containers to efficiently distribute load across the edge-cloud architecture by minimizing resource utilization, and (iv) enabling multi-domain verification of functional and non-functional constraints before deployment. The article presents the results obtained

关键词： ROS Docker Kubernetes K3S robotic applications Edge-Cloud computing

来源：评论

学校读者我要写书评

暂无评论

Energy-efficient Personalized Federated Search with Graph for Edge computing

引用

ACM transactions on embedded computing systems 2023年第5期22卷 1-24页

作者： Yang, Zhao Sun, Qingshuang Changan Univ Coll Future Transportat Middle partSecond Ring South Rd Xian 710064 Shaanxi Peoples R China Northwestern Polytech Univ Sch Comp Sci 1 Dongxiang Rd Xian 710129 Shaanxi Peoples R China Vrije Univ Brussel Fac Sci & Bioengn Sci Bd Plaine 2 B-1050 Brussels Belgium

Federated Learning (FL) is a popular method for privacy-preserving machine learning on edge devices. However, the heterogeneity of edge devices, including differences in system architecture, data, and co-running applications, can significantly impact the energy efficiency of FL. To address these issues, we propose an energy-efficient personalized federated search framework. This framework has three key components. Firstly, we search for partial models with high inference efficiency to reduce training energy consumption and the occurrence of stragglers in each round. Secondly, we build lightweight search controllers that control the model sampling and respond to runtime variances, mitigating new straggler issues caused by co-running applications. Finally, we design an adaptive search update strategy based on graph aggregation to improve personalized training convergence. Our framework reduces the energy consumption of the training process by lowering the training overhead of each round and speeding up the training convergence rate. Experimental results show that our approach achieves up to 5.02% accuracy and 3.45x energy efficiency improvements.

关键词： Personalized FL federated search energy-efficient graph-based aggregation

来源：评论

学校读者我要写书评

暂无评论

IP2FL: Interpretation-Based Privacy-Preserving Federated Learning for Industrial Cyber-Physical systems

IEEE Transactions on Industrial Cyber-Physical Systems

引用

IEEE transactions on Industrial Cyber-Physical systems 2024年 2卷 321-330页

作者： Namakshenas, Danyal Yazdinejad, Abbas Dehghantanha, Ali Parizi, Reza M. Srivastava, Gautam University of Guelph Cyber Science Lab Canada Cyber Foundry GuelphONN1G2W1 Canada Kennesaw State University Decentralized Science Lab College of Computing and Software Engineering MariettaGA30060 United States Brandon University Department of Math and Computer Science ManitobaMBR7A6A9 Canada China Medical University Research Centre for Interneural Computing Taichung40402 Taiwan Lebanese American University Department of Computer Science and Mathematics Beirut1102 Lebanon

The expansion of Industrial Cyber-Physical systems (ICPS) has introduced new challenges in security and privacy, highlighting a research gap in effective anomaly detection while preserving data confidentiality. In the ICPS landscape, where vast amounts of sensitive industrial data are exchanged, ensuring privacy is not just a regulatory compliance issue but a critical shield against industrial espionage and cyber threats. Existing solutions often compromise data privacy for enhanced security, leaving a significant void in protecting sensitive information within ICPS networks. Addressing this, our research presents the IP2FL model, an Interpretation-based Privacy-Preserving Federated Learning approach tailored for ICPS. This model combines Additive Homomorphic Encryption (AHE) for privacy with advanced feature selection methods and Shapley Values (SV) for enhanced explainability. The proposed solution mitigates privacy concerns in federated learning, where traditional methods fall short due to computational constraints and lack of interpretability. By integrating AHE, the IP2FL model minimizes computational overhead and ensures data privacy. Our dual feature selection approach optimizes system performance while incorporating SV to provide critical insights into model decisions, advancing the field towards more transparent and understandable AI systems in ICPS. The validation of our model using ICPS-specific datasets demonstrates its effectiveness and potential for practical applications. © 2023 IEEE.

关键词： embedded systems

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：