With the increasing size of Deep Neural Networks (DNNs) and datasets, DNN training will consume a lot of time. Various distributed strategies have been utilized to speed up the training process. Nevertheless, the freq...
详细信息
Publish/subscribe systems are widely used as messaging middleware in various scenarios, such as online business, financial trading, and social media. However, the network overhead intro-duced by the pub/sub system can...
详细信息
In recent years, wireless sensor networks have been deployed on a large scale. Their ability to capture information and transmit it has made them key players in the smart city and industry. Most often, data streamed f...
详细信息
ISBN:
(数字)9781728186719
ISBN:
(纸本)9781728186719
In recent years, wireless sensor networks have been deployed on a large scale. Their ability to capture information and transmit it has made them key players in the smart city and industry. Most often, data streamed from sensor nodes are collected in a centralized cloud to form a vector of features. They are then used to make future predictions using advanced machine learning (ML) techniques such as Deep Learning (DL). Due to the increasing volumes of data being collected and transmitted over the network, this approach turns out to have a high communication cost, which wastes energy and often saturates the network. In this work, we study an alternative approach that enables collaborative inference throughout the network devices (edge, fog and cloud). Our proposition consists of a fuzzy logic controller which will decide at what level of the network (edge, fog or cloud) the inference should be carried out. Three parameters are taken into account to make this decision: the available energy level of the edge device (smart node), the available network bandwidth and the amount of available data for inference. This work presents preliminary results achieved with the proposed approach using generated configuration input data. We show that the implementation of our controller on edge devices (smart nodes) that run RNN-LSTM for multivariate time series predictions can reduce their energetic cost by around 50%.
Continual wavering of outside weather degrades the efficiency of inside building envelope over time and leads to additional energy consumption, various structural damages, etc. Frequent monitoring of the indoor built ...
Continual wavering of outside weather degrades the efficiency of inside building envelope over time and leads to additional energy consumption, various structural damages, etc. Frequent monitoring of the indoor built environment with thermal images can assist in identifying the energy-leaking and potentially damage-prone areas. Although in recent years different researches performed deep learning and computer vision based thermal anomaly detection in built environment, several issues related to conducting strategic non-intrusive indoor thermal inspection using temporal thermal images, are still unresolved in uncontrolled environment of residential buildings. In this work, we propose a scalable thermal image-based monitoring approach for building envelopes combining the visual knowledge of structural joint information among different building components and their corresponding temporal thermal status. We collected longitudinal thermal images from indoor scenes of different building components (e.g., door, window, wall) and employed a high-level spatio-temporal graph (st-graph) to represent the structural connection among different building components and their temporal self-changes. Our proposed novel unsupervised spatio-temporal clustering framework assigns the cluster label to nodes in st-graph, combining its structural (the self and neighboring component) and temporal features which achieves better performance in identifying thermal variation compared to other clustering based approaches. We demonstrate thermal variation across the spots which indicates the potential energy leakage areas inside the built environment. The cluster patterns obtained from our proposed model assist in understanding the thermal characteristics of various surfaces at certain conditions, such as sun reflection and airflow in the inside built environment.
With the growth of cities and population, urban corporation faces potential challenges. One such challenge is regulating parking spaces. This become a tedious job during weekends and festival season. Improper regulati...
详细信息
Cold data contributes a large portion of the big data today and is usually stored in secondary storage. Various sketch data structures are implemented to represent the stored elements and provide constant-time members...
详细信息
Time Slotted Channel Hopping (TSCH) is a medium access protocol defined in the ieee 802.15.4 standard which have been proven to be one of the most reliable options when it comes to industrial applications. TSCH has be...
详细信息
ISBN:
(数字)9781665495127
ISBN:
(纸本)9781665495134
Time Slotted Channel Hopping (TSCH) is a medium access protocol defined in the ieee 802.15.4 standard which have been proven to be one of the most reliable options when it comes to industrial applications. TSCH has been designed to be utilized in static network topologies. Thus, if an application scenario requires a mobile network topology, TSCH does not perform reliably. In this paper we introduce active connectivity for mobile application scenarios, such as mobile robots. This is a feature that enables the option to regulate physical characteristics such as the speed of a node as it moves, in order to keep being connected to the TSCH network. We model the active connectivity approach through a basic example where two nodes are moving towards the same direction to infer the main principles of the introduced approach. We evaluate the active connectivity feature through simulations and quantify trade-off between connectivity and application-layer performance.
The primary contribution of this paper is designing and prototyping a real-time edge computing system, RhythmEdge, that is capable of detecting changes in blood volume from facial videos (Remote Photoplethysmography;r...
详细信息
ISBN:
(纸本)9781665481526
The primary contribution of this paper is designing and prototyping a real-time edge computing system, RhythmEdge, that is capable of detecting changes in blood volume from facial videos (Remote Photoplethysmography;rPPG), enabling cardiovascular health assessment instantly. The benefits of RhythmEdge include non-invasive measurement of cardiovascular activity, real-time system operation, inexpensive sensing components, and computing. RhythmEdge captures a short video of the skin using a camera and extracts rPPG features to estimate the Photoplethysmography (PPG) signal using a multi-task learning framework while offloading the edge computation. In addition, we intelligently apply a transfer learning approach to the multi-task learning framework to mitigate sensor heterogeneities to scale the RhythmEdge prototype to work with a range of commercially available sensing and computing devices. Besides, to further adapt the software stack for resource-constrained devices, we postulate novel pruning and quantization techniques (Quantization: FP32, FP16;Pruned-Quantized: FP32, FP16) that efficiently optimize the deep feature learning while minimizing the runtime, latency, memory, and power usage. We benchmark RhythmEdge prototype for three different cameras and edge computing platforms while evaluating it on three publicly available datasets and an in-house dataset collected under challenging environmental circumstances. Our analysis indicates that RhythmEdge performs on par with the existing contactless heart rate monitoring systems while utilizing only half of its available resources. Furthermore, we perform an ablation study with and without pruning and quantization to report the model size (87%) vs. inference time (70%) reduction. We attested the efficacy of RhythmEdge prototype with a maximum power of 8W and a memory usage of 290MB, with a minimal latency of 0.0625 seconds and a runtime of 0.64 seconds per 30 frames.
The use of autonomous systems is burgeoning in the world today for many applications in many fields from scientific, industrial, to military. At the same time, advances in semiconductor technology have enabled ever sm...
详细信息
ISBN:
(纸本)9781665439923
The use of autonomous systems is burgeoning in the world today for many applications in many fields from scientific, industrial, to military. At the same time, advances in semiconductor technology have enabled ever smaller, complex and dedicated microprocessors. This work details a control system architecture that takes advantage of these advances for use in resource-constrained autonomous systems. The architecture consists of a real time hardware controller, a guidance and navigation computer, and an edge TPU for machine learning inferences. While the latter two processors are commercially available, a dedicated, modular real time controller is not. Therefore we present an open source design for a real time controller that is intended to be adapted to many types of autonomous systems. We present three different vehicle platforms that implement this control system including a ground vehicle, a surface vessel, and a quadcopter. Finally, we present results using this control architecture on a few key topics of interest in autonomous systems. The first is a novel spatial estimation algorithm called partitioned ordinary kriging that is designed for resource-constrained systems and can be used for path finding during mapping missions. The second result pertains to a sensor calibration utilizing sensor data collected from the real time controller and performed on the navigation computer. Finally, we demonstrate using the edge TPU (tensor processing unit) for object detection using an onboard camera and an object detection algorithm using machine learning.
The proceedings contain 96 papers. The topics discussed include: distributed MQTT brokers at network edges: a study on message dissemination;multi-stage low error localization based on krill herd optimization algorith...
ISBN:
(纸本)9781665454179
The proceedings contain 96 papers. The topics discussed include: distributed MQTT brokers at network edges: a study on message dissemination;multi-stage low error localization based on krill herd optimization algorithm in WSNs;a decentralized framework with dynamic and event-driven container orchestration at the edge;a novel harmony search cat swarm optimization algorithm for optimal bridge sensor placement;on-ramp merging for connected autonomous vehicles using deep reinforcement learning;on-device training of deep learning models on edge microcontrollers;user position-based wireless sensor network deployment algorithm;edge-cloud cooperation for DNN inference via reinforcement learning and supervised learning;node deployment and confident information coverage for WSN-based air quality monitoring;and demand-oriented allocation with fairness in multi-operator dynamic spectrum sharing systems.
暂无评论