Towards statistical machine learning for edge analytics in large scale networks: Real-time Gaussian function generation with generic DSP
01 Apr 2018-
TL;DR: In this paper, Gaussian distribution functions are designed using C28x real-time digital signal processor (DSP) that is embedded in the TMS320C2000 modem designed for powerline communication (PLC) at the low voltage distribution end of the smart grid, where numerous devices that generate massive amount of data exist.
Abstract: The smart grid (SG) is a large-scale network and it is an integral part of the Internet of Things (IoT). For a more effective big data analytic in large-scale IoT networks, reliable solutions are being designed such that many real-time decisions will be taken at the edge of the network close to where data is being generated. Gaussian functions are extensively applied in the field of statistical machine learning, pattern recognition, adaptive algorithms for function approximation, etc. It is envisaged that soon, some of these machine learning solutions and other gaussian function based applications that have low computation and low-memory footprint will be deployed for edge analytics in large-scale IoT networks. Hence, it will be of immense benefit if an adaptive, low-cost, method of designing gaussian functions becomes available. In this paper, gaussian distribution functions are designed using C28x real-time digital signal processor (DSP) that is embedded in the TMS320C2000 modem designed for powerline communication (PLC) at the low voltage distribution end of the smart grid, where numerous devices that generate massive amount of data exist. Open-source embedded C programming language is used to program the C28x for real-time gaussian function generation. The designed gaussian waveforms are stored in lookup tables (LUTs) in the C28x embedded DSP, and could be deployed for a variety of applications at the edge of the SG and IoT network. The novelty of the design is that the gaussian functions are designed with a generic, low-cost, fixed-point DSP, different from state of the art in which gaussian functions are designed using expensive arbitrary waveform generators and other specialized circuits. C28x DSP is selected for this design since it is already existing as an embedded DSP in many smart grid applications and in other numerous industrial systems that are part of the large scale IoT network, hence it is envisaged that integration of any gaussian function based solution using this DSP in the smart grid and other IoT systems may not be too challenging.
Citations
More filters
28 May 2019
TL;DR: The paper proposes the introduction of a new component in the family of distributed computer networks, with specific functions for edge computing devices, called DEW, designed as a Mist counterpart that can be used in Distributed Dynamical Networks (DDN).
Abstract: The paper proposes the introduction of a new component in the family of distributed computer networks, with specific functions for edge computing devices. The DEW component is a network layer located between Fog and Edge, a position currently attributed to the Mist concept. Specifically, DEW is designed as a Mist counterpart that can be used in Distributed Dynamical Networks (DDN). The DEW characteristics and the associated specific algorithms are discussed on a particular type of DDN - microgrids and microgrids clusters.
13 citations
02 Jul 2018
TL;DR: Con Convolution results indicates that the distributed computing approach for low-cost generic devices considered in this paper is suitable for use in large IIoT networks.
Abstract: Low-cost, real-time digital signal processors (DSPs) embedded in generic Internet of Things (IoT) edge devices can make significant contributions to distributed edge computing for industrial IoT (IIoT) networks. The DSP considered in this paper is the Texas Instruments (TI) TMS320C28x DSP (C28x). At the edge of the network, the C28x is programmed using low-level Embedded C programming language to construct the Morlet wavelet. Our implementation at this layer is the first known construction of the Morlet wavelet for C28x DSP using Embedded C. At the fog layer, near the edge of the IoT network, where more computing resources exist, the wavelet is then convolved with healthcare (electrocardiogram) and electrical network signals, using Matlab to reduce signal noise, and to identify important parts of examined signals. Convolution results indicates that the distributed computing approach for low-cost generic devices considered in this paper is suitable for use in large IIoT networks
10 citations
01 Oct 2018
TL;DR: The osmotic collaborative computing method advocated in this paper will be crucial in ensuring the possibility of shifting many complex applications such as novelty detection and other machine learning based cybersecurity applications to edges of large scale IoT networks using low-cost widely available DSPs.
Abstract: To implement machine learning algorithms and other useful algorithms in industrial Internet of Things (IIoT), new computing approaches are needed to prevent costs associated with having to install state of the art edge analytic devices. A suitable approach may include collaborative edge computing using available, resource-constrained IoT edge analytic hardware. In this paper, collaborative computing method is used to construct a popular and very useful waveform for IoT analytics, the Gaussian Mixture Model (GMM). GMM parameters are learned in the cloud, but the GMMs are constructed at the IIoT edge layer. GMMs are constructed using C28x, a ubiquitous, low-cost, embedded digital signal processor (DSP) that is widely available in many pre-existing IIoT infrastructures and in many edge analytic devices. Several GMMs including 2-GMM and 3-GMMs are constructed using the C28x DSP and Embedded C to show that GMM designs could be achieved in form of an osmotic microservice from the IIoT edge to the IIoT fog layer. Designed GMMs are evaluated using their differential and zero-crossings and are found to satisfy important waveform design criteria. At the fog layer, constructed GMMs are then applied for novelty detection, an IIoT cybersecurity and fault-monitoring application and are found to be able to detect anomalies in IIoT machine data using Hampel identifier, 3-Sigma rule, and the Boxplot rule. The osmotic collaborative computing method advocated in this paper will be crucial in ensuring the possibility of shifting many complex applications such as novelty detection and other machine learning based cybersecurity applications to edges of large scale IoT networks using low-cost widely available DSPs.
10 citations
Cites background or methods from "Towards statistical machine learnin..."
...Clipped signals are then fetched from the C28x LUT, sequenced, and concatenated together [37] - [40] (using Embedded C) to generate different 2-GMM and 3-GMM models for the C28x DSP....
[...]
...0139) ensures that the positive half cycle part (half sine) will approximate the needed Gaussian distribution with a low mean square error (MSE) as discussed extensively in [24]....
[...]
...Needed parts of sine waves that represents different Gaussian shapes are then clipped and stored in the C28x DSP lookup table (LUT) [24]....
[...]
...The signal processing method adopted is first to construct sine waves that have different frequencies for the C28x DSP using Embedded C. Needed parts of sine waves that represents different Gaussian shapes are then clipped and stored in the C28x DSP lookup table (LUT) [24]....
[...]
...8 peak SDI value hence could be obtained for the needed GMM plots in CCS since for the C28x DSP, the DAC analog voltage waveform [24], [48] is given is...
[...]
03 Oct 2019
TL;DR: A new perspective on edge architectures is presented and a model for a new edge gateway is designed, which aims to facilitate new distributed computing methods while being able to handle both operational and functional requirements.
Abstract: With the emergence of IoT applications Cloud architecture proves to be inefficient in handling massive amounts of data, mainly because of the variable latency and limited bandwidth. More specific, major requirements of Industrial Internet of Things (IIoT) like control and real-time decision making could not be addressed. These limitations along with the increasing intelligence in the lower levels of the data transmission architecture led to the development of an intermediate edge processing layer, closer to the process, enabling distributed computing and near real-time communication. In this paper a new perspective on edge architectures is presented and a model for a new edge gateway is designed. This device aims to facilitate new distributed computing methods while being able to handle both operational and functional requirements. Three case studies analyse how this device can be used to improve existing solutions: a hydroponic greenhouse, Smart Grid implementation for power systems and a video surveillance system in a manufacturing application.
10 citations
TL;DR: STEAM is proposed, a framework for developing data stream processing applications in the edge targeting hardware-limited devices and enables the development of applications for different platforms, with standardized functions and class structures that use consolidated IoT data formats and communication protocols.
Abstract: Edge devices are usually limited in resources. They often send data to the cloud, where techniques such as filtering, aggregation, classification, pattern detection, and prediction are performed. This process results in critical issues such as data loss, high response time, and overhead. On the other hand, processing data in the edge is not a simple task due to devices’ heterogeneity, resource limitations, a variety of programming languages and standards. In this context, this work proposes STEAM, a framework for developing data stream processing applications in the edge targeting hardware-limited devices. As the main contribution, STEAM enables the development of applications for different platforms, with standardized functions and class structures that use consolidated IoT data formats and communication protocols. Moreover, the experiments revealed the viability of stream processing in the edge resulting in the reduction of response time without compromising the quality of results.
5 citations
References
More filters
Book•
23 Nov 2005TL;DR: The treatment is comprehensive and self-contained, targeted at researchers and students in machine learning and applied statistics, and deals with the supervised learning problem for both regression and classification.
Abstract: A comprehensive and self-contained introduction to Gaussian processes, which provide a principled, practical, probabilistic approach to learning in kernel machines. Gaussian processes (GPs) provide a principled, practical, probabilistic approach to learning in kernel machines. GPs have received increased attention in the machine-learning community over the past decade, and this book provides a long-needed systematic and unified treatment of theoretical and practical aspects of GPs in machine learning. The treatment is comprehensive and self-contained, targeted at researchers and students in machine learning and applied statistics. The book deals with the supervised-learning problem for both regression and classification, and includes detailed algorithms. A wide variety of covariance (kernel) functions are presented and their properties discussed. Model selection is discussed both from a Bayesian and a classical perspective. Many connections to other well-known techniques from machine learning and statistics are discussed, including support-vector machines, neural networks, splines, regularization networks, relevance vector machines and others. Theoretical issues including learning curves and the PAC-Bayesian framework are treated, and several approximation methods for learning with large datasets are discussed. The book contains illustrative examples and exercises, and code and datasets are available on the Web. Appendixes provide mathematical background and a discussion of Gaussian Markov processes.
11,357 citations
Additional excerpts
..., [9], [10], [11], [12], are designed using widely available, industrial, real-time DSP....
[...]
TL;DR: The definition of edge computing is introduced, followed by several case studies, ranging from cloud offloading to smart home and city, as well as collaborative edge to materialize the concept of edge Computing.
Abstract: The proliferation of Internet of Things (IoT) and the success of rich cloud services have pushed the horizon of a new computing paradigm, edge computing, which calls for processing the data at the edge of the network. Edge computing has the potential to address the concerns of response time requirement, battery life constraint, bandwidth cost saving, as well as data safety and privacy. In this paper, we introduce the definition of edge computing, followed by several case studies, ranging from cloud offloading to smart home and city, as well as collaborative edge to materialize the concept of edge computing. Finally, we present several challenges and opportunities in the field of edge computing, and hope this paper will gain attention from the community and inspire more research in this direction.
5,198 citations
"Towards statistical machine learnin..." refers background in this paper
...With memory and computing power shortages, adequate real-time analytics, inference and decisions that will lead to IoT system optimization will be almost impossible [2], [3]....
[...]
Book•
01 Jan 2007
TL;DR: This is the first book to develop both the theory and the practice of synthesizing musical sounds using computers, covering a wide range of applications.
Abstract: This is the first book to develop both the theory and the practice of synthesizing musical sounds using computers. Each chapter starts with a theoretical description of one technique or problem area and ends with a series of working examples (over 100 in all), covering a wide range of applications. A unifying approach is taken throughout; chapter two, for example, treats both sampling and wavetable synthesis as special cases of one underlying technique. Although the theory is presented quantitatively, the mathematics used goes no further than trigonometry and complex numbers. The examples and supported software along with a machine-readable version of the text are available on the web and maintained by a large online community. The Theory and Techniques of Electronic Music is valuable both as a textbook and as professional reading for electronic musicians and computer music researchers.
113 citations
"Towards statistical machine learnin..." refers methods in this paper
...The sine wave part corresponding to the height and width of the gaussian distribution is clipped [21] and stored in C28x LUT....
[...]
01 Jan 2017
TL;DR: The article gives valuable information on building(construction) IoT systems based on PLC technology and some problems of development IoT are noted.
Abstract: Internet of Things (IoT) — a unified network of physical objects that can change the parameters of the environment or their own, gather information and transmit it to other devices. It is emerging as the third wave in the development of the internet. This technology will give immediate access to information about the physical world and the objects in it leading to innovative services and increase in efficiency and productivity. The IoT is enabled by the latest developments, smart sensors, communication technologies, and Internet protocols. This article contains a description of lnternet of things (IoT) networks. Much attention is given to prospects for future of using IoT and it's development. Some problems of development IoT are were noted. The article also gives valuable information on building(construction) IoT systems based on PLC technology.
40 citations
"Towards statistical machine learnin..." refers background in this paper
...One of the suggested solutions to this problem is that various communications systems, such as wireless, wired (including PLC [4]) and fiber optics should be made to work together (as shown in Fig....
[...]
...TMS320C2000 is designed for PLC at the low voltage end of the SG....
[...]
...In this paper, gaussian distribution functions are designed using C28x real-time digital signal processor (DSP) that is embedded in the TMS320C2000 modem designed for powerline communication (PLC) at the low voltage distribution end of the smart grid, where numerous devices that generate massive amount of data exist....
[...]
01 Dec 2017
TL;DR: An industrial machine condition-monitoring open-source software database, equipped with a dictionary and small enough to fit into the memory of edge data-analytic devices is created, which will prevent excessive industrial and smart grid machine data from being sent to the cloud.
Abstract: The Industrial Internet of Things (IIoT) is quite different from the general IoT in terms of latency, bandwidth, cost, security and connectivity. Most existing IoT platforms are designed for general IoT needs, and thus cannot handle the specificities of IIoT. With the anticipated big data generation in IIoT, an open source platform capable of minimizing the amount of data being sent from the edge and at the same time, that can effectively monitor and communicate the condition of the large-scale engineering system by doing efficient real-time edge analytics is sorely needed. In this work, an industrial machine condition-monitoring open-source software database, equipped with a dictionary and small enough to fit into the memory of edge data-analytic devices is created. The database-dictionary system will prevent excessive industrial and smart grid machine data from being sent to the cloud since only fault report and requisite recommendations, sourced from the edge dictionary and database will be sent. An open source software (Python SQLite) situated on Linux operating system is used to create the edge database and the dictionary so that inter-platform portability will be achieved and most IIoT machines will be able to use the platform. Statistical analysis at the network edge using well known industrial methods such as kurtosis and skewness reveal significant differences between generated machine signal and reference signal. This database-dictionary approach is a new paradigm since it is different from legacy methods in which databases are situated only in the cloud with huge memory and servers. The open source deployment will also help to satisfy the criteria of Industrial IoT Consortium and the Open Fog Architecture.
32 citations
"Towards statistical machine learnin..." refers background in this paper
...1) to provide adequate communication bandwidth for interacting IoT devices [5], [6], [7]....
[...]