Analog, parallel, sorting circuit for the application in Neural Gas learning algorithm implemented in the CMOS technology

Concurrency and Computation

Dalecki

et al. 2019

Self Cite

Summary This paper presents a contribution in the development of a wireless sensor network, which can be used for building, in real time, dense air pollution maps, for compact urban areas. Such system may be useful for cyclists and pedestrians moving through the city. Based on such data, they can select the route in such a way, as to avoid the most polluted areas. An important step here may be development of miniaturized and cheap intelligent sensors, capable not only of data recording and transmitting, but also of some on‐site data processing and prediction. Such sensors require a development of small and power efficient circuit, including data processing unit integrated with an artificial neural network (ANN) in a single chip. We present a prototype chip that contains main components of such sensors, which include a programmable 10‐bit analog‐to‐digital converter, a programmable clock generator, and selected blocks of the ANN. The chip is a reconfigurable device, with many testing abilities. For this reason, one of the main challenges was a fast and efficient programming and testing tool. Such tool has been developed by us and is described in this work in detail along with selected measurement results. The presented work is an extended version of our conference paper (“Novel solutions for smart cities—creating air pollution maps based on intelligent sensors”).

Section: Proposed Contribution To the Development Of The Intelligent mentioning

confidence: 89%

“…The adaptation process is performed according to a formula

W_{j} false(k + 1 false) = W_{k} false(l false) + η false(k false) \cdot G false(false) \cdot false[X false(k false) - W_{j} false(k false) false],

where W k is the weights vector of a j th neuron. The neurons that belong to the winner's neighborhood, are trained with the intensities determined by the applied neighborhood function G () …”

Section: Proposed Contribution To the Development Of The Intelligent mentioning

confidence: 99%

New technologies for smart cities – high‐resolution air pollution maps based on intelligent sensors

Banach

Concurrency and Computation

Dalecki

et al. 2019

Self Cite

“…In this formula W k is the weights vector of a j th neuron. The neurons that belong to the winner's neighborhood, are trained with the intensities determined by the applied neighborhood function G() [8], [9]. The prototype chip, shown in Fig.…”

Section: Intelligent Pollution Sensorsmentioning

confidence: 99%

Novel Solutions for Smart Cities—Creating Air Pollution Maps Based on Intelligent Sensors

Banach

Annals of Computer Science and Information Systems

Długosz

2018

Self Cite

The paper presents novel solutions for systems used to create air pollutions maps in smart cities. Ability to record pollution levels with a function of a short-term prediction of their fluctuations may be useful for cyclists and pedestrians moving through the city. Based on such data they can choose their route through the city in such a way, as to avoid the most polluted areas. Systems of this type are in the range of solutions characteristic for smart cities. Their effectiveness requires a relatively dense wireless sensor network (WSN) composed of miniaturized and cheap intelligent pollution sensors, capable not only of data recording and transmitting, but also of some data processing with the prediction abilities. Sensors of this type require a development of various circuit components that feature small sizes and ultra-low energy consumption. One of the main blocks, in this case, should be an artificial neural network (ANN) implemented at the transistor level. In this work, we present prototype circuits designed by us for the described purposes. The realized blocks include a finite impulse response (FIR) filter, programmable analog-to-digital converters (ADCs) with internal controlling clock generators and main building blocks of a parallel ANN. The specialized chips (ASIC -application specific integrated circuit) with the described components were implemented in the CMOS technology in the full custom style.

“…Self-organizing ANNs feature relatively simple structure, which is suitable for low power and low chip area hardware realizations, as shown in our previous works [3,4,24]. This group of ANNs include the following learning algorithms: winner takes all (WTA), winner takes most (WTM) and the neural gas (NG).…”

Section: Introductionmentioning

confidence: 99%

“…In this formula, the Wj(k) is the weights vector of a j th neuron, in k cycles (iteration) of the learning process of the ANN, while η is the learning rate that determines the intensity of the learning process. The neurons that belong to the winner’s neighborhood, are trained with the intensities determined by the applied neighborhood function (NF) G() [3,24]. The values of the NF for particular neurons depend on distances between these neurons and the winning neuron.…”

Section: Introductionmentioning

confidence: 99%

Components of Artificial Neural Networks Realized in CMOS Technology to be Used in Intelligent Sensors in Wireless Sensor Networks

2018

Sensors

Self Cite

The article presents novel hardware solutions for new intelligent sensors that can be used in wireless sensor networks (WSN). A substantial reduction of the amount of data sent by the sensor to the base station in the WSN may extend the possible sensor working time. Miniature integrated artificial neural networks (ANN) applied directly in the sensor can take over the analysis of data collected from the environment, thus reducing amount of data sent over the RF communication block. A prototype specialized chip with components of the ANN was designed in the CMOS 130 nm technology. An adaptation mechanism and a programmable multi-phase clock generator—components of the ANN—are described in more detail. Both simulation and measurement results of selected blocks are presented to demonstrate the correctness of the design.