The Proceedings of the 2018 International Conference on ...embedded.seoultech.ac.kr/bbs/paper/foreign_conference/selection.pdf · where the master sends a command to all the slaves

IEEE Catalog Number: CFP18B77-ART (Xplore)

ISBN: 978-1-5386-4710-3 (Xplore)

IEEE Catalog Number: CFP18B77-CDR (CD-ROM)

ISBN: 978-1-5386-4709-7 (CD-ROM)

Proceedings

2018 International Conference on Platform Technology and Service (PlatCon-18)

xiv

Table of Contents

Platform Technology and Service

A Study on Dynamic Role-Based User Service Authority Control and

Real-Time Service Configuration························································································ 1

Yejin Kwon; Young B Park

Data Analysis Platform Using Open Source Based Deep Learning Engine ···································· 6

Ahyoung Kim; Junwoo Lee

Detecting Flood Vulnerable Areas in Social Media Stream Using Association Rule Mining ············· 11

Maria Rosario Rodavia; Arne Barcelo; Lilibeth Cuison

Analyzing Overhead from Security and Administrative Functions in Virtual Environment ············· 17

Manjinder Singh; Sergey Butakov; Fehmi Jaafar

Resource-Oriented Architecture for Smart Home Operations Management Platforms··················· 23

Chun-Feng Liao; Ying-Ru Chen

A Case Study for the Approach of Converting a Mobile RPG to a Multiplayer VR Game ··············· 29

Wonseok Chae; Halim Choi; Baek Seob Kim; Jonghee Sang; Junwoo Lee

Movie Recommendation Using Metadata Based Word2Vec Algorithm ······································· 33

Yeo-Chan Yoon; Junwoo Lee

Design of Lighting Control System Considering Lighting Uniformity and

Discomfort Glare for Indoor Space ···················································································· 38

Woon Jo Byun; Ye-Seul Jin; Young-Won Kim; JaeHyun Lim

A Design of IoT Based Contextual Adaptation Management System ·········································· 42

Hyunjoong Kang; SoonHyun Kwon; Jaehak Yu; Woongshik You

Deep Residual Convolutional Network for Natural Image Denoising and

Brightness Enhancement ································································································· 46

Wenjie Xu; Jie You; Yujia Zhang; malrey lee; Jae-young Choi; Sungyoung Suk

BER of Alamouti STBC MC-CDMA with Imperfect Channel State Information ·························· 52

Chirawat Kotchasarn

Reference Security Architecture for Body Area Networks in Healthcare Applications ··················· 57

Tasha-Gaye Braham; Sergey Butakov; Ron Ruhl

2018 International Conference on Platform Technology and Service (PlatCon-18)

xv

Development of 9-Axis Sensor-Based Motion Extraction Program for

Generating a Motion Accuracy Determination Factor of the Personal Training···························· 65

Hyeong. Seok. Kim; Seung. Taek. Oh; JaeHyun Lim

Design of Band-type Device to Measure UV Index and UVB Irradiance in Everyday ····················· 69

DaeHwan Park; Kyeong-Mi Kim; Seung. Taek. Oh; JaeHyun Lim

Analysis of Association Between Students' Mathematics Test Results

Using Association Rule Mining ························································································· 73

SungSik Park; Young B Park

A Framework on Semantic Thing Retrieval Method in IoT and IoE Environment ························ 77

Jaehak Yu; SoonHyun Kwon; Hyunjoong Kang; Sunjin Kim; Ji-Hoon Bae; Cheol Sig Pyo

An Efficient Video Series Decision Algorithm Based on Actor-Face Recognition··························· 82

Hong-Rae Lee; Hwee Kim; Yo-Sub Han; Yongseok Seo; Dong-Hyuck Im

Robot Soccer Using Deep Q Network·················································································· 86

Jinwon Kim; Bongsu Kim; Jinwoo Yoon; Sunah Jung; Jae-young Choi; malrey lee

Online Disaster Preparedness Application for Kids································································ 91

Maria Rosario Rodavia; Isabelle Soleil Curato; Ferdinand Pitagan

MPC Based Ramp-Rate Control Strategy with Battery Energy Storage Systems for

Wind Farms ················································································································· 97

Marlon Capuno; Jung-Su Kim; Hwachang Song

Ultra-Low Power Communication for Infrastructure Monitoring During a Disaster ··················· 101

Hwanwoong Hwang; Ji-Hoon Yun; Hwachang Song

Application of EtherCAT in Microgrid Communication Network: A Case Study ························ 105

Raimarius Delgado; Byoung Wook Choi; Hwachang Song

Discrete-time Control Design for Three-phase Grid-connected Inverter

Using Full State Observer ······························································································ 110

Seung-Jin Yoon; Ngoc-Bao Lai; Kyeong-Hwa Kim; Hwachang Song

Context-driven Mobile Learning Using Fog Computing ························································ 114

Hameed Pinjari; Anand Paul; Gwanggil Jeon; Seungmin Rho

Automated Security Configuration Checklist for Apple iOS Devices Using SCAP V1.2 ················ 120

Akinbobola Adetunji; Sergey Butakov; Pavol Zavarsky

Application of EtherCAT in Microgrid

Communication Network: A Case Study

Raimarius Delgado, Byoung Wook Choi, and Hwachang Song

Dept. of Electrical and Information Engineering

Seoul National University of Science and Technology

Seoul, South Korea

{raim223, bwchoi, hcsong}@seoultech.ac.kr

Abstract—Microgrids can operate either connected to the

utility grid or disconnected, respectively called grid-connected

and islanding mode. If an emergency occurs, a rapid transition to

islanding mode is executed to evaluate the cause of problem and

to prevent further damage. However, seamless transition control

is not realizable because of speed constraints introduced by the

communication protocols being used today. This paper reviews

the current technologies applied in a microgrid communication

system and presents a new study to solve this problem using

EtherCAT, a real-time Ethernet protocol. The advantages of this

solution with respect to the state-of-the-art are summarized. We

also conducted a performance analysis using minimum cycle time

as the performance index to verify the feasibility of EtherCAT in

a microgrid application. Potential future work is suggested

related to integration of EtherCAT in an actual microgrid system

for practical field experiments.

Keywords—microgrid; EtherCAT; real-time; communication;

control system

I. INTRODUCTION

Microgrid research has developed exponentially in recent years due to environmental and social concerns such as natural disasters, global warming, and energy shortage resulting from the world-wide increase of consumer demands. The world is now looking for a clean and efficient solution that utilizes renewable energy sources to produce electricity that can replace the current technology that is having various issues like restrictions in planning, variable market, and lesser scope of generation expansion. A lot of microgrids are currently being constructed as an access way for this new source of energy [1-4].

As shown in Fig. 1, a microgrid is an interconnection of, storage devices, protection units, distributed generators such as solar photovoltaics and wind turbines, a fuel cell generator, and controllable loads. Operation mode of a microgrid depends on its connection to the main power grid (utility grid). In grid-connected mode, the microgrid operates side-by-side the electric power system and power is distributed in four quadrants [5]. In islanding mode operation, the microgrid is separated from the utility grid and supplies only the load connected within its boundaries.

When operating normally, the microgrid is operating in grid-connected mode and is forcibly switched to islanding mode in case of a fault in either the utility grid, or the microgrid itself. Transition between these modes can result to harmful transient overcurrent which is a huge issue in microgrid control. Many researchers have presented control strategies for seamless mode transition [6-8], but in all these works, the importance of implementing a fast and accurate solution is highlighted. However, current protocols used inside a microgrid communication system is relatively slow that can limits smooth transition between operation modes [9].

In this paper, we review the communication standards and protocols in control and monitoring systems used in microgrids. An emerging protocol called EtherCAT based on real-time Ethernet scheme is introduced and compared to the traditional technologies being used. The performance of this new solution is verified by conducting a simple experiment with minimum cycle time as the performance indicator. The minimum cycle time is defined as the time that it takes for a data to pass from the main controller, around the whole network and back to main controller.

Fig. 1. Basic microgrid components and communication network.

2018 International Conference on Platform Technology and Service (PlatCon)

978-1-5386-4710-3/18/$31.00 ©2018 IEEE

105

II. CURRENT COMMUNICATION PROTOCOLS IN A MICROGRID

A. Modbus

Modbus is an application layer messaging protocol, positioned at level 7 of the OSI model (Fig. 2) providing server and client communication between devices connected on distinct types of buses or networks. Communication between the connected devices can either be master to a single slave (can be compared to unicast protocol), or a broadcast type where the master sends a command to all the slaves in parallel. In a microgrid network, all components communicate with each other independently using either of these methods. A transaction comprises one single query and single response frame, or one single broadcast frame. A Modbus frame message contains the address of the intended receiver, the command the receiver must execute and eventually the data needed for the execution of such command.

Modbus can be encapsulated into different physical links such as RS-232, RS-485, and TCP/IP. Modbus/TCP basically embeds a Modbus frame into a TCP frame [10]. All the functions supported by the Modbus protocol are identified by an index number (e.g. 01 (0x01) Read Coils is a function used to read the status of the coils in a remote device). The Modbus communication interface was published in 1979 by Modicon. With the original purpose of working for a multidrop network based on a master/client architecture. Because, when designed, the target was isolated systems, all security aspects were not taken into consideration.

B. DNP3

Interoperability between devices of different manufacturers is provided by the DNP3. It has the particularity of supporting multi-slave, peer-to-peer and multi-master communications. As for the OSI 7-layer model (Fig. 2), during transmission, data is passed from the user application through these three layers of the stack to produce a stream of data over the physical communications medium. Also, Pseudo-Transport layer is used that allows large block of data transmission. However, with the interconnection of SCADA networks to the public networks, an IP version of DNP3 appeared called DNP3-over-TCP/IP. The functions described above are designated in the transport layer. The application layer handles disassembly and reassembly functions through the DNP3 Pseudo-Transport layer.

In comparison with Modbus, DNP3 messages are not restricted to transferring one data point at a time. Multiple data types (i.e. boolean, floating point) can be encapsulated in a single message to reduce data traffic. Data quality information and timestamps can also be included in the message. DNP3 slave devices can send updates as values change, without having to wait for a poll from the Master. The DNP3 data are organized into a large library of objects. This library is divided into separate groups where each group represents a specific object. For example group number 1 represents all the binary input in one slave. Moreover, within each group, each object may be represented following different structures. Thus, inside each object group there are several variations which represent the separate ways to describe an object [11].

Fig. 2. Open Systems Interconnection (OSI) 7-layer model.

C. IEC 61850

The IEC 61850 is a standard that defines data models, exchanges, and events between power systems substations. This standard can be mapped to a number of legacy protocols such as Manufacturer Message Specification (MMS), Generic Object Oriented Substation Events (GOOSE), and Sampled Measured Values (SMV). However, IEC 61850 was not designed for serial communication protocols as it is intended to run over Ethernet networks. This should have a positive impact on the cost and operation of power system components at the application layer and above. Leveraging the services of this standard is proposed to achieve reliability and security. Services in this standard are: retrieving the device description, fast and reliable host–to–host status information exchange of status, reporting data or sequence of events, data logging, retrieving sample values from sensors, time synchronization, and file transfer for online configuration of components.

Communication protocols define how data bits are transmitted on the transmission medium. However, they do not define the data organization in an IED or device with communication capability. For microgrid components, each DER has unique functionality, and a subset of data attributes may not be present in all DERs. Each IED manufacturer may have different naming conventions for data. The IEC 61850 data naming conventions are based on a power systems context, which helps ensure interoperability between devices from various vendors used in the same communication system [12].

III. ETHERCAT

In comparison to other Real-time Ethernet Protocols, EtherCAT offers very high real-time performance and determinism developed by Beckhoff Automation [12]. The typical EtherCAT network supports single-master network configurations where the master communicates with the slaves by sending them suitable telegrams.

106

The EtherCAT master sends a telegram that passes through each node. Each EtherCAT slave device reads the data addressed to it on the fly, and inserts its data in the frame, as the frame is moving downstream. The frame is delayed only by hardware propagation delay. The last node in a segment detects an open port and sends the message back to the master using Ethernet technology’s full duplex feature. The telegram’s maximum effective data rate increases to over ninety percent, and due to the utilization of the full duplex feature, the theoretical effective data rate is even higher than 100 Mbit/s.

EtherCAT offers efficient clock synchronization for slaves down to hundreds of nanoseconds through special timing functionality, known as the Distributed Clock (DC) mechanism. The first slave connected to the EtherCAT Master serves as the reference clock for the entire network. Before activating the master, the controller calculates the offsets between the reference clock and the other DC-enabled slaves. The calculated offset is written to the slaves with which the value of the global clock is calculated based on the measurement at the local clock. The controller periodically distributes the value of the reference clock to all the slaves for the compensation of the clock skew caused by the local clock drift . Although the DC mechanism is easily implemented, it can accurately synchronize the clocks of numerous slaves with a difference of less than one microsecond.

In comparison to the traditional communication protocols used in microgrid communication systems described in the last section, EtherCAT has major advantages including flexible topology, efficient and fast data transmission, and highly accurate clock synchronization between connected nodes.

A. Transmission efficiency and speed

When the frames from the master station pass through the slave nodes, they are not encoded or decoded. Only the data are read and written as the bits passing the EtherCAT slave controller (ESC). After the data are accessed, the frames are automatically forwarded to the next port by the EtherCAT Processing Unit in ESC. In comparison to the traditional network protocols, EtherCAT only uses three layers in the OSI model (Fig. 2), physical, data link, and application layers. In perspective, less processing means faster speed.

EtherCAT uses logical addressing, for an effective data exchange between master and slaves. The master does not directly address the slaves but, rather, the slaves read and write their data from and to a logical memory. Practically, each slave has been assigned a set of fieldbus memory management units (FMMU) which are logical structures directly mapped on the slave physical input/output data. In an EtherCAT telegram, the master may specify a logical address referring to several FMMUs, possibly located on different slaves. During operation, each slave analyzes the logical address of the frame which is passing through: if the address matches with one or more of its FMMUs, then that slave extracts the output data and writes the input ones. In this way, a unique header (telegram) may be used to address several slaves with the consequent improvement of the protocol efficiency.

Fig. 3. Structure of an EtherCAT frame.

As shown in Fig. 3, the structure of the EtherCAT frame

can be as large as 1498 bytes excluding the EtherCAT header. The EtherCAT datagram begins with a 10-byte header specifying the type of operation it is used for, read or write. It is followed by the data filed which represents the data to or from the slave. To check the correct execution of operation, the datagram is ended with a 2-Byte working counter.

B. Synchronization

In comparison to other Ethernet protocols, EtherCAT has an accurate clock synchronization technique called distributed clock (DC). To guarantee accurate clock synchronization, the following operation must be finished beforehand. Firstly, master station passes the propagation delay measurement frame to all slaves and the slave controller records the receive time of the frame. And the master station collects the time stamps afterwards and calculates the propagation between all slaves. Secondly, the local offset time to reference clock of every slave station must be compensated. In order to do that, the time of each slave clock is compared with that of the reference clock, and the difference is stored in each slave to make the slave get the same absolute system time. Finally, the drift between reference clock and slave clock cannot be avoided because each slave station has respective free running oscillator with different parameters. With the parameters information, the master can calculate the delay of each node. The master repeats the calculation for every frame it sends. As the network operates, the enormous sample size means that the master has incredibly accurate data. The inherent ring topology creates an incredibly efficient clock mechanism that increased in accuracy with every message [13].

IV. TIMING ANALYSIS

To verify the feasibility of EtherCAT as a better option of communication protocol in a microgrid network, we performed a timing analysis using minimum cycle time as the performance index. Minimum cycle time is defined as the time necessary to transmit a data from the master, along the whole network, and back to the master. As shown in the structure of an EtherCAT frame, we can theoretically express the minimum cycle time as:

MCT header ethercat delay idle

T T T T T= + + + (1)

107

Fig. 4. Experimental testbed using EtherCAT-based servo motors for timing

analysis.

TABLE I. EXPERIMENTAL TESTBED SPECIFICATION

Master CPU Intel Atom D2700 @ 2.13 GHz

Memory 2 GB DDR3 SDRAM

Network Card Realtek RTL8169 Gigabit

Slaves Servo Drive Sanmotion R Advanced Model

PDOs 24 Bytes for each slave (144 Bytes)

(TxPDO 12 Bytes, RxPDO 12 Bytes)

Here, Theader is the time to transmit all fields of the Ethernet frame excluding the EtherCAT payload. Tethercat denotes the actual time for transmitting the EtherCAT payload and Tdelay denotes the wire propagation delay time which we assume as 1 µs at 100 Mbps for each slave and Tidle is referred to as the interframe gap, or the time to be waited before issuing a new frame which is also 1 µs in the original Ethernet specification.

For practical results, we implemented an open source EtherCAT master to an embedded hardware with specifications detailed in Table 1. Six servo motors are connected to the EtherCAT master using CANopen-over-EtherCAT connection to serve as the load of the system, with each configured to transmit 24 Bytes of process data objects at each iteration.

We have observed that the computing time of the control task inside the EtherCAT master will have an effect at the actual cycle time of the whole network. Thus, we consider the execution time, Texec, defined as the overall time duration to acquire the data stored in the buffer from the network interface card (NIC), processing of the next command, and the for sending out the frame through the NIC. Thus, the actual minimum cycle time from (1) is transformed to:

MCT header ethercat delay idle exec

T T T T T T= + + + + (2)

The control task that handles EtherCAT data transmission is configured to have the highest priority running in a real-time operating system with a sampling time of 1 millisecond. The test program was carried out for 10 minutes producing 600,000 samples. The experiment results are processed offline and statistically calculated to acquire the maximum, minimum, average, and standard deviation values. From the 600,000 samples, we have calculated 298.709 µs, 53.958 µs, 61.494 µs

respectively for the metrics above, with a standard deviation of 3.565 µs. From here, we only consider the maximum execution time for the calculation of minimum cycle time for it is better to assume the worst-case value in any performance test.

For our experimental testbed running EtherCAT in 100 Mbps, the total minimum cycle time is calculated considering that Theader is 2.08 µs with 26 Bytes of header data. 144 Bytes of EtherCAT load and 12 Bytes of EtherCAT header results to a delay of 12.48 µs. Thus, the minimum cycle time considering the worst-case execution time for the testbed is 322.269 µs.

V. CONCLUSION AND DISCUSSION

In this paper, we present a new protocol that is for microgrid application called EtherCAT. EtherCAT is a real-time Ethernet protocol that possesses great advantages over traditional microgrid communication protocols such as Modbus, DNP3, and IEC 61850 in terms of transmission speed and efficiency, flexibility, and synchronization.

Theoretically, the structure of EtherCAT is assumed to be faster than the other protocols since it only uses 3 layers in the OSI 7-layer model, thus reducing processing delay. It also uses a technique called logical addressing, where the master is not required to directly address the slaves but, rather, the slaves read and write their data from and to a logical memory, which minimizes data loss in the transmission line. We also conducted a timing analysis using minimum cycle time of an EtherCAT network using an actual experimental testbed composed of an EtherCAT master and 6 EtherCAT slave devices. Results shows that 144 Bytes of data in the transmission line has a worst-case minimum cycle time of 322.269 µs. The results of this study would serve as a promising guideline in application of EtherCAT in microgrid communication systems. As for future work, we would be developing a prototype microgrid using EtherCAT slaves for field experiments of different microgrid control strategies.

ACKNOWLEDGMENT

This work was supported by the Human Resources Development of the Korea Institute of Energy Technology Evaluation and Planning (KETEP) grant funded by the Korea government Ministry of Trade, Industry & Energy. (NO. 20154030200720).

REFERENCES

[1] F. Katirael, M.R. Iravani, and P.W. Lehn, “Micro-grid autonomous

operation during and subsequent to islanding process,” IEEE Trans. Power Delivery. vol. 2, pp. 248-257, Jan. 2005.

[2] X. Liu, P. Wang, and P.C. Loh, “A hybrid AC/DC microgrid and its coordination control,” IEEE Transactions on Smart Grid. vol. 2, pp. 278-286, Jun. 2011.

[3] B. Satish, and S. Bhuvaneswari, “Control of microgrid – A review,” International Conference on Advances in Green Energy. Thiruvananthapuram, India, pp. 18-25, Dec. 2014.

[4] Z. Lu, C. Wang, Y. Min, S. Zhou, J. Lu, and Y. Wang, “Overview on microgrid research,” Automation of Electric Power Systems. vol. 31, pp. 100-106, Oct. 2007.

108

[5] X. Zhang, J, Guan, and B. Zhang, “A master slave peer to peer integration microgrid control strategy based on communication,” IEEE PES Asia-Pacific Power and Energy Conference. Xi’an, China, pp. 1106-1110, Oct. 2016.

[6] J. Rocabert, G.M.S. Azevedo, and A. Luna, “Intelligent connection agent for three-phase grid-connected microgrids,” IEEE Trans. Power Electron. vol. 26, pp. 2993-3005, Feb. 2011.

[7] C. Cho, J.H. Jeon, and J.Y. Kim, “Active synchornizing control of a microgrid,” IEEE Trans. Power. Electron. vol. 26, pp. 3707-3719, Dec. 2011.

[8] T.M.L. Assis, and G.N. Taranto, “Automatic reconnection from intentional islanding based on remote sensing of voltage and frequency signals,” IEEE Trans. Smart. Grid, pp. 1877-1884, May. 2012.

[9] M.A. Ahmed, Y.C. Kang, and Y-.C. kim, “Communication network architectures for smart-house with renewable energy resources,” Energies. vol. 8, pp. 1877-1884, Aug. 2015.

[10] N. Goldenberg, and A. Wool, “Accurate modeling of Modbus/TCP for intrusion detection in SCADA systems,” International Journal of Critical Infrastructure Protection. vol. 6, pp. 63-75, June. 2013.

[11] R. Amoah, S. Camtepe, and E. Foo, “Securing DNP3 broadcast communications in SCADA systems,” IEEE Transactions on Industrial Informatics. vol. 12, pp. 1474-1485, Jul. 2016.

[12] F. Clavel, E. Savary, P. Angays, and A. Vieux-Melchior, “Intergration of a new standard: A network simulator of IEC61850 architectures for electrical substations,” IEEE Industry Applications Magazine. vol 21, pp. 41-48, Jan. 2015.

[13] M. Cereia, I.C. Bertolotti, and S. Scanzio, “Performance of a real-time EtheraCAT master under Linux,” IEEE Transactions on Industrial Informatics. vol. 7, pp. 679-687, Nov. 2011.

.

109

/ColorImageDict > /JPEG2000ColorACSImageDict > /JPEG2000ColorImageDict > /AntiAliasGrayImages false /CropGrayImages true /GrayImageMinResolution 150 /GrayImageMinResolutionPolicy /OK /DownsampleGrayImages true /GrayImageDownsampleType /Bicubic /GrayImageResolution 300 /GrayImageDepth -1 /GrayImageMinDownsampleDepth 2 /GrayImageDownsampleThreshold 2.00333 /EncodeGrayImages true /GrayImageFilter /DCTEncode /AutoFilterGrayImages true /GrayImageAutoFilterStrategy /JPEG /GrayACSImageDict > /GrayImageDict > /JPEG2000GrayACSImageDict > /JPEG2000GrayImageDict > /AntiAliasMonoImages false /CropMonoImages true /MonoImageMinResolution 1200 /MonoImageMinResolutionPolicy /OK /DownsampleMonoImages true /MonoImageDownsampleType /Bicubic /MonoImageResolution 600 /MonoImageDepth -1 /MonoImageDownsampleThreshold 1.00167 /EncodeMonoImages true /MonoImageFilter /CCITTFaxEncode /MonoImageDict > /AllowPSXObjects false /CheckCompliance [ /None ] /PDFX1aCheck false /PDFX3Check false /PDFXCompliantPDFOnly false /PDFXNoTrimBoxError true /PDFXTrimBoxToMediaBoxOffset [ 0.00000 0.00000 0.00000 0.00000 ] /PDFXSetBleedBoxToMediaBox true /PDFXBleedBoxToTrimBoxOffset [ 0.00000 0.00000 0.00000 0.00000 ] /PDFXOutputIntentProfile (None) /PDFXOutputConditionIdentifier () /PDFXOutputCondition () /PDFXRegistryName () /PDFXTrapped /False

/CreateJDFFile false /Description >>> setdistillerparams> setpagedevice















































































/ColorImageDict > /JPEG2000ColorACSImageDict > /JPEG2000ColorImageDict > /AntiAliasGrayImages false /CropGrayImages true /GrayImageMinResolution 150 /GrayImageMinResolutionPolicy /OK /DownsampleGrayImages true /GrayImageDownsampleType /Bicubic /GrayImageResolution 300 /GrayImageDepth -1 /GrayImageMinDownsampleDepth 2 /GrayImageDownsampleThreshold 2.00333 /EncodeGrayImages true /GrayImageFilter /DCTEncode

The Proceedings of the 2018 International Conference on ...embedded.seoultech.ac.kr/bbs/paper/foreign_conference/selection.pdf · where the master sends a command to all the slaves

Documents