+ All Categories
Home > Documents > Lucas Henrique Cavalcanti Santostg/2019-2/TG_EC/tg-lhcs-final.pdf · IoT introduz conectividade nos...

Lucas Henrique Cavalcanti Santostg/2019-2/TG_EC/tg-lhcs-final.pdf · IoT introduz conectividade nos...

Date post: 14-Aug-2020
Category:
Upload: others
View: 0 times
Download: 0 times
Share this document with a friend
47
Lucas Henrique Cavalcanti Santos Optimized Real Time Radio Frequency Network for Multiple Mobile Robots Communication Federal University of Pernambuco [email protected] www.cin.ufpe.br/~secgrad Recife 2019
Transcript
Page 1: Lucas Henrique Cavalcanti Santostg/2019-2/TG_EC/tg-lhcs-final.pdf · IoT introduz conectividade nos dispositvos e hoje já tem expecitativa de vender 214 bilhões de dólares em 2020.

Lucas Henrique Cavalcanti Santos

Optimized Real Time Radio Frequency Network for Multiple MobileRobots Communication

Federal University of [email protected]

www.cin.ufpe.br/~secgrad

Recife2019

Page 2: Lucas Henrique Cavalcanti Santostg/2019-2/TG_EC/tg-lhcs-final.pdf · IoT introduz conectividade nos dispositvos e hoje já tem expecitativa de vender 214 bilhões de dólares em 2020.

Lucas Henrique Cavalcanti Santos

Optimized Real Time Radio Frequency Network for Multiple MobileRobots Communication

A B.Sc. Dissertation presented to the Center of Informaticsof Federal University of Pernambuco in partial fulfillmentof the requirements for the degree of Bachelor in ComputerEngineering.

Concentration Area: Wireless NetworkAdvisor: Edna Natividade da Silva Barros

Recife2019

Page 3: Lucas Henrique Cavalcanti Santostg/2019-2/TG_EC/tg-lhcs-final.pdf · IoT introduz conectividade nos dispositvos e hoje já tem expecitativa de vender 214 bilhões de dólares em 2020.

ACKNOWLEDGEMENTS

I want to thank my family, who supported and guided me through my life and gave methe best conditions they can. I want to thank especially my mother and girlfriend to share andhold up the whole graduation journey by my side.

I would like to thank my advisor, Prof. Edna Barros, who guided me during all thegraduation, trusting and supporting me in all my projects and ideas. I also would like to thankProf. Hansenclever Bassani, because without him and Edna there isn’t RobôCIn.

I want to thank the Centro de Informática (CIn), and all the professors, for the besteducation, facilities, and infrastructure, that leads the course and its students to excellence incomputation. I would like to thank especially Prof. Daniel Cunha that accepted the invitation toexamine that work.

I cannot forget to thank my RobôCIn’s colleges and classmates, who shared their routinewith me and taught me to become a better person. Also, I would like to thank the RobôCIn’syoung members, that motivates me daily and brought a purpose for my position on the team.Finally, I want to thank my friends Roberto Fernandes, Renato Sousa, Heitor Rapela, GabrielBandeira, Cristiano Santos, Carlos Pena, Jailson Gomes, Geovanny Lucas, Raphael Brito, PedroMagalhães, Mariana Barros, Walber Macedo, Victor Sabino, and many others that shares theirjourneys, studies, worries and dreams with me.

Page 4: Lucas Henrique Cavalcanti Santostg/2019-2/TG_EC/tg-lhcs-final.pdf · IoT introduz conectividade nos dispositvos e hoje já tem expecitativa de vender 214 bilhões de dólares em 2020.

"If you want to go fast, go alone. If you want to go far, go together."

–African Proverb

Page 5: Lucas Henrique Cavalcanti Santostg/2019-2/TG_EC/tg-lhcs-final.pdf · IoT introduz conectividade nos dispositvos e hoje já tem expecitativa de vender 214 bilhões de dólares em 2020.

ABSTRACT

Industrial robots increased 30% from 2017 to 2018 due to the rise of Industry 4.0 andthe Internet of Things (IoT). Both concepts may change the production chain and the people’sinteraction with the environment. IoT that expects to reach 214 billion dollars in 2020, tries toembedded connectivity to every device. Additionally, Industry 4.0 uses the IoT contributionstogether with robotics, to increase production customization and autonomy. In this context, thechallenge of adequately controlling and monitoring robots arises. In some robotics competitions,these challenges are present and attract researches and teams to bring their research and solutions.In one of these competitions, the RobôCIn team plays soccer with its autonomous system.Composed by a computer and six mobile robots, the RobôCIn’s system controls their robotsthrough wireless communication and aims to monitor them. Because of the dependence aroundthe wireless network, this work intends to build an optimized architecture that deliveries wirelessmessages quickly and receives robot telemetry as well. Starting from the computer that sendsencoded data, passing through a base station, where the data entry in wireless media, until itreaches the robot that receives and decodes it in actions, this work proposes an architecture andprotocol to fulfill the real-time requirements while monitoring the robots. The delivery time ofthe messages was exhaustively tested to find the best approach and its limitations. It results indelivering control messages, with telemetry enabled, in 4,39ms, i.e., a control system of 227messages per second.

Keywords: Wireless communication, motion control network, mobile robots, embeddedsystems, autonomous robots, RoboCup competition.

Page 6: Lucas Henrique Cavalcanti Santostg/2019-2/TG_EC/tg-lhcs-final.pdf · IoT introduz conectividade nos dispositvos e hoje já tem expecitativa de vender 214 bilhões de dólares em 2020.

RESUMO

Robôs industriais cresceram 30% de 2017 a 2018, e este aumento no número de robôsveio junto com a Industria 4.0 e Internet da Coisas (IoT). Conceitos que podem mudar a formacom que os produtos são produzidos e como as pessoas interagem com o ambiente a sua volta. AIoT introduz conectividade nos dispositvos e hoje já tem expecitativa de vender 214 bilhões dedólares em 2020. Além de que a Industria 4.0 usa o concento de IoT, junto com robótica, paraaumentar a customização e autonomia nos processos industriais. Neste contexto, o desafio deadequadamente mover e monitorar os robôs surge. Assim, algumas competições de robóticaatraem pesquisadores a levarem suas pesquisas e soluções. Em uma destas competições a equipeRobôCIn joga futebol de robôs com seu sistema autônomo. Composto por um computadore 6 robôs, o sistema da equipe controla seus robôs através de comunicação sem-fio, e desejamonitora-los da mesma forma. Por conta da depedência em redes sem-fio, este trabalho visaconstruir uma arquitetura otimizada que entregue mensagens sem-fio de forma rápida e comtelemetria presente. Começando pelo computador que envia dados, passando pela estação baseque os move para a comunicação sem-fio, até o robô que recebe e interpreta as mensagens, essetrabalho propõe uma arquitetura junto com um protocolo que entregue mensagens, em temporeal, enquanto recebe mensagens de monitoramento. O tempo de entrega de mensagens foi deexaustivamente testado, para descobrir a melhor abordagem e suas limitações. Então o resultadoalcançado em entrega de mesanges de controle, com a telemetria ativa, foi de 4,39ms, ou seja,um sistema de controle com 227 mensagens por segundo.

Palavras-chave: Comunicação sem-fio, rede de controle, robôs móveis, sistemas embarcados,robôs autonomos, RoboCup.

Page 7: Lucas Henrique Cavalcanti Santostg/2019-2/TG_EC/tg-lhcs-final.pdf · IoT introduz conectividade nos dispositvos e hoje já tem expecitativa de vender 214 bilhões de dólares em 2020.

LIST OF FIGURES

Figure 1 – General Structure of the Small Size League Competition AutonomousSystem (WEITZENFELD, 2015). . . . . . . . . . . . . . . . . . . . . 13

Figure 2 – SPI Signals (Leens, 2009). . . . . . . . . . . . . . . . . . . . . . . . . 16Figure 3 – RS232 and UART singal voltage (SparkFun, 2010). . . . . . . . . . . . 16Figure 4 – Transistor-Transistor Logic method with sent bits and its function (Spark-

Fun, 2014). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17Figure 5 – The types of power supplies for different wireless technologies (Proskochylo

et al., 2015). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19Figure 6 – The specifications of WiFi, ZigBee, and Bluetooth modules (Proskochylo

et al., 2015). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20Figure 7 – nRF24l01 packet bytes and its functions (Nordic, 2008) . . . . . . . . . 21Figure 8 – Control scheme of a mobile robot (Siegwart et al., 2011a). . . . . . . . 22Figure 9 – Dependence of lost packets rate between distance and throughput (Kordas

et al., 2010). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24Figure 10 – SSL network layout using the Tigers’ team approach of a network-

capable base station (Andre Ryll, 2016). . . . . . . . . . . . . . . . . 25

Figure 11 – Project network overview with a computer that sends commands to abase station which transmits the message wirelessly to multiple robots. 26

Figure 12 – RobôCIn’s Small Size Soccer (SSL) robot. . . . . . . . . . . . . . . . 29Figure 13 – Base station architecture of the proposed wireless network with Serial

and Ethernet options between the Computer and Base Station. . . . . . 30Figure 14 – Diagram of radios used for control and telemetry network on the Base

Station and Robot, together with its way of communication. . . . . . . 31Figure 15 – Proposed Base Station hardware. . . . . . . . . . . . . . . . . . . . . 32Figure 16 – Architecture of the method developed for testing the network configura-

tions and interfaces. . . . . . . . . . . . . . . . . . . . . . . . . . . . 34Figure 17 – Reception delay for 30 tests at each different sending interval, using a

Serial base station transmitting computer messages at the configuredsending intervals. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

Figure 18 – Reception delay for 30 tests at each different sending interval, using aEthernet base station transmitting computer messages at the configuredsending intervals. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36

Figure 19 – Delivery time of base station approaches with the minimum sendinginterval. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36

Page 8: Lucas Henrique Cavalcanti Santostg/2019-2/TG_EC/tg-lhcs-final.pdf · IoT introduz conectividade nos dispositvos e hoje já tem expecitativa de vender 214 bilhões de dólares em 2020.

Figure 20 – Delivery time of base station with different distance between transceivers. 39Figure 21 – Delivery time analysis between interfaces (a) and network size (b). . . 40Figure 22 – Comparing delivery time with two and six nodes network at two different

robots. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40Figure 23 – Delivery time of Ethernet base station packets with different telemetry

sampling time at 6 robots control network. . . . . . . . . . . . . . . . 41Figure 24 – Angular speed on time of each wheel is the robot, monitored by the

developed telemetry network. . . . . . . . . . . . . . . . . . . . . . . 41

Page 9: Lucas Henrique Cavalcanti Santostg/2019-2/TG_EC/tg-lhcs-final.pdf · IoT introduz conectividade nos dispositvos e hoje já tem expecitativa de vender 214 bilhões de dólares em 2020.

LIST OF TABLES

Table 1 – User Datagram Protocol packet with bits offset and its content. . . . . . 18Table 2 – Nordic nRF24l01 Specifications. . . . . . . . . . . . . . . . . . . . . . 21

Table 3 – Message protocol created to control RobôCIn’s robots. . . . . . . . . . . 27Table 4 – Telemetry protocol created to receive RobôCIn’s robots’ information. . . 28Table 5 – Project Modules and Configuration in Base Station and Robots . . . . . 30Table 6 – Ethernet Base Station with Telemetry Different Sampling Interval . . . . 37

Page 10: Lucas Henrique Cavalcanti Santostg/2019-2/TG_EC/tg-lhcs-final.pdf · IoT introduz conectividade nos dispositvos e hoje já tem expecitativa de vender 214 bilhões de dólares em 2020.

LIST OF ACRONYMS

CPU Control Processing UnitCRC Cyclic Redundancy CheckCSV Comma-Separated ValuesIEEE Institute of Electrical and Electronics EngineersIoT Internet of ThingsLARC Latin American Robotics CompetitionMAC Medium Access ControlmbedOS mbed Operating SystemRS232 Recommended Standard 232SPI Serial Peripheral InterfaceSSL Small Size SoccerTCP Transmission Control ProtocolTTL Transistor-Transistor LogicUART Universal Asynchronous Receiver / TransmitterUDP User Datagram Protocol

Page 11: Lucas Henrique Cavalcanti Santostg/2019-2/TG_EC/tg-lhcs-final.pdf · IoT introduz conectividade nos dispositvos e hoje já tem expecitativa de vender 214 bilhões de dólares em 2020.

LIST OF ALGORITHMS

Algorithm 1 – Transmission Concept for Base Station Serial Interface Approach . . 31Algorithm 2 – Transmission Concept for Base Station Ethernet Interface Approach . 32

Page 12: Lucas Henrique Cavalcanti Santostg/2019-2/TG_EC/tg-lhcs-final.pdf · IoT introduz conectividade nos dispositvos e hoje já tem expecitativa de vender 214 bilhões de dólares em 2020.

CONTENTS

1 INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

2 BACKGROUND . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152.1 COMMUNICATION INTERFACES . . . . . . . . . . . . . . . . . . . . . 152.1.1 Serial Peripheral Interface . . . . . . . . . . . . . . . . . . . . . . . . . . 152.1.2 Universal Asynchronous Receiver / Transmitter . . . . . . . . . . . . . . 162.2 ETHERNET NETWORK . . . . . . . . . . . . . . . . . . . . . . . . . . . 172.3 WIRELESS MODULES . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182.4 MOBILE ROBOTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212.4.1 Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 232.4.2 Telerobotics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 232.5 RELATED WORK . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24

3 SYSTEM ARCHITECTURE . . . . . . . . . . . . . . . . . . . . . . . . 263.1 COMMUNICATION PACKETS AND PROTOCOL . . . . . . . . . . . . . 273.2 SYSTEM IMPLEMENTATION . . . . . . . . . . . . . . . . . . . . . . . . 293.3 TIME ANALYSES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 333.3.1 Test Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 333.3.2 Control Message Interval Time . . . . . . . . . . . . . . . . . . . . . . . 343.3.3 Telemetry Message Interval Time . . . . . . . . . . . . . . . . . . . . . . 37

4 RESULTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 384.1 DIFFERENT DISTANCES . . . . . . . . . . . . . . . . . . . . . . . . . . 384.2 MULTIPLE ROBOTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 394.3 TELEMETRY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41

5 CONCLUSION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42

REFERENCES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44

Page 13: Lucas Henrique Cavalcanti Santostg/2019-2/TG_EC/tg-lhcs-final.pdf · IoT introduz conectividade nos dispositvos e hoje já tem expecitativa de vender 214 bilhões de dólares em 2020.

121212

1INTRODUCTION

Today there are more cell phones than people in the world (BankMyCell, 2019). However,despite the high number of cell phones, other embedded systems arrived people’s live. Therefore,day after day, the interaction with an increasing number of dedicated devices that help people’sdaily tasks is becoming normal. The technology entrance on people’s lives came supported byan increase of the semiconductor industry, which showed significant growth in sales, passing 1trillion units sold in 2018, 10% higher than in 2017 (IC Insights, 2019).

The embedded system exchanges information with the Internet gives the sensors andactuators the IoT concept. The IoT is a reality with an expectation of selling 214 billion dollars in2020 (Radiant Insights, 2015). This market takes the power of internet wireless communicationto facilitate the interaction between the people and the environment.

The semiconductors and IoT concept are fundamental to Industry 4.0, an industry conceptthat brings autonomy and customization to products manufacture (Wang et al., 2017). In otherwords, Industry 4.0 ends with "dumb" robots that only do programmed movements, and bringssensors and external communication so the robots can react and adapt to the necessities.

According to the International Federation of Robotics (IFR), the industrial robots in-creased 30% from 2017 to 2018 (IFR, 2018). IFR also affirms that since 2017, the robots canwork alongside humans, and no more cages are needed. Also, in its report, IFR (2018) concludesthat robot-human interactions are increasing. On the other hand, robots are not in daily lifebecause people create a highly dynamic environment that is even more challenging than anindustrial environment. As people live in homes, that environment is a complicated place whererobots need sensors to interact autonomously safely.

In an autonomous system, we cannot prioritize some parts over others. Every pieceshould work as a symphony, for example, if there is a lack of sensors, there is not enoughinformation to analyze, and without actuators, there is not an action to take. Additionally,without wireless communication in systems, robots will not move correctly in the environment,even though it has input information and functional agents.

Competitions like RoboCup, the most significant autonomous robotics competition, cre-ates high dynamic environments where robots should play soccer autonomously against anotherteam (Kitano et al., 1997). The soccer competition is perfect for developing interdisciplinary

Page 14: Lucas Henrique Cavalcanti Santostg/2019-2/TG_EC/tg-lhcs-final.pdf · IoT introduz conectividade nos dispositvos e hoje já tem expecitativa de vender 214 bilhões de dólares em 2020.

131313

technologies, as it has obstacles and target position moving at every second. That dynamicsrequires accurate sensing of the field situation, combined with fast decision algorithms andreal-time communication.

Figure 1: General Structure of the Small Size League Competition Autonomous System(WEITZENFELD, 2015).

RoboCup (2018) soccer categories, like Small Size Soccer (SSL), where a team of sixrobots plays soccer autonomously against another team, the sensing comes from a vision systemthat captures images from the field and process camera images in a computer that detects thepositions of the objects. After the positions go to each team, decision algorithms are executedtogether with a wireless station to send movements to each team robot on the field, as illustratedin Figure 1.

At the Small Size Soccer (SSL), where a complex robotic system is required, motioncontrol is considered the output of the system. Furthermore, to wireless control robots, highaccuracy and speed is needed. Additionally, systems capable of precisely move the robots need aclosed loop of control; in other words, it sends the desired movements to the robots and measuresit to fix any mistake. The closed-loop benefits are proportional to the updating frequency ofthe loop, so, besides fast and accurate algorithms and sensors, it is essential to use real-timecommunication in order to move the robots precisely. Robots movements act in order to playsoccer, and, as competitions are, the best team with the best system performance and strategywins.

The main goal of this work is to leverage the network efficiency and quality of the SSLrobots made by the team of the Centro de Informática (CIn), the RobôCIn. The first step relies onefficiency, as it reduces the wireless latency when sending control packets to the robots. For that,this work proposes an optimized wireless controller communication through the study of thetechnologies available. With wireless and embedded technologies, this work goes more in-depth,

Page 15: Lucas Henrique Cavalcanti Santostg/2019-2/TG_EC/tg-lhcs-final.pdf · IoT introduz conectividade nos dispositvos e hoje já tem expecitativa de vender 214 bilhões de dólares em 2020.

141414

building and analyzing peripherals connections, configurations, and protocols that minimizethe delivery time on the robots. In other words, the work develops protocols that, together withinterfaces, works to optimize communications between computer, base stations, and transceivers.Moreover, without it, no matter how fast the robot’s sensing and cognition are, the motion controldoes not perform the path planned.

Taking into consideration that systems should be monitored, this work also aims tobring quality for the network with the development of telemetry without affecting the controllatency. The telemetry, which was not in the RobôCIn’s system, consists of sending informationfrom robots to the computer, closing the loop between the vision, motion control, and robot.The challenge appears when the system has multiple robots using only one base station, whichcontrols robots and receives telemetry.

The contribution of this works is a control network that improves path execution accuracyby the delivery of real-time messages. The network, different from other control networks, con-tributes to detect and prevent failures in robots by its closed-loop communication. Furthermore,the cognition code is benefited by the telemetry contribution that enables monitoring the robotsonline and its displacement.

This work is divided as follows: Chapter 2 is a short introduction to computer andembedded systems communication interfaces, followed by an explanation of wireless communi-cation, and its available modules together with possible configurations, benefits, and limits. Suchinformation supports architecture decisions; we also introduce, in Chapter 2, what is meant bymobile robots, and the challenges to control them, and to receive telemetry information, essentialto understanding the project impact. Also in Chapter 2, other teams’ network is shortly presentedin order to explain the differences in these work requirements and decisions. In Chapter 3,the proposed system is described. Its architecture is explained together with the connectioninterfaces, algorithms, protocols, experimental system parameters, and test methods. In Chapter4, the results are shown, and the analyses of the proposed system at different conditions arediscussed. Finally, in Chapter 5, the conclusion of this work is presented.

Page 16: Lucas Henrique Cavalcanti Santostg/2019-2/TG_EC/tg-lhcs-final.pdf · IoT introduz conectividade nos dispositvos e hoje já tem expecitativa de vender 214 bilhões de dólares em 2020.

151515

2BACKGROUND

This section goes through the technologies and its characteristics necessary to build aradio frequency network that is capable of controlling a mobile robot in real-time.

2.1 COMMUNICATION INTERFACES

Nowadays, there are tons of sensors and actuators present on the market. Additionally,there are modern embedded processors that, combined with these peripherals, are capable ofcreating quality systems and solutions. To spread the numbers of embedded solutions on theworld and to facilitate the use, were created development boards and peripheral modules thatmasks the complexity behind the new microcontrollers, sensors, and actuators. These facilitiescame alongside the popularization of communications interfaces, that made possible a uniformand fast integration.

2.1.1 Serial Peripheral Interface

Serial Peripheral Interface (SPI) is a protocol present in almost all development boardsin the world. Made to facilitate the data exchange between devices, it uses one device as master,which controls the communication bus, and one or more slaves that answer and respect themaster signals.

The SPI works with only four signals (Leens, 2009). The first one, called SCLK, is theclock signal sent from the master to all devices, to provide synchronous signals. The second, theslave select, known as SS, is responsible for choosing the slave which the master communicates.In other words, the master has one signal as SS for each slave that he wants to exchange bytes.The third and fourth signals are analogous, one called MOSI is the signal that outputs from themaster and goes to the slaves as input, and the other is the MISO, which is input signal to themaster that came as output from the slaves.

The downside of this communication protocol is the number of signals needed. It startswith four signals to communicate with one peripheral and increases one SS signal for each slaveadded. Otherwise, this type of communication reaches high throughput (Leens, 2009). The high

Page 17: Lucas Henrique Cavalcanti Santostg/2019-2/TG_EC/tg-lhcs-final.pdf · IoT introduz conectividade nos dispositvos e hoje já tem expecitativa de vender 214 bilhões de dólares em 2020.

161616

speed comes because the MOSI and MISO signals only change when there is clock changing,an effect illustrated by the signals transitions in Figure 2. The synchronization between devicesreduce bit error, and overall, the synchronization and multiple devices’ communication make theSPI a powerful and widely used interface.

Figure 2: SPI Signals (Leens, 2009).

2.1.2 Universal Asynchronous Receiver / Transmitter

The Universal Asynchronous Receiver / Transmitter (UART) is a well-known type ofcommunication that a few years ago was the principal communication between computers andexternal peripherals. Before the USB protocol, printers and computers used to use UARTcommunication with the Recommended Standard 232 (RS232). Although the RS232 is a methodintroduced in 1962, with low speed and no support to multiple devices communication (Hazen,2003), it strongly standardizes the UART communication and suffered updates that made possiblethe continuous use in an ample amount of micro-controllers as an interface with programmers.

Figure 3: RS232 and UART singal voltage (SparkFun, 2010).

Figure 3 shows the original RS232 with its voltage varying between -13v and +13v, butalso shows a RS232 with a reduced operating voltage that floats between 0v and 5v, for 0 and 1

Page 18: Lucas Henrique Cavalcanti Santostg/2019-2/TG_EC/tg-lhcs-final.pdf · IoT introduz conectividade nos dispositvos e hoje já tem expecitativa de vender 214 bilhões de dólares em 2020.

171717

bits. Named as Transistor-Transistor Logic (TTL) method of communication, that reduction ofoperating voltage also reduces the necessity of power components, which usually increases theprice. Otherwise, that improvement was possible only with circuits improvement since 1962,because the voltage reduction brings more noise, interference, and degradation (Santitoro, 2003).

Like all serial interfaces, the TTL method sends and receives bit after bit until it makesone complete byte. The synchronization implemented is the starting and stopping bits thatdelimits one byte sent asynchronously. Figure 4 illustrates the start and stop bit together withthe bits that composes the two byte sent. Serial interface has only in two wires, the first one isthe receiver signal, called RX, and the other is the transmitter, called TX. As every device sendsin the TX and receives in the RX, two devices need to cross the RX and TX to communicate.Additionally, two devices in the same serial bus need a configuration on the same throughput ofbits, or as commonly called, the same baud rate.

The need for two signals and a baud rate configuration makes this UART a simple andeasy implemented interface because, with a small setup and fewer pins, the cost decreases inevery device. The downside comes with the necessity of speed configuration in both devices andno clock synchronization, different from SPI.

Figure 4: Transistor-Transistor Logic method with sent bits and its function (SparkFun, 2014).

Although it is possible to connect multiple devices by adequately connecting the TXsand RXs pins, there is not support implemented. In other words, devices cannot communicate atthe same time, because no collision and interference are supported.

2.2 ETHERNET NETWORK

Ethernet is the most common network technology used in the world. No matter if it isa local, metropolitan, or wide area network, it plays a crucial part in the Internet web. It is anetwork that connects devices on the same bus, directly or through switches and routers, thetechnology enables a vast number of applications.

Standardized by the IEEE organization, Ethernet uses Medium Access Control (MAC), asix-byte number unique to each device, to identify the device in the network. Ethernet initiallymade with coaxial cables, nowadays are made by twisted pairs of cable that reduce noiseinterference. The noise reduction and the IEEE standard updates made the Ethernet reach10Gbps and contribute to the development of a safe and resilient communication (Santitoro,2003). Significant on the Internet connections, and to locals applications, Ethernet link devices,and process, even at the same machine.

Page 19: Lucas Henrique Cavalcanti Santostg/2019-2/TG_EC/tg-lhcs-final.pdf · IoT introduz conectividade nos dispositvos e hoje já tem expecitativa de vender 214 bilhões de dólares em 2020.

181818

Table 1: User Datagram Proto-col packet with bits offset and itscontent.

Bit Offset Content0 Source port

16 Destination port32 Length48 Checksum

At the seven layers of the OSI model of the Internet (Briscoe, 2000), the Ethernet comesright after the hardware. The second layer, called the link layer, has Ethernet and is a strategicpoint to build and control the physicals connections using MAC. Upon the link layers, there is thenetwork layer, where the Internet Protocol (IP) addresses lives and helps to establish, maintain,and terminate connections. Different from MAC, where the device hardware chooses it, the IP isgiven to the device by routers or switches that control network routing and device connections.

The fourth OSI layer, known as the transport layer, is where Transmission ControlProtocol (TCP) and User Datagram Protocol (UDP) lives. Both protocols are the most used inthe Ethernet networks; consequently, it is the most used on the Internet. The TCP can maintainconnections, so it guarantees order, byte streaming, and reliability. On the other side, UDPsends independents packets without maintaining connections, so it does not guarantee order orreliability. However, as illustrated by the bits offset and the its content in Table 1, the UDP has alight packet suitable for Internet application that requires real-time data, like video streaming.

2.3 WIRELESS MODULES

Over the years, people empower more and more devices with wireless technology. Inother words to reach the Internet or share information locally, devices like cell-phones, laptops,and even television use wireless modules intending to create a unified ecosystem of devices.

Wireless is commonly known as WiFi, but they mean different things. While wirelessincludes all technologies that exchange data without wires, WiFi is a wireless technology withpattern and protocol defined by Institute of Electrical and Electronics Engineers (IEEE) (IEEE,2007), same as Bluetooth. Furthermore, even using similar frequencies, WiFi is optimized inother to access the Internet and Bluetooth to exchange data locally between devices. In otherwords, they were developed differently to increase advantages in it is desired applications.

Most of the modern wireless technology uses frequency modulation of waves to exchangedata. Even with the high complexity and high cost to build wireless modules from scratch, thereare different needs, that just one module cannot resolve. So, the companies continuouslybuild different modules that try to fill the application’s requirements. At Figure 5, there is apower comparison and possible power supply to each technology. Commonly with less power

Page 20: Lucas Henrique Cavalcanti Santostg/2019-2/TG_EC/tg-lhcs-final.pdf · IoT introduz conectividade nos dispositvos e hoje já tem expecitativa de vender 214 bilhões de dólares em 2020.

191919

consumption, less is the signal coverage, but in our daily life, we use some of the technologies atFigure 5 to different applications.

Figure 5: The types of power supplies for different wireless technologies (Proskochylo et al.,2015).

With different standards and protocols, there are some options for wireless modulespresent on the market. Usually, the modules are made suitable for building embedded systems,using SPI to integrate it at the microcontroller. The variety and specification of wireless mod-ules bring the necessity of choosing only one module, with one standard, one protocol, andconsequently, a group of optimizations and limitations.

Authors are continuously analyzing the wireless modules present in the market. Threemodules that always come as options to embedded systems are WiFi, Bluetooth, ZigBee.Proskochylo et al. (2015) tested these three modules to create a detailed comparison of technicalspecifications and limitations. The comparison in Figure 6 shows that the WiFi module is faster,has a larger operating frequency band, and more configuration options. On the other hand, theWiFi module consumes ten times more energy than the other modules.

The WiFi technology has a consumption not suitable for batteries used in embeddeddevices. Looking between ZigBee and Bluetooth, the ZigBee has more flexible characteristics,connects to a large number of devices while consuming only 30mA, less than the Bluetoothmodule. Proskochylo et al. (2015) also affirms that ZigBee is low cost and oriented to tasks liketelemetry.

Page 21: Lucas Henrique Cavalcanti Santostg/2019-2/TG_EC/tg-lhcs-final.pdf · IoT introduz conectividade nos dispositvos e hoje já tem expecitativa de vender 214 bilhões de dólares em 2020.

202020

Figure 6: The specifications of WiFi, ZigBee, and Bluetooth modules (Proskochylo et al., 2015).

A few years, Nordic (2008) launched the nRF24l01 and, year after year, increases itsusage inside the embedded systems, including IoT applications. The Nordic module came tocompete with the ZigBee modules because it has similar characteristics. On the other hand, thenRF24l01 does not follow the IEEE pattern as ZigBee, in other to bring more flexibility andNordic’s frequency protocols.

At embedded systems, the Nordic modules send wireless packets from one module toanother if they have a proper configuration. In other words, it sends from one module to anotherif they are operating in the same frequency, and the sender packet has the receiver address. Thisoperation is similar to UDP but different from ZigBee that changes packets only with paireddevices.

In the end, the nRF24l01 module is way more flexible and lower power than ZigBee(Saha et al., 2017). Although it does not reach the WiFi modules speed (Wang et al., 2014), theNordic module has high performance allied to low consumption, shown at Table 2, that fits inembedded applications.

Page 22: Lucas Henrique Cavalcanti Santostg/2019-2/TG_EC/tg-lhcs-final.pdf · IoT introduz conectividade nos dispositvos e hoje já tem expecitativa de vender 214 bilhões de dólares em 2020.

212121

Table 2: Nordic nRF24l01 Specifications.

Characteristics nRF24l01

Frequency 2.4 GHz

Speed Rate 2 Mbit/s

Level of OSI ShockBurst

Modulation GFSK

Power Lever 0 - 18 dBm

Power Consumption 13.5mA TX, standby 26 uA

The SPI appears as an interface to configure and send packets through the nRF24l01.The configuration step chooses the transceiver mode of operation, frequency channel, address,throughput, payload, and specific options like the use of acknowledgment reception packet andCyclic Redundancy Check (CRC). The packet transmission occurs sending the message bytesto the radio sending register, where the nRF24l01 gets the bytes to fill the packet illustrated atFigure 7.

Figure 7: nRF24l01 packet bytes and its functions (Nordic, 2008)

At the nRF24l01 packet, shown at Figure 7, the preamble is a byte of zeros and ones tohelp the receiver demodulation and address is the receiver address, defined at the configurationstep. The packet control field has the message payload, packet identification, and the requirementof acknowledgment packet or not. The payload is the bytes to send, and the CRC is a code thathelps detection of bit corruption at the message.

2.4 MOBILE ROBOTS

As the industry was the first application of robotics. The first robots came to facilitate andspeed up industries process, so in general, they were articulated arms that precisely manipulateproducts. However, the increase of other areas such as computer vision, embedded systems, andneural networks brought to robotics much more power and options to robotics (Siegwart et al.,2011b).

With the increase of different areas, robotics integrated most of them in robots. Theinterdisciplinary enabled the creation of more sophisticated robotics systems. The mobile robots

Page 23: Lucas Henrique Cavalcanti Santostg/2019-2/TG_EC/tg-lhcs-final.pdf · IoT introduz conectividade nos dispositvos e hoje já tem expecitativa de vender 214 bilhões de dólares em 2020.

222222

came as a challenge to unify mobility, with autonomy and interaction; in other words, they arerobots capable of moving and doing tasks without human guidance.

Siegwart et al. (2011a) affirms that a mobile robot control scheme is made from aperception, localization, cognition, and motion control, as shown at Figure 8. Perception is theextraction of the information based on sensors; this information should serve the environmentmodel to start the localization task. The localization uses the perception together with a localbuilt map or an online map process to discover the position of the robot. The robot and objectpositions go to the cognition step, where the robot interacts with the ambient planning movementsthat may accomplish simple or complex tasks. The final step is motion control that the robotshould act as planned, moving the robot’s wheels and arms or any actuator based on the cognitionplanning. Motion control is also considered as the robot output to the real world, as it is theprocess that interacts with the environment.

Figure 8: Control scheme of a mobile robot (Siegwart et al., 2011a).

No matter the task, the process of sensing, planning, and moving robots should workcontinuously, to not fail the task process even with technology impressions, and above all, tocontinue the task with environment changes (Siegwart et al., 2011b). The steps to build a mobilerobot can goes into an embedded system on the robot or not. That advantage reduces the cost andcomplexity of some systems because the designer can have a global vision of the environmentand share information with multiple agents.

Some organizations, as RoboCup (www.robocup.org), creates controlled environ-ments to gradually increase the researches quality and challenges in the robotics domain. Themain challenge proposed by the RoboCup is to build robotics teams that autonomously play

Page 24: Lucas Henrique Cavalcanti Santostg/2019-2/TG_EC/tg-lhcs-final.pdf · IoT introduz conectividade nos dispositvos e hoje já tem expecitativa de vender 214 bilhões de dólares em 2020.

232323

soccer. Soccer is a highly dynamic environment with multiple agents, and all kinds of roboticssystems, from wheeled to humanoid robots, need to implement perception, localization, cog-nition, and motion control, locally or not. In other words, the soccer competition serves as abenchmark for a lot of technologies and researches, opening doors to innovation (Kim et al.,2014).

The success of some mobile robot systems animated the industry, not to fill the factorieswith fully autonomous robots, but also to bring some autonomous systems that can increase thespeed and quality of productions (Brogårdh, 2007).

2.4.1 Control

After years of robotics development, a challenge present from the beginning until todayis properly control actuators. Siciliano and Siegwart affirm that exists mathematics modelsfor controlling manipulators and mobile robots. Made a few years ago, the models take inconsideration robot structure and output control equations to actuators (Sanz, 2009) (Siegwartet al., 2011b). However, the models are a reality; the robot interaction with the real world has alot of complex effects like wheel slipping, which is undesired and cannot be predicted by themodels (Cerkala & Jadlovska, 2015).

Solving actuators control interaction challenges is part mobile robot loop. As someundesired effects depend on mechanics, electronics, and environment, they are specific for therobot in question and not predictable. So, repeating the process of perception and motion controlgives the robot the ability to recover from wrong movements to the desired position.

2.4.2 Telerobotics

The telerobotics comes to overcome the complexity inside robots because it proposesto divide the embedded computation with external computers. As telerobotics proposes tocommunicate robots with an external computer, it may divide the computational load, which canprocess perception, localization, or cognition out of the robot. Besides the reduction of load,the external computer can reduce robot cost, power consumption and give a broad vision ofthe environment, as it can access multiple robots or even access different sensors like externalcameras (Liu Jianbang et al., 2008).

External process downside comes when is needed real-time operations, like at soccerrobots, where the process from sensing to motion control needs to occur in a few milliseconds. Tomaintain the telerobotics system in a few milliseconds, it needs resilient wireless communicationand optimized data exchange for sending crucial information. To achieve that time requirementis essential to develop a dedicated communication embedded system (Liu Jianbang et al., 2008),but together with proper software and infrastructure development (Kang et al., 2013).

Page 25: Lucas Henrique Cavalcanti Santostg/2019-2/TG_EC/tg-lhcs-final.pdf · IoT introduz conectividade nos dispositvos e hoje já tem expecitativa de vender 214 bilhões de dólares em 2020.

242424

2.5 RELATED WORK

This section briefly summarizes related works that have been applying wireless networksto mobile robots. Most authors have been applying wireless networks with IoT concept to createsensors network. As IoT usually require low power technologies, authors also uses its transceiverat robots. In other words, robotics take advantage of the lower consumption, but adds tightrequirements of message delivery time.

Talking about the construction of the network embedded system, the work of Kordas et al.

(2010) proposed a transceiver embedded device to control mobile robots using the nRF24l01(Nordic, 2008). They tested performance varying network topology between peer and broadcastcommunication. At peer communication, each robot receives its message through a specificaddress of the radio. On the broadcast topology, all robots message are condensed in only onepacket and received by all robots simultaneously. Although the broadcast strategy is fast, themaximum radio bytes per packet is 32, and this limits the number of robots in the network.

Figure 9: Dependence of lost packets rate between distance and throughput (Kordas et al., 2010).

Furthermore, Kordas et al. shows that more bytes in packets or further the transceiversare, more packets are lost. On the other hand, the Figure 9 compares 1 Mbps and 2 Mbpstransmission speed, and shows that a higher throughput do not increase the lost packet rate. AsKordas et al. analyzed the transmissions lost rate and delivery time of bytes, they didn’t analyzedthe whole system flow, including the computer, that sends messages to the embedded devicewhere the messages are transmitted.

Moreover, it is possible to find works that analyze different communications technologiesapplied to soccer robotics requirements. Nadarajah & Sundaraj (2012) work does not comparewith nRF24l01 technology, but gives an excellent overview of the network requirements for theapplication, and also put in balancing the pros and cons of each technology compared (Nadarajah

Page 26: Lucas Henrique Cavalcanti Santostg/2019-2/TG_EC/tg-lhcs-final.pdf · IoT introduz conectividade nos dispositvos e hoje já tem expecitativa de vender 214 bilhões de dólares em 2020.

252525

& Sundaraj, 2012).The Tigers robotic soccer team proposes a shared interface of communication, using

a transceiver base station connected to the network (Andre Ryll, 2016), shown in Figure 10.Although Andre et al. affirms that it speeds up the communication between perception andtransceiver due to the Ethernet network, Tigers’ base station has no isolation between its computerand adversary computer.

Figure 10: SSL network layout using the Tigers’ team approach of a network-capable basestation (Andre Ryll, 2016).

The Tigers’ protocol sends and receives much information with high precision betweencomputer and robot, so it almost uses the maximum number of bytes in one packet (Andre Ryll,2016), which may increase the number of lost packets on the network (Nadarajah & Sundaraj,2012).

Page 27: Lucas Henrique Cavalcanti Santostg/2019-2/TG_EC/tg-lhcs-final.pdf · IoT introduz conectividade nos dispositvos e hoje já tem expecitativa de vender 214 bilhões de dólares em 2020.

262626

3SYSTEM ARCHITECTURE

In this Chapter, the proposed architecture is presented. It consists in an optimized wirelessnetwork architecture that controls multiples mobile robots and monitors them, as illustrated byFigure 11. It starts with commands from a computer, that generally is called as the commanderand has the cognition. The commander’s code sends messages, defined by the cognition, thatincludes identifications and actions, to each robot at a time. This information leaves the computerand needs to reach the robots, so the data needs to reach through wireless media. For that, thecommander connects itself to a base station, which is an embedded system that receives thecommander’s message and transmits it using a wireless module. Each robot also needs a wirelessmodule, responsible for receiving packets sent from the base station. Finally, with the message,the robots interpret the data and execute the command sent by the computer.

Figure 11: Project network overview with a computer that sends commands to a base stationwhich transmits the message wirelessly to multiple robots.

The opposite way of communication is needed to create the telemetry ability. Althoughthe robot can use the modules of the control network to send telemetry information, an additionaltransceiver is added, creating then a second channel of data exchange that not interfere with theflow of control messages.

Page 28: Lucas Henrique Cavalcanti Santostg/2019-2/TG_EC/tg-lhcs-final.pdf · IoT introduz conectividade nos dispositvos e hoje já tem expecitativa de vender 214 bilhões de dólares em 2020.

272727

Around the message flow, there are two communication points. The first betweenthe computer and the base station, and the latter between the base station and robots. Bothconnections need common patterns to exchange data; in other words, the system needs the samelanguage of communication.

3.1 COMMUNICATION PACKETS AND PROTOCOL

The first protocol, defined as the communication between two transceivers, communicateswirelessly. It was essential to define a communication topology that minimizes the configurationoverhead, not depending on messages order, neither fails due to lost packets. When consideringthe nature of the control and telemetry messages, there is no need to recover them, because theyrequire only the current information. So, the communication between the base station and robotsis entirely broadcast; in other words, the messages transmitted by the base station are received inall robots, as shown in Figure 11.

The broadcasted topology of messages removes the overhead in dynamic changing thereceiver address at wireless modules. Using broadcast, the system not requires acknowledgmentpackets from receivers, which permits disabling the re-transmission of lost packets. On the otherhand, the message should include a robot identification, to enable the option of addressing eachrobot independently.

Table 3: Message protocol created to controlRobôCIn’s robots.

Bits Offset Bits Size Information

0-3 4 Message Type

4-7 4 Robot Id

8 - 27 20 vx - Linear Speed

28 - 47 20 vy - Linear Speed

48 - 67 20 ω - Angular Speed

68 - 87 20 θ - Robot Angle

88 1 Kick Front

89 1 Kick Chip

90 1 Charge the Kick

91 - 98 8 Strength of the Kick

99 1 Turn on the Dribbler

100 - 107 8 Speed of the Dribbler

108 - 111 4 Additional Command

Moreover, after choosing the robot to control, the commander needs to send desiredmovements and peripheral functions, so the project needs a control protocol. The developed

Page 29: Lucas Henrique Cavalcanti Santostg/2019-2/TG_EC/tg-lhcs-final.pdf · IoT introduz conectividade nos dispositvos e hoje já tem expecitativa de vender 214 bilhões de dólares em 2020.

282828

protocol aims to minimize the payload size while enabling accurate control of the robots. So,the bits offset and message interpretation works according to Table 3 description. It includes, inthe first byte, the message identification, which defines the interpretation of the message bits.At the first byte, there are also bits that identifies which robot the commander wants to control.These two initial slots, of four bits each, permit the robots to act only when the commanderaddresses the message to it. In other words, even receiving all broadcasted messages, the robotsidentify when they need to perform some action. Finally, the last thirteen bytes define the robot’smovements and its peripherals actions, like the kicker that allows the robot to kick the ball.

After controlling the robots, the network should enable the ability to monitors them.Then, another messaging protocol is created, now described at Table 4. It also uses the minimumbits required to represent the robot’s information. Equally to the control protocol, the telemetrymessage has the message and robot identification in one byte, to inform the commander whichmessage arrived and from which robot it came. Differently from the control protocol, thetelemetry protocol has each motor speed, measured by the robot’s sensors, and that informationassures the commander the correct behavior of the motors. Alongside with motors speed, thetelemetry packet includes essential information measured at the robot, like battery level andmore, which enables the identification of malfunction in robots.

Table 4: Telemetry protocol created to receiveRobôCIn’s robots’ information.

Bits Offset Bits Size Information0-3 4 Message Type4-7 4 Robot Id

8 - 23 16 m1 - Motor 1 Speed24 - 39 16 m2 - Motor 2 Speed40 - 55 16 m3 - Motor 3 Speed56 - 71 16 m4 - Motor 4 Speed72 - 86 15 Dribbler’s Motor Speed87 - 94 8 Kick’s Capacitor Load

95 1 Ball on the Robot96 - 103 8 Robot’s Battery

After defining the communication between transceivers and the protocol to control andmonitor robots, the communication between the computer and base station is defined. Althoughit uses wired communication, there is data flowing, and its correctness is essential to transmittingand receiving messages. Independently of the interface used, the base station does not need toencode or decode messages of the control and telemetry protocols. It only needs to transmitmessages from computer to robots, and from robots to computer. Furthermore, there is nonecessity for defining a protocol, as the information exchanged between the computer and basestation only needs the size of the packet to receive and send the right message. The definedprotocols of messages makes the commander sends control messages of 14 bytes, while receives

Page 30: Lucas Henrique Cavalcanti Santostg/2019-2/TG_EC/tg-lhcs-final.pdf · IoT introduz conectividade nos dispositvos e hoje já tem expecitativa de vender 214 bilhões de dólares em 2020.

292929

telemetry messages of 13 bytes from the base station. On the other hand, the base station sends13 bytes to the commander and receives 14 bytes from it.

3.2 SYSTEM IMPLEMENTATION

The network implementation started by choosing the best modules that fit the projectrequirements and budget. The RobôCIn’s robot, shown at Figure 12, already has two nRF24l01(Nordic, 2008) transceivers modules and uses a NUCLEO-F767ZI (ST, 2019a) board pro-grammed with mbed Operating System (mbedOS) (Arm, 2019), which is a real-time operatingsystem for ARM microcontrollers that allows the development of C++ codes, gives support forARM peripherals, and emulates virtual thread.

Figure 12: RobôCIn’s Small Size Soccer (SSL) robot.

The mbedOS comes with built-ins libraries that implement different interfaces of commu-nication while controlling the board peripherals. So, the SPI between the board and transceiversused the read and write functions implemented in the mbedOS. On the other hand, the transceiversonly accept and return messages if the program communicates with the correct register’s address.

The chosen module, the nRF24l01 transceiver, has specific registers for configuring theradio (address, power, and function), sending and receiving bytes. All of them are presentedin the module datasheet (Nordic, 2008). After configuring the nRF24l0, with the specificationshown in Table 5, the radio sends and receives data with Enhanced Shock Burst (Nordic, 2008),a proprietary technology. That technology allows the exchange of at most 32 bytes of payloadand includes a CRC of bits, calculated inside the transceiver module.

Page 31: Lucas Henrique Cavalcanti Santostg/2019-2/TG_EC/tg-lhcs-final.pdf · IoT introduz conectividade nos dispositvos e hoje já tem expecitativa de vender 214 bilhões de dólares em 2020.

303030

Table 5: Project Modules and Configuration in Base Station and Robots

Base Station Robots

Transceiver nRF24l01 nRF24l01

Embedded Board Nucleo F767ZI Nucleo H743ZI

CPU Frequency 400 MHz 216 MHz

Operating System mbedOS mbedOS

Radio Frequency (Control) 2504 MHz 2504 MHz

Radio Frequency (Telemetry) 2529 MHz 2529 MHz

Radio Address 0x753FAD299ALL 0x753FBD299ALL

Together with the RobôCIn’s robots and its software, this work proposes an embeddedsystem for the base station. The Figure 13 shows the base station architecture and interfaces.There are two nRF24l01 modules coupled by SPI to a NUCLEO-H743ZI (ST, 2019b) pro-grammed by mbedOS, using the SPI library. The base station takes into consideration thehardware and software used in robots to maintain the same operating system while adding afaster microcontroller board, as shown by its Control Processing Unit (CPU) frequency at Table 5.The choice of using a faster board is because even the robots receiving all broadcasted message,there is only one base station for multiple robots, which creates a bottleneck of messages.

Figure 13: Base station architecture of the proposed wireless network with Serial and Ethernetoptions between the Computer and Base Station.

The control communication from the computer to the robot uses one transceiver at thebase station and another at the robot, as shown in Figure 14. In the base station, the transceiveris configured as a sender radio that targets the message to the address of the robots, shown inTable 5. On the other hand, the transceiver configuration in the robots uses the receiver modewith robot’s network address.

Finally, the computer, base station, and the robot share information with the protocolsmade. The base station and robots exchange bytes with nRF24l01 transceivers using the

Page 32: Lucas Henrique Cavalcanti Santostg/2019-2/TG_EC/tg-lhcs-final.pdf · IoT introduz conectividade nos dispositvos e hoje já tem expecitativa de vender 214 bilhões de dólares em 2020.

313131

configuration presented, but between the computer and station, this work brings two optionsof the interface: Serial and Ethernet (Figure 15). Moreover, both interfaces have support atthe NUCLEO board and have built-in libraries in the mbedOS, allowing the transmission andreception of bytes between embedded boards and computers.

Figure 14: Diagram of radios used for control and telemetry network on the Base Station andRobot, together with its way of communication.

Due to the characteristics of the Serial interface, it is easy to implement, but it can onlyreceive and send one byte at a time. Furthermore, to control the robots, was developed a basestation that read each byte until they compose a message. Only with the message complete, thebase station transmits packets to the nRF24l01. This concept is shown by Algorithm 1.

Algorithm 1: Transmission Concept for Base Station Serial Interface Approach1 Initialize byte array and counter;2 while true do3 //Read one byte at Serial;4 byte array[counter] = read serial();5 counter++;6 if counter == message size then7 //Send message to transceiver;8 send bytes (byte array, message size);9 clear (byte array);

10 counter = 0;

The second interface, the Ethernet, is used with UDP protocol present in the chosenembedded board, by the mbedOS support, and present at every personal computer. The UDPprotocol is suitable for streaming data, as the control network. Moreover, different from theSerial approach, the Ethernet sends a group of bytes containing the sent message and also aheader, as described in Chapter 2. In other words, the developed algorithm only waits for amessage and re-transmits its bytes by the payload of the nRF24l01, as shows in Algorithm 2.

Page 33: Lucas Henrique Cavalcanti Santostg/2019-2/TG_EC/tg-lhcs-final.pdf · IoT introduz conectividade nos dispositvos e hoje já tem expecitativa de vender 214 bilhões de dólares em 2020.

323232

Algorithm 2: Transmission Concept for Base Station Ethernet Interface Ap-proach

1 Initialize byte array;2 while true do3 //Receive complete message and send.;4 byte array = read packet();5 send bytes(byte array);6 clear (byte array);

Finally, the telemetry network to monitors the robot was developed using an additionaltransceiver present in the robot and the base station, as illustrated in Figure 14. First, the robotcode was modified to measure robot’s status on top of RobôCIn’s robot code (RobôCIn, 2019).In the code, the telemetry transceiver is configured in sender mode, but different from the controlnetwork it uses the telemetry network frequency and addresses the base station (Table 5). Withthe robot status, the code encodes it, following the telemetry protocol, and sends through thetelemetry transceiver to the base station.

At the base station, a virtual thread, parallel to the control code, was developed, and itconfigures the second transceiver of the base station as a receiver, expecting messages at thetelemetry frequency. So, whenever a wireless message reaches the telemetry transceiver at thebase station, it forwards the payload bytes to the computer, where the message is appropriatelydecoded and identified.

Figure 15: Proposed Base Station hardware.

Page 34: Lucas Henrique Cavalcanti Santostg/2019-2/TG_EC/tg-lhcs-final.pdf · IoT introduz conectividade nos dispositvos e hoje já tem expecitativa de vender 214 bilhões de dólares em 2020.

333333

3.3 TIME ANALYSES

3.3.1 Test Method

After implementing the network configurations, the system does not have a definedperiod between sending each message. That interval is essential to the network operation and itis different for the Serial base station, for the Ethernet one, and even for the telemetry messages.Then, this work proposes a method for testing the delivery delay of packets at the robots, whereis the main requirement of the network.

The proposed method tests the delivery time, independent of the interface, and alsoevaluates the telemetry impact on the network. It is robust to analyze delay performance,because, no matter the parameters, it measures the interval between messages that should arrivecontinuously at the robots.

The interval between messages delivery varies a lot because the network broadcasttopology works asynchronously, and sometimes wireless packets are lost. So, the delivery timebetween messages in one configuration means the average interval between each of the 500messages received. Moreover, tests performed with a slow network or losing packets increasethe average interval time, calculated with

timeinterval =∑

500i=1 (messageTimei −messageTimei−1)

500

� �3.1

and, because of the varying time interval between messages, a not optimized networkalso increases the standard deviation, calculated by

sinterval =

√√√√ 1499

500

∑i=1

(messageTimei −messageTimei−1)� �3.2

The test flow described at Figure 16 begins with the commander sending messages to aSerial or Ethernet base station, with manually configured interval. At the base station, it maintainsits code and configuration. On the other hand, the robot has an additional code that measures theinterval between the reception of 500 messages. Then, the robot sends the measured intervals toan additional computer through USB connection encoding the data in Comma-Separated Values(CSV) file. At that computer, the data is plotted and analyzed. After manually analyzing thetime performance of the test, another configuration or experimental intervals is configured at thecommander to realize another test.

Page 35: Lucas Henrique Cavalcanti Santostg/2019-2/TG_EC/tg-lhcs-final.pdf · IoT introduz conectividade nos dispositvos e hoje já tem expecitativa de vender 214 bilhões de dólares em 2020.

343434

Figure 16: Architecture of the method developed for testing the network configurations andinterfaces.

Although the test of telemetry impact on the network uses the same flow, the telemetrysending interval is configured at the robots and not in the commander. So, the same analyseswere performed to assure the efficiency of the control network, even monitoring the robots.

3.3.2 Control Message Interval Time

To choose the interval in which the commander should send messages, at the Serial andEthernet base station, the control message interval time was experimentally found. For eachinterface, multiples intervals are tested and compared in order to define an optimal interval.

When we search for the optimized interface, the computer and base station code have asignificant effect on the tests. As described in Chapter 3, the base station uses the Algorithm 1for the Serial communication approach with the maximum Serial speed supported (115200bits per second), and Algorithm 2 for Ethernet approach. The interval that the computer sendsmessages means the throughput that each base station supports. Moreover, that interval shouldexist because the computer processor is a way faster than the embedded system.

An interval in microseconds is defined based on RobôCIn’s cognitive software processingtime (few milliseconds) and according to what each base station approach supports. This processtries to find the minimum interval between messages which each base station algorithm supports.Finally, to find the maximum throughput, the interval time was decreased until the robot correctlyreceives the messages. Then, a larger interval was tested in order to search for the optimal one.

Analysing the results in Figure 17, the network with base station Serial approach has thebest throughput with the 1900us of interval. However, the minimum interval at Figure 17 was

Page 36: Lucas Henrique Cavalcanti Santostg/2019-2/TG_EC/tg-lhcs-final.pdf · IoT introduz conectividade nos dispositvos e hoje já tem expecitativa de vender 214 bilhões de dólares em 2020.

353535

1850us, that throughput made the base station lost some bits when reading the commander, whichcaused undesired behaviors on the robots. With a smaller interval, like 1800us, the messagesdo not even reach the robot. In the end, the optimal result is with 1900us of interval, that alsohas a small amplitude of the box-plot limits, meaning that the message delivery time was moreconstant than in other intervals.

Figure 17: Reception delay for 30 tests at each different sending interval, using a Serial basestation transmitting computer messages at the configured sending intervals.

At Figure 18 is shown that the optimum interval time between messages in the networkwith Ethernet interface is 500us. Almost four times smaller than the Serial, the Ethernet approachdoes not corrupt the bits with a shorter interval time than 500us. However, as shown at Figure 18,smaller intervals increases and vary the delivery time at the robot.

Page 37: Lucas Henrique Cavalcanti Santostg/2019-2/TG_EC/tg-lhcs-final.pdf · IoT introduz conectividade nos dispositvos e hoje já tem expecitativa de vender 214 bilhões de dólares em 2020.

363636

Figure 18: Reception delay for 30 tests at each different sending interval, using a Ethernet basestation transmitting computer messages at the configured sending intervals.

The analysis presented previously found different intervals of time for each communica-tion interface — an optimal interval of 1900us for Serial and 500us for Ethernet. So, Figure 19presents the delivery time at 25 tests with each communication interface and its optimal interval.There, one test is represented by one point in the figure, and it is the average reception betweeneach one of 500 messages received together with its standard deviation. However, the Serialapproach is three times slower than Ethernet; each approach has its test interval similar, with asimilar standard deviation, which means a uniform network.

(a) Serial station with 1900us of interval (b) Ethernet station with 500us of interval

Figure 19: Delivery time of base station approaches with the minimum sending interval.

Page 38: Lucas Henrique Cavalcanti Santostg/2019-2/TG_EC/tg-lhcs-final.pdf · IoT introduz conectividade nos dispositvos e hoje já tem expecitativa de vender 214 bilhões de dólares em 2020.

373737

3.3.3 Telemetry Message Interval Time

Finally, with the optimal interval for each base station approach, the telemetry is devel-oped on the network in order to the commander receive robots’ feedback. However, the intervalfor the control network was found; no interval between telemetry packets sent from the robotis known. Again, the tests are applied to find an optimal interval for the telemetry network,and they analyze the delivery time of control packets, as it is essential to move robots evenwith telemetry enabled. In order words, the robot should send telemetry packets, and the testscontinue as described at Figure 16, but changing the telemetry interval in robots code.

At the initial tests with the Serial interface, the bits exchanged between the commanderand base station are usually corrupted, which made it impossible to use Serial to send andreceive a message between the base station and the computer. Then, the Ethernet approach is theinterface that supports duplex communication without affecting the throughput or corruptingmessages. In other words, monitoring the robots was only possible on the Ethernet base station.

Similar to the computer sending interval time, the robot should have an interval betweeneach telemetry packet sent. Furthermore, the tests initially used the interval of 200ms betweentelemetry messages because of the RobôCIn’s requirement. With the baseline of 200 millisecondsor five messages per second, the tests decreased the telemetry interval to find a balance thatguarantees quickly sampling without damaging the control network.

Table 6: Ethernet Base Station with Telemetry Different Sampling Interval

Test Condition Delivery Time Time Increase

Average Standard Deviation Average Standard Deviation

Without Telemetry 721.98ms 140.12

200ms Sampling 724.78ms 159.44 0.39% 13.79%

50ms Sampling 730.89ms 196.07 1.23% 39.93%

10ms Sampling 768.80ms 217.69 6.48% 55.36%

Table 6 shows the network time performance of an Ethernet base station, with a sendinginterval of 500 microseconds. It compares the delivery time of a network using different telemetryintervals, with a network without telemetry. So, Table 6 presents an impact of 0.39% with a200ms telemetry interval; at 50ms of sampling, the receive interval increases 1.23%. Additionally,the standard deviation without telemetry compared to 200ms and 50ms of telemetry intervalshows that telemetry increases the variation of messages delivery interval. Furthermore, the10ms sampling increases 6.48% of average delivery time, together with an increase of more than50% at the standard deviation, which may affect the control network performance.

Finally, the Ethernet base station receiving telemetry packets at every 50 millisecondsguarantees a control update in every 730.89 milliseconds.

Page 39: Lucas Henrique Cavalcanti Santostg/2019-2/TG_EC/tg-lhcs-final.pdf · IoT introduz conectividade nos dispositvos e hoje já tem expecitativa de vender 214 bilhões de dólares em 2020.

383838

4RESULTS

This Chapter presents the results from the performance tests applied to the network.Described in Chapter 3, the test methodology aims to analyze network time performance at thechosen configurations. To achieve minimum delivery time at robots, no matter the approach, theperformance was measured in the robots using a network of six robots. Therefore, tests wererealized in the same environment, hardware, and when not the goal of the test, the same software.

The network efficiency is tested changing the number of robots, its distances to the basestation, and enabling and disabling the telemetry. These tests were designed and performed inorder to assure the robustness and scalability of the proposed network for RobôCIn’s soccerrobots.

4.1 DIFFERENT DISTANCES

An important characteristic of a network that controls soccer robots is a short deliverytime with different distances. As the robot moves around the field, it may move next to the basestation transceiver or not. The tests that simulate different distances needed to repeat the testsusing both base station approaches and positioning the RobôCIn’s robot at different parts of thefield. First, the robot was next to the transceiver, 0.4m of distance, after it was at midfield, 2.5mof distance, and finally at the opposite side of the field, 5m of distance.

Page 40: Lucas Henrique Cavalcanti Santostg/2019-2/TG_EC/tg-lhcs-final.pdf · IoT introduz conectividade nos dispositvos e hoje já tem expecitativa de vender 214 bilhões de dólares em 2020.

393939

(a) Delivery time of Serial base station differentdistances

(b) Delivery time of Ethernet base station differentdistances

Figure 20: Delivery time of base station with different distance between transceivers.

Figure 20 shows that the only difference in delivery time is the interface used betweenthe computer and the base station. So the delay in reaching farther distances does not increase thedelivery time. In other words, the time to navigate to a farther point is insignificant. Furthermore,the small standard deviation at the box-plot shows that the number of lost packets was alsoinsignificant.

4.2 MULTIPLE ROBOTS

The robot soccer game that RobôCIn plays have 12 robots, six from each team. So,the network performance test analyzed multiple robots to evaluate the delivery time and itsconsistency and equality between robots. Using the optimal interval discovered for Serial(1900us) and Ethernet (500us) interfaces, and each robot is placed at 2.5m away from the basestation.

Figure 21 shows that the delivery time in a network with six robots increased aroundsix times compared with one robot network. This outcome is expected because the number ofrobots increased six times, and, even with the broadcast topology, each message address only onerobot. In other words, the base station sends six messages to control all robots as one messageonly controls one robot. That outcome does not change with different communication interfacesbetween computer and base station because it is intrinsic of network topology. Figure 21(b) alsoshows that the delivery delay at one robot increases together with the number of robots in thenetwork.

Page 41: Lucas Henrique Cavalcanti Santostg/2019-2/TG_EC/tg-lhcs-final.pdf · IoT introduz conectividade nos dispositvos e hoje já tem expecitativa de vender 214 bilhões de dólares em 2020.

404040

(a) Delivery time of six robots network (b) Delivery time with different number of nodes atEthernet base station

Figure 21: Delivery time analysis between interfaces (a) and network size (b).

Therefore, to test consistency in the delivery time capture at two different robots sepa-rately, two different tests were realized in the Ethernet base station network. The first one wasanalyzing the delay at two different robots in a two robots’ network, and the other one at thesame base station and robots but with six robots network. The result of both tests, presented atFigure 22, confirms the expectation that a network with two robots (Figure 22(a)) has a deliverytime three times shorter than the six robot network. Similar result presented at Figure 21.

(a) Delivery time of Ethernet base station at tworobots network

(b) Delivery time of 6 robots network with Ethernetbase station at two nodes

Figure 22: Comparing delivery time with two and six nodes network at two different robots.

Although the delivery time increased with more robots in the network, both networks,when comparing Robot 1 and Robot 2 delivery times, shown at Figure 22(b), are similar at bothnodes. That result proves that the proposed broadcast protocol works equally for every robot,and even though the robots receive all sent packets, it correctly filters messages by the robot id.

Page 42: Lucas Henrique Cavalcanti Santostg/2019-2/TG_EC/tg-lhcs-final.pdf · IoT introduz conectividade nos dispositvos e hoje já tem expecitativa de vender 214 bilhões de dólares em 2020.

414141

4.3 TELEMETRY

Analyzing the control network with telemetry activated for the whole team, six robots,was crucial to confirm the ability to use telemetry at the games. The results, shown at Figure 23,had a higher gap between the delivery delay at a network with and without telemetry whencompared with the tests realized with one robot, shown at Table 6. The increased gap shows adelay coming from the base station higher load of messages because now it needs to transmit sixpackets of control and six of telemetry.

Figure 23: Delivery time of Ethernet base station packets with different telemetry sampling timeat 6 robots control network.

Although, the delivery time of the control network increased from 4308us, when it iswithout telemetry, to 4672us, when it is with a 50ms sampling telemetry in each of the six robots.Now, with telemetry, it is possible to monitor the robot’s status, like the angular speed of eachwheel on time, represented at Figure 24.

Figure 24: Angular speed on time of each wheel is the robot, monitored by the developedtelemetry network.

Page 43: Lucas Henrique Cavalcanti Santostg/2019-2/TG_EC/tg-lhcs-final.pdf · IoT introduz conectividade nos dispositvos e hoje já tem expecitativa de vender 214 bilhões de dólares em 2020.

424242

5CONCLUSION

In the context of houses, Industry 4.0, Internet of Things (IoT), and robotics, wirelesscommunication is essential. It can share information across devices, give functionalities, or,like the soccer robots at SSL, may control mobile robots movements. Therefore, technologiesand systems should be developed and integrated to fill the necessity present at each applicationdeveloped. Although there are no perfect modules or technologies, there are options to find abalance that maximizes the desired efficiency at each application.

At robotics systems, the challenge is higher. Besides the perception and cognition, thereare tasks like wireless communication, which is fundamental to accomplish real-time motioncontrol at the robots. At communication wireless, this work proposed architecture of the basestation that efficiently controls the RobôCIn’s SSL robots. The work reviewed modules andinterfaces to reach a system architecture that, when analyzed in a real robotics soccer environment,fulfill the requirements.

The results show a system that reached 4.33 milliseconds of latency at robot when it ison a six robot network. Only the Ethernet base station proposed reaches that latency, and it isalmost three times faster than the serial base station. What endorsed the benefits of using a fasterand updated interface of communication, even between the computer and base station.

Even though the number of robots in the network increases the latency proportionallybecause the station should divide its throughput to more robots, continuously emitting messagesat one, two, and six robots’ networks, showed consistent message reception and consistentlatency. Even when measuring in two different robots and also at different distances from thebase station, the proposed network delivered at least 230 messages per second, faster enough toSSL competition, that usually uses 60 fps cameras.

When developed the telemetry at the robot and base station, the delivery time of thepackets in the six robots’ network, with 50ms of telemetry sampling enabled, reached 4.39milliseconds. This latency permits the delivery of 227 messages per second per robot. Moreover,although 4.39 milliseconds represents an increased of 1,39% compared with the network withouttelemetry, the telemetry network brings the benefit of monitoring each robot and should avoidfailures and increase precision.

Results reliability working with wireless networks is trick, because many phenomena

Page 44: Lucas Henrique Cavalcanti Santostg/2019-2/TG_EC/tg-lhcs-final.pdf · IoT introduz conectividade nos dispositvos e hoje já tem expecitativa de vender 214 bilhões de dólares em 2020.

434343

may occur and noise the results. So, every test made measured 30 receptions intervals, and eachinterval is the average time between the delivery of 500 packets. This amount of measures in areal environment statistically endorse the real-time network efficiency, which also has telemetryability at each robot.

With the results shown at this work, the RobôCIn team already used the proposed networkat the Latin American Robotics Competition (LARC) - 2019. Without any failure or delayedcontrol, it was a pleasure to see the team using the proposed network to control and monitor itsrobots. At the competition, RobôCIn did not play any match with malfunction robots due totelemetry. The robot status, received at the computer, enabled the team prevent robot’s failuresbefore and during the game. This work also brought to RobôCIn a quickly maintenance andconsistency in the championship.

After the RobôCIn’s experience, the proposed network proved its ability to supportmultiple devices in real-time and is suitable for IoT and Industry 4.0 applications.

For future work, a study of new modules and its technologies should be considered tospeed up the network data exchange. Moreover, future work may find alternatives to reducethe delivery time impact when adding more robots to the network. Changes like the embeddedboards or different interfaces may increase the network cost but may reduce the delivery of thewhole system. Together with bets specifications, future work may develop an automatic timeanalyzes that search the optimal parameters using the hardware in the loop.

Page 45: Lucas Henrique Cavalcanti Santostg/2019-2/TG_EC/tg-lhcs-final.pdf · IoT introduz conectividade nos dispositvos e hoje já tem expecitativa de vender 214 bilhões de dólares em 2020.

444444

REFERENCES

Andre Ryll, Mark Geiger, N. O. A. S. L. M. (2016). Extended team description for robocup2016. [Online; accessed 01-September-2019].

Arm (2019). mbed operating system - open-source rtos. https://os.mbed.com/. [Online;accessed 05-February-2019].

BankMyCell (2019). How many phones are in the world? BankMyCell Blog.

Briscoe, N. (2000). Understanding the osi 7-layer model. PC Network Advisor, 120(2).

Brogårdh, T. (2007). Present and future robot control development—an industrial perspective.Annual Reviews in Control, 31:69–79.

Hazen, M. E. (2003). Understanding some basic recommended standards for se-rial data communication, a comparison of rs-232, rs-422 and rs-485. http://kando.prociweb.hu/letoltes/data/3.evfolyam/MAI/F2/Integralt_felugyeleti_rendszerek_kialakitasa/RS232_485/wp0585.pdf.

IC Insights (2019). Semiconductor unit shipments exceeded 1 trillion devices in 2018. Technicalreport, IC Insights.

IEEE (2007). Ieee standard for information technology - telecommunications and informationexchange between systems - local and metropolitan area networks - specific requirements - part11: Wireless lan medium access control (mac) and physical layer (phy) specifications. IEEE Std802.11-2007 (Revision of IEEE Std 802.11-1999), 1–1076.

IFR (2018). Executive summary world robotics 2016 industrial robots.

Kang, J. S., Kim, Y. G., & Park, H. S. (2013). Remote data transmission middleware fortelerobotics. In 2013 International Conference on IT Convergence and Security (ICITCS), 1–4.

Kim, J., Kim, D., Kim, Y., & Seow, K. (2014). Soccer Robotics. Springer Tracts in AdvancedRobotics. Springer Berlin Heidelberg.

Kitano, H., Asada, M., Kuniyoshi, Y., Noda, I., & Osawa, E. (1997). Robocup: The robot worldcup initiative. In Proceedings of the First International Conference on Autonomous Agents,340–347.

Kordas, J., Wagner, P., & Kotzian, J. (2010). Wireless transceiver for control of mobile embed-ded devices. In Proceedings of the International Multiconference on Computer Science andInformation Technology, 869–872.

Leens, F. (2009). An introduction to i2c and spi protocols. IEEE Instrumentation MeasurementMagazine, 12(1):8–13.

Liu Jianbang, Lai Xuzhi, Wu Min, & Chen Xin (2008). Design of embedded telerobotics system.In 2008 27th Chinese Control Conference, 229–233.

Nadarajah, S. & Sundaraj, K. (2012). Wireless communication in robot soccer: A case study ofexisting technologies. In 2012 IEEE Conference on Sustainable Utilization and Development inEngineering and Technology (STUDENT), 33–38.

Page 46: Lucas Henrique Cavalcanti Santostg/2019-2/TG_EC/tg-lhcs-final.pdf · IoT introduz conectividade nos dispositvos e hoje já tem expecitativa de vender 214 bilhões de dólares em 2020.

454545

Nordic (2008). nrf24l01+ single chip 2.4ghz transceiver. https://www.sparkfun.com/datasheets/Components/SMD/nRF24L01Pluss_Preliminary_Product_Specification_v1_0.pdf. [Online; accessed 15-March-2019].

Proskochylo, A., Vorobyov, A., Zriakhov, M., Kravchuk, A., Akulynichev, A., & Lukin, V.(2015). Overview of wireless technologies for organizing sensor networks. In 2015 SecondInternational Scientific-Practical Conference Problems of Infocommunications Science andTechnology (PIC S T), 39–41.

Radiant Insights (2015). Embedded system market size, share, trends, report, 2020. Technicalreport, Radiant Insights.

RoboCup (2018). Laws of the RoboCup Small Size League 2019. Small Size League TechnicalCommittee.

RobôCIn (2019). Small size league soccer robot’s code. https://github.com/robocin/ssl-embedded. [Online; accessed 18-July-2019].

Saha, H., Mandal, S., Mitra, S., Banerjee, S., & Saha, U. (2017). Comparative performanceanalysis between nrf24l01+ and xbee zb module based wireless ad-hoc networks. InternationalJournal of Computer Network and Information Security, 9:36–44.

Santitoro, R. (2003). Metro ethernet services - a technical overview. Metro Ethernet Forum.

Sanz, P. (2009). Robotics: Modeling, planning, and control (siciliano, b. et al; 2009) [on theshelf]. Robotics & Automation Magazine, IEEE, 16:101–101.

Siegwart, R., Nourbakhsh, I. R., & Scaramuzza, D. (2011a). Autonomous mobile robots. ABradford Book.

Siegwart, R., Nourbakhsh, I. R., & Scaramuzza, D. (2011b). Introduction to autonomous mobilerobots. MIT press.

SparkFun (2010). Rs-232 vs. ttl serial communication. https://www.sparkfun.com/tutorials/215. [Online; accessed 05-September-2019].

SparkFun (2014). Serial communication. https://learn.sparkfun.com/tutorials/serial-communication. [Online; accessed 10-August-2019].

ST (2019a). Nucleo-f767zi. https://www.st.com/en/evaluation-tools/nucleo-f767zi.html. [Online; accessed 01-September-2019].

ST (2019b). Nucleo-h743zi. https://www.st.com/en/evaluation-tools/nucleo-h743zi.html. [Online; accessed 12-September-2019].

Wang, Y., Hu, C., Feng, Z., & Ren, Y. (2014). Wireless transmission module comparison. In2014 IEEE International Conference on Information and Automation (ICIA), 902–907.

Wang, Y., Ma, H.-S., Yang, J.-H., & Wang, K.-S. (2017). Industry 4.0: a way from masscustomization to mass personalization production. Advances in Manufacturing, 5.

WEITZENFELD, A.; BISWAS, J. A. M. S. K. (2015). Robocup small-size league: Past, presentand future.

Page 47: Lucas Henrique Cavalcanti Santostg/2019-2/TG_EC/tg-lhcs-final.pdf · IoT introduz conectividade nos dispositvos e hoje já tem expecitativa de vender 214 bilhões de dólares em 2020.

464646

Cerkala, J. & Jadlovska, A. (2015). Nonholonomic mobile robot with differential chassismathematical modelling and implementation in simulink with friction in dynamics. ActaElectrotechnica et Informatica, 15:3–8.


Recommended