京都大学 / 新制・課程博士 / 博士(工学) / 甲第24606号 / 工博第5112号 / 新制||工||1978(附属図書館) / 京都大学大学院工学研究科機械理工学専攻 / (主査)教授 松野 文俊, 教授 松原 厚, 教授 泉井 一浩 / 学位規則第4条第1項該当 / Doctor of Philosophy (Engineering) / Kyoto University / DGAM

Today’s society is facing a large increase of automation and smart devices. Everything from coffee machines to fridges include some kind of electronics and embedded systems. The focus of this Bachelor’s thesis was to dive deeper into how these automated devices can be controlled and more specifically a robot arm. The main purpose revolved around constructing a robotic arm that could be controlled through three different methods using MATLAB. These three were manual control, numerical analysis control and with a neural network based control. The prototype was created by assembling six servo motors onto 3D-printed parts. The arm consisted of three main parts which were a base, an arm and a gripper. The system was controlled by an Arduino micro-controller connected to a computer. The results show that the manual control method was easy to implement, fast and reliable. It allows control of all the angels for each servo motor, which also means controlling each individual degree of freedom. The numerical way, using Newton-Raphson’s method, broadened the abilities to control the arm but was slower. The third and final solution was to use fuzzy-logic. This ended up being a powerful method allowing for great control with low latency. While unreliable, the method showed great potential and with refinement could surpass the others. The conclusion was that the neural network method was the overall best method for controlling and manoeuvring the robot arm using MATLAB. / Dagens samhälle står inför en stor ökning av automatisering och smarta produkter. Allt från kaffemaskiner till kyl och frys innehåller någon form av elektronik och inbäddade system. Det huvudsakliga syftet med detta kanditatexamensarbete var att gräva djupare i hur dessa automatiserade produkter kan kontrolleras och mer specifikt i detta fall, en robotarm. Projektet handlade om att konstruera en robotarm som kunde styras och kontrolleras genom tre olika metoder i programmet MATLAB. Dessa tre har vi valt att kalla manuell kontroll, numerisk kontroll och neuralt nätverksbaserad kontroll. Prototypen tillverkades genom att montera sexservomotorer på 3D-utskrivna delar. Armen bestod av tre huvuddelar, en bas, en arm och en gripklo. Systemet styrdes av en Arduino mikrokontroll ansluten till en dator. Resultaten visar att den manuella kontrollmetoden var enkel att implementera, snabb samt var tillförlitlig. Den gav precis styrning av alla vinklar för varje servomotor, vilket också innebar att den gav god styrning av varje frihetsgrad. Den numeriska metoden, mer bestämt Newton Raphson’s metod, vidgade möjligheterna att kontrollera armen men var långsammare. Den tredje och sista lösningen var att använda ett neuralt nätverk, fuzzy logic. Detta visade sig vara ett kraftfullt sätt att styra roboten med låg latens. Det neurala nätverket visade sig dock vara opålitligt, men metoden visade stor potential för vidare utveckling och kan då prestera mycket bättre än de andra två metoderna. Slutsatsen var att det neurala nätverket var den generellt bästa metoden för att kontrollera och manövrera robotarmen via programmeringsprogrammet MATLAB.

Yes / A fuzzy neural Petri Nets (FNPN) controller is utilized for controlling a three-links robot arm which considers a nonlinear dynamic system. The incorporation of the classical FNN with a Petri net (PN) has been suggested to produce a new representing system called FNPN structure to alleviate the computation burden. The motion equation of three links robot arm is derived from Lagrange’s equation. This equation has been incorporated with the motion equations of DC Servo motors which motivate the robot. For nonlinearity dynamic problems, this paper presents a direct adaptive control technique to control three links robot arm utilizing the FNPN controller. The computer simulation depicts that the present FNPN controller accomplished better performance with fast response and minimum error.

Research and development leading to the realisation of a fully autonomous and robust multi-fingered hand has been going on for three decades. Yet none can be found in an industrial application. This is largely because we do not fully understand the fundamental mechanics of multi-finger grasping. / This thesis is a study of the mechanics of multi-finger grasping, with particular attention being paid to applying the analysis to experimental co-operative motion tasks between a hand-arm system and grasped object. / Fine manipulation with multi-fingered robot hands is critically influenced by the capacity to achieve stable grasps. By exploring the fundamental mechanics involved, a method for establishing the stability of spatial four finger-contact grasps is obtained. This work examines both frictionless and frictional grasps in two and three dimensions and develops the stability requirements for grasping. The conditions for a stable grasp are expressed as simple equations relating the line coordinates of (i) transitory sliding actuator and (ii) the normal to the tangent plane at every contact location. This is achieved by using the principle of virtual work and a branch of statics known as astatics. / After specifying a grasp in terms of its contact locations and forces the object can be grasped. However, in general the configuration of the hand-arm combination will not be unique, as such a manipulator system has more than six degrees of freedom and is said to be super-abundant. The choice of appropriate shares taken by the arm and hand in delivering the manipulation task needs to be resolved. This can be done making use of a kinematic performance measure based on aligning the grip triangle with the hand line of symmetry and maximising the available manipulation range. The hand-arm combination can then be driven to this desired grasp enabling the manipulator to carry out the specified task effectively. A Salisbury hand and PUMA 760 robot arm are used to demonstrate these co-operative motion tasks. / All the experimental results are presented along with a detailed description of the implementation of a hierarchical robot controller system which incorporates force control of the PUMA 760.

Orientador: Marconi Kolm Madrid / Tese (doutorado) - Universidade Estadual de Campinas, Faculdade de Engenharia Eletrica e de Computação / Made available in DSpace on 2018-08-10T09:28:57Z (GMT). No. of bitstreams: 1 Monteiro_DionneCavalcante_D.pdf: 1351119 bytes, checksum: e13d3adc10bf45c4ea22d6ef1b5a7117 (MD5) Previous issue date: 2003 / Resumo: Neste trabalho os algoritmos genéticos e as redes neurais artificiais, técnicas de inteligência artificial, são empregadas para algumas das tarefas que podem ser realizadas por um braço de robô. Inicialmente os algoritmos genéticos são empregados para o controle de trajetória de um robô em um espaço de trabalho que possui a presença de um obstáculo. Operações como crossover e mutação são apresentadas, principalmente por estar-se tratando de trajetórias que são formadas por segmentos de retas. As redes neurais artificiais são testadas no controle direto de dois processos reais usados como paradigma: uma mesa XY e um pêndulo invertido acionado. Para tais processos, é utilizada uma estrutura bastante simplificada, onde a rede neural artificial fornece um ganho para o controlador proporcional que calcula o sinal de controle a ser aplicado. O erro do processo serve para treinar a rede neural sem ser considerado nenhum tipo de treinamento anterior, ou seja, todo o controle neural é executado em tempo real, além disso, uma função determina a taxa de aprendizagem do algoritmo back-propagation em função dos erro existentes nas malhas de controle dos processos. Como existem diversas variáveis para tais controladores neurais, foi também considerado que não existia a possibilidade de se definir o melhor controlador para um determinado processo. Para resolver tal problema, um algoritmo genético foi utilizado para designar qual o melhor controlador para um determinado espaço de trabalho no qual o número de neurônios das camadas de entrada e escondida, constantes de configuração do controlador, e a topologia da rede são otimizados dentro do espaço considerado pelo algoritmo. Todos os resultados importantes obtidos são mostrados, visando mostrar que as técnicas de inteligência artificial podem ser aplicadas à robótica com a vantagem de diminuir, principalmente, o tempo de planejamento de tarefas, tais como: planejamento de trajetória, rastreamento de trajetória, e projeto de controladores eficientes / Abstract: In this work genetic algorithms and artificial neural networks are used for robot arm tasks. Initially, the genetic algorithms are employed to control the trajectory of a robot arm in a limited workspace with an obstacle. Operations like crossover and mutation are presented to manipulate trajectories determined by line segments. Artificial neural networks are tested to control two realtime processes: a XY-Table and an inverted pendulum. For these processes, it is used a simple structured control where the neural network provides a gain to the proportional control, generating a control signal to the processes. The process error is used for training a neural network, without any kind of off-line training, i.e., the training of the neural network is in realtime. Also, a function determines the learning rate of the back-propagation algorithms as a function of the errors of the process control. Since the neural controller have multiple variables, it was not possible to define an optimal controller for the processes. To solve this problem, a genetic algorithm was used to determine the best neural controller in the workspace used, where the number of neurons in the input and hidden layers, constants to configure the neural controller and the network topology are optimized. The results obtained show that artificial intelligent techniques can be applied to robotics reducing the time of task planning, like: trajectory planning, track planning and the project of efficient controllers / Doutorado / Engenharia de Computação / Doutor em Engenharia Elétrica

In the field of mechatronics, different types of robotic arms are used for various applications. Control of robotic arms from a distance is required in certain situations, such as hazardous environments. The purpose of this thesis was to investigate the feasibility, speed, and accuracy of the movement of a robotic arm following the movement of one handheld Inertial Measuring Unit (IMU). The assessment of accuracy was determined through experiments with pre-established movements and examined responses from the arm. The robot arm has four degrees of freedom and is controlled by integrating and filtering the IMU data to obtain the linear position, and inverse kinematics are used to obtain the arm joint angles required to reach the position. The robotic arm was constructed using Solid Edge 3D CAD, 3D printed in PLA plastic. After construction electronic components were connected and assembled. The programs were implemented in MATLAB, and the data was processed and transferred through the Arduino Integrated Development Environment (IDE). The results indicate that the robotic arm demonstrates good capability in executing given coordinates. The accuracy of the IMU-based position tracking is inconsistent and not suitable for all applications. The system's total speed for reading and executing movements is found to be satisfactory, but improvements in precision are necessary for more demanding implementations. The primary causes of errors in the system are attributed to the precision of the measuring device, manufacturing deviations, and limitations in the IMU calculation. This study contributes to the understanding of linear position tracking using inertial sensors, filtering techniques, and communication between microcontrollers, providing insights for future research and development in the field. / Inom området mekatronik används olika typer av robotarmar för olika tillämpningar. Kontroll av robotarmar på avstånd krävs i situationer som farliga miljöer. Syftet med denna avhandling var att undersöka genomförbarhet, hastighet och rörelseprecision hos en robotarm som följer rörelsen av en handhållen tröghetsmätningsenhet (IMU). Bedömningen av noggrannheten fastställdes genom experiment med förutbestämda rörelser och undersökta svar från armen. Robotarmen har fyra frihetsgrader och styrs genom att integrera och filtrera IMU-data för att erhålla den linjära positionen, och inverskinematik används för att erhålla de nödvändiga armledsvinklarna för att nå positionen. Robotarmen konstruerades med hjälp av Solid Edge 3D CAD, 3D-utskrivet i PLA-plast och elektroniska komponenter monterades. Programmen implementerades i MATLAB och data bearbetades och överfördes via Arduino Integrated Development Environment. Resultaten visar att robotarmen har god förmåga att utföra givna koordinater, men noggrannheten i IMU-baserad positionsspårning är inkonsekvent och inte lämplig för alla tillämpningar. Systemets totala hastighet för att läsa av och utföra rörelser anses vara tillfredsställande, men förbättringar av precisionen är nödvändiga för mer krävande implementeringar. De främsta orsakerna till fel i systemet tillskrivs precisionen hos mätinstrumentet, tillverkningsavvikelser och begränsningar i IMU-beräkningen. Denna studie bidrar till förståelsen av linjär positionsspårning med tröghetssensorer, filtreringstekniker och kommunikation mellan mikrokontroller, vilket ger insikter för framtida forskning och utveckling inom området.

Robots are increasingly being used in personal households for service-related tasks such as floor cleaning, lawn mowing or entertainment. However, there is still a lack of household robots performing tactile service-related tasks that requires human-robot interaction. This bachelor’s thesis considers the possible improvement of a robotic head massager performing a head massage by using feedback from a distance sensor. The construction consists of a robot arm with two degrees of freedom. The robot arm is holding a head massager that is implemented with a distance sensor. The distance sensor is used to explore the possible improvement of the safety of the product by stopping the massage whenever the distance to the head was less or equal to 3 cm. The embedded system was controlled by an Arduino Uno that was connected to a computer. The results showed that the distance sensor can in fact improve the safety of the robotic head massager, however it did not always work which opened up a discussion about counter productivity to claim the product as safer. / Användandet av robotar ökar mer och mer i privata hushåll för service-relaterade uppgifter såsom dammsugning, gräsklippning och underhållning. Emellertid finns det fortfarande en brist på hushållsrobotar som utför taktila service-relaterade uppgifter som kräver människa-robot interaktion. Den här kandidatuppsatsen undersöker en möjlig förbättring av en robotarm som utför huvudmassage med hjälp av en huvudkliare och återkoppling. Konstruktionen består av en robot arm med två frihetsgrader. Robotarmen håller en huvudkliare som är implementerad med en distanssensor. Distanssensorn används för att undersöka möjligheten att förbättra säkerheten hos produkten genom att stoppa massagen när avståndet till huvudet var mindre eller lika med 3 cm. Det inbyggda systemet kontrolleras av en Arduino Uno som är kopplad till en dator. Resultaten visar att en distanssensor faktiskt kan förbättra säkerheten hos robot-kliaren, däremot fungerar den inte varje gång vilket öppnar upp för diskussion gällande kontraproduktivitet att benämna produkten som mer säker.

The purpose of this thesis was to design, construct and control a robotic arm with six degrees of freedom. The arm should be able to do simple tasks such as pick and place with good accuracy and without using external sensors. This thesis investigates the precision and the strength of the constructed robot arm. The arm was constructed using 3D printed parts and commonly available hardware such as threaded rods, bearings, screws and nuts. Each axis uses a combination of pulleys and belts in order to achieve desired torque. A differential transmission was implemented in four of the axes in order to combine the power of the motors and reduce weight in the upper parts of the arm. The robot is driven by six stepper motors that are controlled by a combination of RAMPS 1.4 shield and Arduino Mega 2560 microcontroller. The user can manipulate each axis by sending commands to the Arduino through an USB cable. The commands are generated with the help of a simple user interface written in Python. Experiments have shown that the arm has an average error increase of 0.0289-0.1356 mm for each movement, depending on the chosen speed. The maximum amount ofweight that the arm can hold in the worst case scenario is 0.84 kg. / Syftet med denna avhandling var att designa, konstruera och kontrollera en robotarm med sex frihetsgrader. Armen ska kunna utföra enkla uppgifter som pick-and-place med god noggrannhet och utan användning av externa sensorer. Denna avhandling underosöker precisionen och styrkan hos den konstruerade robotarmen. Armen konstruerades med 3D-printade delar och läattillgänglig hårdvara som gängstänger, lager, skruvar och muttrar. Varje axel använder en kombination av kuggremskivor och kuggremmar för att uppnå önskat moment. En differentialväxel användes i fyra av axlarna för att kombinera motorernas moment och minska vikten i armens övre delar. Roboten drivs av sex stegmotorer som styrs av en kombinationav RAMPS 1.4-shield och Arduino Mega 2560 mikrokontroller. Användaren kan styra varje axel genom att skicka kommandon till Arduinon via en USB-kabel. Kommandona genereras med hjälp av ett enkelt användargränssnitt skrivet i Python. Experiment har visat att armen har en genomsnittlig felökning på 0,0289-0,1356 mm för varje rörelse, beroende på vald hastighet. Den högsta vikt som armen i värsta fallkan håalla är 0,84 kg.

The automation revolution already helps in many tasks that are now performed by robots.  Increases in the complexity of problems regarding robot manipulators require new approaches or alternatives in order to solve them. This project comprises a research in different available software for implementing easy and fast visual servoing tasks controlling a robot manipulator. It focuses on out-of-the-box solutions. Then, the tools found are applied to implement a solution for controlling an arm from Universal Robots. The task is to follow a moving object on a plane with the robot manipulator. The research compares the most popular software, the state-of-the-art alternatives, especially in computer vision and also robot control. The implementation aims to be a proof of concept of a system divided by each functionality (computer vision, path generation and robot control) in order to allow software modularity and exchangeability. The results show various options for each system to take into consideration. The implementation is successfully completed, showing the efficiency of the alternatives examined. The chosen software is MATLAB and Simulink for computer vision and trajectory calculation interfacing with Robotic Operating System (ROS). ROS is used for controlling a UR3 arm using ros_control and ur_modern_driver packages.  Both the research and the implementation present a first approach for further applications and understanding over the current technologies for visual servoing tasks. These alternatives offer different easy, fast, and flexible methods to confront complex computer vision and robot control problems.

Made available in DSpace on 2014-12-17T14:56:01Z (GMT). No. of bitstreams: 1 AllanAAFS.pdf: 1304979 bytes, checksum: 691afb58d42d67dd5158ddcf00b02ce5 (MD5) Previous issue date: 2005-03-22 / This work proposes a kinematic control scheme, using visual feedback for a robot arm with five degrees of freedom. Using computational vision techniques, a method was developed to determine the cartesian 3d position and orientation of the robot arm (pose) using a robot image obtained through a camera. A colored triangular label is disposed on the robot manipulator tool and efficient heuristic rules are used to obtain the vertexes of that label in the image. The tool pose is obtained from those vertexes through numerical methods. A color calibration scheme based in the K-means algorithm was implemented to guarantee the robustness of the vision system in the presence of light variations. The extrinsic camera parameters are computed from the image of four coplanar points whose cartesian 3d coordinates, related to a fixed frame, are known. Two distinct poses of the tool, initial and final, obtained from image, are interpolated to generate a desired trajectory in cartesian space. The error signal in the proposed control scheme consists in the difference between the desired tool pose and the actual tool pose. Gains are applied at the error signal and the signal resulting is mapped in joint incrementals using the pseudoinverse of the manipulator jacobian matrix. These incrementals are applied to the manipulator joints moving the tool to the desired pose / Este trabalho prop?e um esquema de controle cinem?tico, utilizando realimenta??o visual para um bra?o rob?tico com cinco graus de liberdade. Foi desenvolvido um m?todo que utiliza t?cnicas de vis?o computacional, para a determina??o da posi??o e orienta??o cartesiana tridimensional (pose) do bra?o rob?tico a partir da imagem do mesmo fornecida por uma c?mera. Um r?tulo triangular colorido ? disposto sobre a garra do manipulador rob?tico e regras heur?sticas eficientes s?o utilizadas para obter os v?rtices desse r?tulo na imagem. M?todos num?ricos s?o usados para recupera??o da pose da garra a partir desses v?rtices. Um esquema de calibra??o de cores fundamentado no algoritmo K-means foi implementado de modo a garantir a robustez do sistema de vis?o na presen?a de varia??es na ilumina??o. Os par?metros extr?nsecos da c?mera s?o calibrados usando-se quatro pontos coplanares extra?dos da imagem, cujas posi??es no plano cartesiano em rela??o a um referencial fixo s?o conhecidas. Para duas poses distintas da garra, inicial e final, adquiridas atrav?s da imagem, interpola-se uma trajet?ria de refer?ncia em espa?o cartesiano. O esquema de controle proposto possui como sinal de erro a diferen?a entre a pose de refer?ncia e a pose atual da garra. Ap?s a aplica??o de ganhos, o sinal de erro ? mapeado em incrementos de junta utilizando-se a pseudoinversa do jacobiano do manipulador. Esses incrementos s?o aplicados ?s juntas do manipulador, deslocando a garra para a pose de refer?ncia