Robotic grasping: State of the art and modular real-world implementation

Robotic Grasping is an important research topic in robotics since for robots to attain more general-purpose utility, grasping is a necessary skill, but very challenging to master. In general the robots may use their perception abilities like an image from a camera to identify grasps for a given object usually unknown. A grasp describes how a robotic end-effector need to be positioned to securely grab an object and successfully lift it without lost it, at the moment state of the arts solutions are still far behind humans. In the last 5–10 years, deep learning methods take the scene to overcome classical problem like the arduous and time-consuming approach to form a task-specific algorithm analytically. In this thesis are present the progress and the approaches in the robotic grasping field and the potential of the deep learning methods in robotic grasping. Based on that, an implementation of a Convolutional Neural Network (CNN) as a starting point for generation of a grasp pose from camera view has been implemented inside a ROS environment. The developed technologies have been integrated into a pick-and-place application for a Panda robot from Franka Emika. The application includes various features related to object detection and selection. Additionally, the features have been kept as generic as possible to allow for easy replacement or removal if needed, without losing time for improvement or new testing.

Structure learning of graphical models for task-oriented robot grasping

In the collective imaginaries a robot is a human like machine as any androids in science fiction. However the type of robots that you will encounter most frequently are machinery that do work that is too dangerous, boring or onerous. Most of the robots in the world are of this type. They can be found in auto, medical, manufacturing and space industries. Therefore a robot is a system that contains sensors, control systems, manipulators, power supplies and software all working together to perform a task. The development and use of such a system is an active area of research and one of the main problems is the development of interaction skills with the surrounding environment, which include the ability to grasp objects. To perform this task the robot needs to sense the environment and acquire the object informations, physical attributes that may influence a grasp. Humans can solve this grasping problem easily due to their past experiences, that is why many researchers are approaching it from a machine learning perspective finding grasp of an object using information of already known objects. But humans can select the best grasp amongst a vast repertoire not only considering the physical attributes of the object to grasp but even to obtain a certain effect. This is why in our case the study in the area of robot manipulation is focused on grasping and integrating symbolic tasks with data gained through sensors. The learning model is based on Bayesian Network to encode the statistical dependencies between the data collected by the sensors and the symbolic task. This data representation has several advantages. It allows to take into account the uncertainty of the real world, allowing to deal with sensor noise, encodes notion of causality and provides an unified network for learning. Since the network is actually implemented and based on the human expert knowledge, it is very interesting to implement an automated method to learn the structure as in the future more tasks and object features can be introduced and a complex network design based only on human expert knowledge can become unreliable. Since structure learning algorithms presents some weaknesses, the goal of this thesis is to analyze real data used in the network modeled by the human expert, implement a feasible structure learning approach and compare the results with the network designed by the expert in order to possibly enhance it.

Robotic patrolling: analisi di strategie e applicazione ad uno scenario

Introduzione alle strategie di robotic patrolling e analisi delle stesse. Applicazione ad uno scenario di una strategia e realizzazione di un robot patroller.

Swarm constructability: designing through embedded tectonic behaviors in swarm robotic systems

Con questa tesi di laurea si muovono i primi passi di una ricerca applicata finalizzata alla costruzione-deposizione di materiale da parte di sciami di mini-robot dal comportamento indipendente che si coordinano tramite segnali lasciati e rilevati nell’ambiente in cui si muovono. Lo sviluppo di tecniche di progettazione e fabbricazione digitale ha prodotto un aumento nel grado di interconnessione tra tecnologia e design, dunque, di nuove possibilità tettoniche. Le relazioni tettoniche tradizionali stanno infatti subendo una trasformazione radicale, potendo essere esplicitamente informate e dunque mediate attraverso gli strumenti digitali dall’ideazione alla produzione. Questa mediazione informata del contenuto tettonico (che opera costantemente) è distintivo di un approccio material-based alla progettazione che aumenta l’integrazione tra struttura, materia e forma entro le tecnologie di fabbricazione (R.Oxman). Dei numerosi processi di fabbricazione per l’architettura che si servono di tecnologia robotica, pochi sono capaci di superare la logica gerarchica, rigida e lineare-sequenziale che serve di fatto agli obiettivi di automazione ed ottimizzazione. La distribuzione di forme di intelligenza semplificata ad un numero elevato di unità robot è quindi qui proposta come alternativa al modello appena descritto. Incorporando semplici decisioni di carattere architettonico negli agenti-robot che costituiscono il sistema distribuito di entità autonome, la loro interazione e le decisioni prese individualmente producono comportamento collettivo e l’integrazione delle suddette relazioni tettoniche. Nello sviluppo del progetto, si è fatto così riferimento a modelli comportamentali collettivi (di sciame) osservabili in specie comunitarie che organizzano strutture materiali -come termiti e vespe- ed in organismi semplici -come le muffe cellulari della specie Physarum polycephalum. Per queste specie biologiche il processo di costruzione non dipende da un ‘piano generale’ ma è guidato esclusivamente da azioni dei singoli individui che comunicano lasciando tracce chimiche nell’ambiente e modificano il loro comportamento rilevando le tracce lasciate dagli altri individui. A questo scopo, oltre alle simulazioni in digitale, è stato indispensabile sviluppare dei prototipi funzionali di tipo fisico, ovvero la realizzazione di mini-robot dal movimento indipendente, in grado di coordinarsi tra loro tramite segnali lasciati nell’ambiente e capaci di depositare materiale.

Integration of robotic inverse kinematic routines in an algorithmic 3D modeling software

An industrial manipulator equipped with an automatic clay extruder is used to realize a machine that can manufacture additively clay objects. The desired geometries are designed by means of a 3D modeling software and then sliced in a sequence of layers with the same thickness of the extruded clay section. The profiles of each layer are transformed in trajectories for the extruder and therefore for the end-effector of the manipulator. The goal of this thesis is to improve the algorithm for the inverse kinematic resolution and the integration of the routine within the development software that controls the machine (Rhino/Grasshopper). The kinematic model is described by homogeneous transformations, adopting the Denavit-Hartenberg standard convention. The function is implemented in C# and it has been preliminarily tested in Matlab. The outcome of this work is a substantial reduction of the computation time relative to the execution of the algorithm, which is halved.

A ROS-based Workspace Control and Trajectory Planner for a seven degrees of freedom Robotic arm

In this Bachelor Thesis I want to provide readers with tools and scripts for the control of a 7DOF manipulator, backed up by some theory of Robotics and Computer Science, in order to better contextualize the work done. In practice, we will see most common software, and developing environments, used to cope with our task: these include ROS, along with visual simulation by VREP and RVIZ, and an almost "stand-alone" ROS extension called MoveIt!, a very complete programming interface for trajectory planning and obstacle avoidance. As we will better appreciate and understand in the introduction chapter, the capability of detecting collision objects through a camera sensor, and re-plan to the desired end-effector pose, are not enough. In fact, this work is implemented in a more complex system, where recognition of particular objects is needed. Through a package of ROS and customized scripts, a detailed procedure will be provided on how to distinguish a particular object, retrieve its reference frame with respect to a known one, and then allow navigation to that target. Together with technical details, the aim is also to report working scripts and a specific appendix (A) you can refer to, if desiring to put things together.

Integration of the computer controlled screwdriver for Robotic manufacturing in the ROS environment

The computer controlled screwdriver is a modern technique to perform automatic screwing/unscrewing operations.The main focus is to study the integration of the computer controlled screwdriver for Robotic manufacturing in the ROS environment.This thesis describes a concept of automatic screwing mechanism composed by universal robots, in which one arm of the robot is for inserting cables and the other is for screwing the cables on the control panel switch gear box. So far this mechanism is carried out by human operators and is a fairly complex one to perform, due to the multiple cables and connections involved. It's for this reason that an automatic cabling and screwing process would be highly preferred within automotive/automation industries. A study is carried out to analyze the difficulties currently faced and a controller based algorithm is developed to replace the manual human efforts using universal robots, thereby allowing robot arms to insert the cables and screw them onto the control panel switch gear box. Experiments were conducted to evaluate the insertion and screwing strategy, which shows the result of inserting and screwing cables on the control panel switch gearbox precisely.

Development of a robotic system for switchgear cabling

Industrial robots are an inalienable part of modern automated production. Typical applications of robots include welding, painting, (dis)assembly, packaging, labeling, palletizing, pick and place and others. Many of that applications includes object manipulation. If the shape and position of the object are known in advance, it is possible to design the trajectory of the robot’s end-effector to take and place. Such a strategy is applicable for rigid objects and widely used in the manufacturing field. But flexible (deformable) objects can change their shape and position upon contact with the robot’s end-effector or environment. That is the reason why the general approach is unacceptable. It means that the robot can fail to grasp such an object and can’t place it in the desired position. This thesis has addressed the problem of cable manipulation by bilateral robotic setup for the industrial manufacturing of electrical switchgear. The considered solution is based on the idea of tensioned cable. If the cable was grasped by the ends and tensioned, it has a line shape. Since the position of the robot’s end-effectors known, the position of the cable is known as well. Such an approach is capable to place cable in cable ducts of switchgear. The considered solution has been tested experimentally on a real bilateral robotic setup.

Integration of ROS for the control of industrial robotic systems in a proprietary environment

The paper deals with the integration of ROS, in the proprietary environment of the Marchesini Group company, for the control of industrial robotic systems. The basic tools of this open-source software are deeply studied to model a full proprietary Pick and Place manipulator inside it, and to develop custom ROS nodes to calculate trajectories; speaking of which, the URDF format is the standard to represent robots in ROS and the motion planning framework MoveIt offers user-friendly high-level methods. The communication between ROS and the Marchesini control architecture is established using the OPC UA standard; the tasks computed are transmitted offline to the PLC, supervisor controller of the physical robot, because the performances of the protocol don’t allow any kind of active control by ROS. Once the data are completely stored at the Marchesini side, the industrial PC makes the real robot execute a trajectory computed by MoveIt, so that it replicates the behaviour of the simulated manipulator in Rviz. Multiple experiments are performed to evaluate in detail the potential of ROS in the planning of movements for the company proprietary robots. The project ends with a small study regarding the use of ROS as a simulation platform. First, it is necessary to understand how a robotic application of the company can be reproduced in the Gazebo real world simulator. Then, a ROS node extracts information and examines the simulated robot behaviour, through the subscription to specific topics.

Robotic Manipulation of DLOs for Wiring Harness Assembly: a Machine Learning Approach

In recent times, a significant research effort has been focused on how deformable linear objects (DLOs) can be manipulated for real world applications such as assembly of wiring harnesses for the automotive and aerospace sector. This represents an open topic because of the difficulties in modelling accurately the behaviour of these objects and simulate a task involving their manipulation, considering a variety of different scenarios. These problems have led to the development of data-driven techniques in which machine learning techniques are exploited to obtain reliable solutions. However, this approach makes the solution difficult to be extended, since the learning must be replicated almost from scratch as the scenario changes. It follows that some model-based methodology must be introduced to generalize the results and reduce the training effort accordingly. The objective of this thesis is to develop a solution for the DLOs manipulation to assemble a wiring harness for the automotive sector based on adaptation of a base trajectory set by means of reinforcement learning methods. The idea is to create a trajectory planning software capable of solving the proposed task, reducing where possible the learning time, which is done in real time, but at the same time presenting suitable performance and reliability. The solution has been implemented on a collaborative 7-DOFs Panda robot at the Laboratory of Automation and Robotics of the University of Bologna. Experimental results are reported showing how the robot is capable of optimizing the manipulation of the DLOs gaining experience along the task repetition, but showing at the same time a high success rate from the very beginning of the learning phase.

Development of a Robotic Wires Storage System for Switchgear Manufacturing Automation

Considering the great development of robotics in industrial automation, the Remodel project aims to reproduce, through the use of Cobots, the wiring activity typical of a human operator and to realize an autonomous storage work. My researches focused on this second topic. In this paper, we will see how to realize a gripper compatible with an Omron TM5X-900, able to perform a pick and place of different types of cables, but also how to compute possible trajectories. In particular, what I needed, was a trajectory going from the Komax, the cables production machine, to a Warehouse taking into account the possible entangles of cables with the robot during its motion. The last part has been dedicated to the synchronization between robot and main machine work.

A comprehensive study on robotic manipulation of deformable linear objects

There are many deformable objects such as papers, clothes, ropes in a person’s living space. To have a robot working in automating the daily tasks it is important that the robot works with these deformable objects. Manipulation of deformable objects is a challenging task for robots because these objects have an infinite-dimensional configuration space and are expensive to model, making real-time monitoring, planning and control difficult. It forms a particularly important field of robotics with relevant applications in different sectors such as medicine, food handling, manufacturing, and household chores. In this report, there is a clear review of the approaches used and are currently in use along with future developments to achieve this task. My research is more focused on the last 10 years, where I have systematically reviewed many articles to have a clear understanding of developments in this field. The main contribution is to show the whole landscape of this concept and provide a broad view of how it has evolved. I also explained my research methodology by following my analysis from the past to the present along with my thoughts for the future.

Integrazione delle immagini ecotomografiche nei sistemi per chirurgia robotica: Stato dell'arte e sviluppi futuri

L'elaborato si propone di analizzare la fusione di due tecnologie diverse, ma allo stesso modo rivoluzionarie. Da un lato abbiamo la chirurgia robotica (Robotic Assisted Surgery), reputata ormai la nuova frontiera della chirurgia mini-invasiva, dall'altro invece l'eco-tomografia, che in molti ipotizzano poter divenire lo 'stetoscopio' del futuro.

Feedback linearization of antagonistically actuated variable stiffness joints with twisted string actuators

The work of this thesis is on the implementation of a variable stiffness joint antagonistically actuated by a couple of twisted-string actuator (TSA). This type of joint is possible to be applied in the field of robotics, like UB Hand IV (the anthropomorphic robotic hand developed by University of Bologna). The purposes of the activities are to build the joint dynamic model and simultaneously control the position and stiffness. Three different control approaches (Feedback linearization, PID, PID+Feedforward) are proposed and validated in simulation. To improve the properties of joint stiffness, a joint with elastic element is taken into account and discussed. To the end, the experimental setup that has been developed for the experimental validation of the proposed control approaches.

La piattaforma ros per lo sviluppo di applicazioni robotiche

Lo sviluppo hardware nel campo della robotica ha raggiunto negli ultimi anni livelli impressionanti ed è in continua crescita, e di pari passo si è espansa l’eterogeneità delle forme che può assumere, dalle tipologie basate su movimento a terra ai droni volanti, fino a forme più sofisticate di robot umanoidi che cercano di emularne il comportamento. Se da un lato ora possiamo disporre di hardware sempre più potente ed efficiente a costi sempre minori, dall’altro programmare il comportamento che un robot deve tenere nelle svariate circostanze in cui può imbattersi, nel poter portare a compimento il proprio obbiettivo, risulta essere sempre più complesso. Dopo una breve introduzione alla robotica e alle difficoltà che deve affrontare e una panoramica sui robot, cosa siano e come siano strutturati, fulcro della tesi sarà l’esposizione delle caratteristiche principali di ROS, Robot Operating System, come piattaforma di sviluppo software nel campo della robotica, e si concluderà con un semplice caso di studio in cui ne verrà messo in mostra concretamente l’utilizzo.