Digital Theses Archive


Tesi etd-12122017-164240

Type of thesis
E-mail address
Robotic Disassembly for the Circular Economy
Scientific disciplinary sector
INGEGNERIA - Biorobotics
relatore Prof. DARIO, PAOLO
  • bimanual dexterous taxonomy
  • circular economy
  • dependability
  • design for disassembly
  • Dual arm manipulation
Exam session start date
This work presents the first results on the development of novel robotic dual-arm vision and sensor guided disassembly methods and practical systems for the Circular Economy. The starting point and the basis of this research is the in-depth analysis of the human manipulation task execution behaviours when disassembling objects of increasing complexity. Our study aims, among other objectives, to show by evidence how the much-needed transition to the Circular Economy paradigm, whose disassembly is one of the main processes, is now technically and economically viable and to provide the basis for some initial concrete guidelines for the development of robotic disassembly.<br>Beginning from the analysis of the manipulation and grasping taxonomies already proposed in the literature and that can be found in the state of the art, the feasibility of the implementation of human-inspired disassembly robotic procedures has been empirically verified by means of our first bimanual dexterous robot experimental model. The model is human-inspired and provides a complete description of the set of grasping and manipulation combinations when developing disassembly tasks, in terms of both hand function and hand configuration. It can be applied to one-hand and to bimanual manipulation operations to comply with all different tasks of a specific application. As such, it represents the first model of human bimanual dexterous manipulation. The model is empirically grounded and results from a study of twenty participants performing non-destructive disassembly of three different objects (a ceiling roller, a pencil sharpener and a portable vacuum cleaner). Human performances in executing the disassembly of the objects have been analysed and the bimanual grasp and manipulation combinations that led to a successful disassembly mapped into the newly developed model. The mapped results constitute a base empirical ground for the development of novel effective and efficient bioinspired robotic disassembly strategies. It has been already used as a tool to define algorithms allowing the implementation of some robotic disassembly task. The model itself could easily be applied to any other application and also to human behavioural studies.<br>The analysis of human behaviours – on top of its per se scientific interest - allows to define a set of objective performance benchmarks to assess the quality and in particular the dependability of disassembly tasks when performed by robots. Dependability, meant as the level of trust that we can rely on the capability of a robotic platform to perform the tasks for which it has been designed, is the study at the basis of the introduction of a new technological process such as the automated disassembly. Such human performances have also been analysed in terms of “failed attempts”, defined as the manipulation combinations that did not led to a successful disassembly. We believe that such analysis could serve as a benchmark for dependability assessments of robotic platforms implementing the same task and a first preliminary analysis is reported.<br>Participants were requested to complete a set of questionnaires to provide their perception in terms of complexity when disassembling the objects, and also estimate the difficulties of a potential robotic platform performing the same tasks. Such analysis might be of interest for further AI and Machine Learning studies. A subset of participants having a background in industrial design also provided an initial feedback in terms of optimized Design For Disassembly.<br>The performance model, being by itself a guideline for robotics implementation of human and robotic disassembly, is intended to be also a tool to compare and evaluate different grippers as well as robotic platforms and motion planning algorithms. <br>Further research will focus on introducing haptic information on objects that have reached their End of Life and consider how their compromised status affects the disassembly process, also from a design for robotic disassembly perspective.<br>Disassembly tasks have been implemented and tested on a bimanual robotic platform as well. The hardware platform utilized to test the control algorithms and perform the experimental tests integrates the SMART5 RML Comau Dual-Arm anthropomorphic robot on a TCP-IP network with a vision system that guided its actions (visual servoing), torque sensing and control and that can be easily expanded to various governing. The system was able to perform preliminary disassembly tasks in a semi-structured work environment implementing a set of complex coordinated operations, such as unscrewing a mechanical piece from its site. A preliminary comparison of human and robot performances when implementing the unscrewing operations is provided in order to assess the reliability of such initial robotic set up.<br>In Chapter 1 an introduction on the goals, and motivations of the research and its methods and tools are introduced.<br>In Chapter 2 the Circular Economy’s main concepts and one of its bottleneck processes, disassembly, are discussed in the perspective of their robotization and automation. A new paradigm is also introduced, the peg-out-of-hole problem, and a set of guidelines for Design for Disassembly provided.<br>In Chapter 3 manipulation and grasping notions are exploited, focusing on artificial hands and dual-arm robotic platform. <br>Chapter 4 provides the state of the art of Dependability of robotic platforms and set the basis to analyse the performance of a new robotic set up (Chapter 6) compared to human performances (Chapter 5). <br>Chapter 5 describes the first human-inspired model of bimanual dexterous manipulation for disassembly tasks. Preliminary studies are shown, and an innovative bimanual dexterous manipulation model described. Two case studies are analysed: a disassembly of a car suspension by skilled and non-skilled technicians and the disassembly of three objects and the consequential mapping of their data on the bimanual model. <br>In Chapter 6 an example of bimanual coordinated manipulation using a Dual-Arm robot, on the basis of the arguments introduced in Chapter 3, is presented. We developed an experimental setup that was able to recognize different object put on a work table (using a 3D vision system), to choose which object to operate on (realization of task planning algorithms) and to perform disassembly operations on the chosen object, adopting both robotic arms in cooperating bimanual manipulation. The robotic performances are analysed based on the notion gained in Chapter 3 and compared to the human ones shown in Chapter 5.<br>In Chapter 7, the Conclusions and future developments are discussed, while Chapter 8 and Chapter 9 are for additional reference and acknowledgments. <br>