Robot Learning from Demonstration with Gaussian Processes
| Autor: | Arduengo, Miguel |
|---|---|
| Přispěvatelé: | Ocampo-Martínez, Carlos, Torras, Carme, Colomé, Adrià, Ministerio de Economía y Competitividad (España), European Commission, Universitat Politècnica de Catalunya. Departament d'Enginyeria de Sistemes, Automàtica i Informàtica Industrial |
| Jazyk: | Catalan; Valencian |
| Rok vydání: | 2021 |
| Předmět: |
Autonomous robots -- Design and construction
Informàtica [Àrees temàtiques de la UPC] Aprenentatge supervisat (Aprenentatge automàtic) Aprenentatge automàtic -- Avaluació -- Models matemàtics Enginyeria electrònica [Àrees temàtiques de la UPC] Matemàtiques i estadística [Àrees temàtiques de la UPC] Machine learning -- Evaluation -- Mathematical models Supervised learning (Machine learning) Robots autònoms -- Disseny i construcció |
| Zdroj: | UPCommons. Portal del coneixement obert de la UPC Universitat Politècnica de Catalunya (UPC) Digital.CSIC. Repositorio Institucional del CSIC instname |
| Popis: | Trabajo fin de máster presentado en la Universidad Politécnica de Cataluña.--2021-06-21 Autonomous systems are no longer confined to factories, but they are progressively spreading to urban, social, and assistive domains. However, in order to become handy co-workers and helpful assistants, robots must be endowed with quite different abilities than their industrial ancestors, and a lot of additional research is still required. A key challenge in intelligent robotics is creating autonomous agents that are capable of directly interacting with the world around them to achieve their goals. Learning plays a central role in intelligent autonomous systems, as the real world contains too much uncertainty and a robot must be capable of dealing with environments that neither it nor its designers have foreseen. Learning from demonstration is a promising paradigm that allows robots to learn complex tasks that cannot be easily scripted, but can be demonstrated by a human teacher. In this thesis, we develop complete learning from demonstration framework. We first present a whole-body teleoperation approach for human motion transfer, which allows a teacher equipped with a motion capture system to intuitively provide demonstrations to a robot. Then, to learn a generalized rep- resentation of the task which can be adapted to unforeseen scenarios, we unify in a single, entirely Gaussian-Process-based formulation, the main components of a state-of-the-art method. We evaluate our approach through a series of real-world experiments with the manipulator robot TIAGo, achieving satisfactory results. Finally, we must be aware that we are in a technological inflection point in which robots are developing the capacity to greatly increase their cognitive and physical capabilities. This will raise complex issues regarding the economy, ethics, law, and the environment, which we provide an overview of in this thesis. Intelligent robotics offer an unimaginable spectrum of possibilities, with the appropriate attention and the right policies they open the doors to new sources of value and growth. However, it is in the hands of scientists and engineers to not look away and anticipate the potential impacts in order to turn robots into the motor of global prosperity |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|