Learning cloth manipulation with demonstrations
Recent advances in Deep Reinforcement learning and computational capabilities of GPUs have led to variety of research being conducted in the learning side of robotics. The main aim being that of making autonomous robots that are capable of learning how to solve a task on their own with minimal requi...
| Autores: | , , |
|---|---|
| Tipo de recurso: | informe técnico |
| Fecha de publicación: | 2019 |
| País: | España |
| Institución: | Universitat Politècnica de Catalunya (UPC) |
| Repositorio: | UPCommons. Portal del coneixement obert de la UPC |
| Idioma: | inglés |
| OAI Identifier: | oai:upcommons.upc.edu:2117/182696 |
| Acceso en línea: | https://hdl.handle.net/2117/182696 |
| Access Level: | acceso abierto |
| Palabra clave: | Classificació INSPEC::Cybernetics::Artificial intelligence::Learning (artificial intelligence) Àrees temàtiques de la UPC::Informàtica::Robòtica |
| Sumario: | Recent advances in Deep Reinforcement learning and computational capabilities of GPUs have led to variety of research being conducted in the learning side of robotics. The main aim being that of making autonomous robots that are capable of learning how to solve a task on their own with minimal requirement for engineering on the planning, vision, or control side. Efforts have been made to learn the manipulation of rigid objects through the help of human demonstrations, specifically in the tasks such as stacking of multiple blocks on top of each other, inserting a pin into a hole, etc. These Deep RL algorithms successfully learn how to complete a task involving the manipulation of rigid objects, but autonomous manipulation of textile objects such as clothes through Deep RL algorithms is still not being studied in the community. The main objectives of this work involve, 1) implementing the state of the art Deep RL algorithms for rigid object manipulation and getting a deep understanding of the working of these various algorithms, 2) Creating an open-source simulation environment for simulating textile objects such as clothes, 3) Designing Deep RL algorithms for learning autonomous manipulation of textile objects through demonstrations. |
|---|