A Deep Reinforcement Learning Approach for Active SLAM
In this paper, we formulate the active SLAM paradigm in terms of model-free Deep Reinforcement Learning, embedding the traditional utility functions based on the Theory of Optimal Experimental Design in rewards, and therefore relaxing the intensive computations of classical approaches. We validate s...
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2020-11-01
|
| Series: | Applied Sciences |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2076-3417/10/23/8386 |
| _version_ | 1797546775696900096 |
|---|---|
| author | Julio A. Placed José A. Castellanos |
| author_facet | Julio A. Placed José A. Castellanos |
| author_sort | Julio A. Placed |
| collection | DOAJ |
| description | In this paper, we formulate the active SLAM paradigm in terms of model-free Deep Reinforcement Learning, embedding the traditional utility functions based on the Theory of Optimal Experimental Design in rewards, and therefore relaxing the intensive computations of classical approaches. We validate such formulation in a complex simulation environment, using a state-of-the-art deep Q-learning architecture with laser measurements as network inputs. Trained agents become capable not only to learn a policy to navigate and explore in the absence of an environment model but also to transfer their knowledge to previously unseen maps, which is a key requirement in robotic exploration. |
| first_indexed | 2024-03-10T14:35:00Z |
| format | Article |
| id | doaj.art-67761e793c1343cdb40dd40ef46adf20 |
| institution | Directory Open Access Journal |
| issn | 2076-3417 |
| language | English |
| last_indexed | 2024-03-10T14:35:00Z |
| publishDate | 2020-11-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Applied Sciences |
| spelling | doaj.art-67761e793c1343cdb40dd40ef46adf202023-11-20T22:17:42ZengMDPI AGApplied Sciences2076-34172020-11-011023838610.3390/app10238386A Deep Reinforcement Learning Approach for Active SLAMJulio A. Placed0José A. Castellanos1Instituto de Investigación en Ingeniería de Aragón (I3A), Universidad de Zaragoza, C/ María de Luna 1, 50018 Zaragoza, SpainInstituto de Investigación en Ingeniería de Aragón (I3A), Universidad de Zaragoza, C/ María de Luna 1, 50018 Zaragoza, SpainIn this paper, we formulate the active SLAM paradigm in terms of model-free Deep Reinforcement Learning, embedding the traditional utility functions based on the Theory of Optimal Experimental Design in rewards, and therefore relaxing the intensive computations of classical approaches. We validate such formulation in a complex simulation environment, using a state-of-the-art deep Q-learning architecture with laser measurements as network inputs. Trained agents become capable not only to learn a policy to navigate and explore in the absence of an environment model but also to transfer their knowledge to previously unseen maps, which is a key requirement in robotic exploration.https://www.mdpi.com/2076-3417/10/23/8386active SLAMrobotic explorationoptimality criteriadeep reinforcement learningdeep Q-networks |
| spellingShingle | Julio A. Placed José A. Castellanos A Deep Reinforcement Learning Approach for Active SLAM Applied Sciences active SLAM robotic exploration optimality criteria deep reinforcement learning deep Q-networks |
| title | A Deep Reinforcement Learning Approach for Active SLAM |
| title_full | A Deep Reinforcement Learning Approach for Active SLAM |
| title_fullStr | A Deep Reinforcement Learning Approach for Active SLAM |
| title_full_unstemmed | A Deep Reinforcement Learning Approach for Active SLAM |
| title_short | A Deep Reinforcement Learning Approach for Active SLAM |
| title_sort | deep reinforcement learning approach for active slam |
| topic | active SLAM robotic exploration optimality criteria deep reinforcement learning deep Q-networks |
| url | https://www.mdpi.com/2076-3417/10/23/8386 |
| work_keys_str_mv | AT julioaplaced adeepreinforcementlearningapproachforactiveslam AT joseacastellanos adeepreinforcementlearningapproachforactiveslam AT julioaplaced deepreinforcementlearningapproachforactiveslam AT joseacastellanos deepreinforcementlearningapproachforactiveslam |