Deep reinforcement learning using compositional representations for performing instructions
Spoken language is one of the most efficientways to instruct robots about performing domestic tasks. However, the state of the environment has to be considered to plan and execute actions successfully. We propose a system that learns to recognise the user’s intention and map it to a goal. A reinforc...
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
De Gruyter
2018-12-01
|
| Series: | Paladyn |
| Subjects: | |
| Online Access: | https://doi.org/10.1515/pjbr-2018-0026 |
| _version_ | 1797428596567965696 |
|---|---|
| author | Zamani Mohammad Ali Magg Sven Weber Cornelius Wermter Stefan Fu Di |
| author_facet | Zamani Mohammad Ali Magg Sven Weber Cornelius Wermter Stefan Fu Di |
| author_sort | Zamani Mohammad Ali |
| collection | DOAJ |
| description | Spoken language is one of the most efficientways to instruct robots about performing domestic tasks. However, the state of the environment has to be considered to plan and execute actions successfully. We propose a system that learns to recognise the user’s intention and map it to a goal. A reinforcement learning (RL) system then generates a sequence of actions toward this goal considering the state of the environment. A novel contribution in this paper is the use of symbolic representations for both input and output of a neural Deep Q-network (DQN), which enables it to be used in a hybrid system. To show the effectiveness of our approach, the Tell-Me-Dave corpus is used to train an intention detection model and in a second step an RL agent generates the sequences of actions towards the detected objective, represented by a set of state predicates. We show that the system can successfully recognise command sequences fromthis corpus aswell as train the deep- RL network with symbolic input.We further show that the performance can be significantly increased by exploiting the symbolic representation to generate intermediate rewards. |
| first_indexed | 2024-03-09T09:01:09Z |
| format | Article |
| id | doaj.art-5a5754a5178344bfac592fe4b31d50c3 |
| institution | Directory Open Access Journal |
| issn | 2081-4836 |
| language | English |
| last_indexed | 2024-03-09T09:01:09Z |
| publishDate | 2018-12-01 |
| publisher | De Gruyter |
| record_format | Article |
| series | Paladyn |
| spelling | doaj.art-5a5754a5178344bfac592fe4b31d50c32023-12-02T11:31:28ZengDe GruyterPaladyn2081-48362018-12-019135837310.1515/pjbr-2018-0026pjbr-2018-0026Deep reinforcement learning using compositional representations for performing instructionsZamani Mohammad Ali0Magg Sven1Weber Cornelius2Wermter Stefan3Fu Di4Knowledge Technology, Department of Informatics, University of Hamburg, Vogt-Koelln-Str. 30,Hamburg, GermanyKnowledge Technology, Department of Informatics, University of Hamburg, Vogt- Koelln-Str. 30,Hamburg, GermanyKnowledge Technology, Department of Informatics, University of Hamburg, Vogt- Koelln-Str. 30,Hamburg, GermanyKnowledge Technology, Department of Informatics, University of Hamburg, Vogt- Koelln-Str. 30,Hamburg, GermanyCAS Key Laboratory of Behavioral Science, Chinese Academy of Sciences,Beijing, ChinaSpoken language is one of the most efficientways to instruct robots about performing domestic tasks. However, the state of the environment has to be considered to plan and execute actions successfully. We propose a system that learns to recognise the user’s intention and map it to a goal. A reinforcement learning (RL) system then generates a sequence of actions toward this goal considering the state of the environment. A novel contribution in this paper is the use of symbolic representations for both input and output of a neural Deep Q-network (DQN), which enables it to be used in a hybrid system. To show the effectiveness of our approach, the Tell-Me-Dave corpus is used to train an intention detection model and in a second step an RL agent generates the sequences of actions towards the detected objective, represented by a set of state predicates. We show that the system can successfully recognise command sequences fromthis corpus aswell as train the deep- RL network with symbolic input.We further show that the performance can be significantly increased by exploiting the symbolic representation to generate intermediate rewards.https://doi.org/10.1515/pjbr-2018-0026deep reinforcement learningspoken language instruction |
| spellingShingle | Zamani Mohammad Ali Magg Sven Weber Cornelius Wermter Stefan Fu Di Deep reinforcement learning using compositional representations for performing instructions Paladyn deep reinforcement learning spoken language instruction |
| title | Deep reinforcement learning using compositional representations for performing instructions |
| title_full | Deep reinforcement learning using compositional representations for performing instructions |
| title_fullStr | Deep reinforcement learning using compositional representations for performing instructions |
| title_full_unstemmed | Deep reinforcement learning using compositional representations for performing instructions |
| title_short | Deep reinforcement learning using compositional representations for performing instructions |
| title_sort | deep reinforcement learning using compositional representations for performing instructions |
| topic | deep reinforcement learning spoken language instruction |
| url | https://doi.org/10.1515/pjbr-2018-0026 |
| work_keys_str_mv | AT zamanimohammadali deepreinforcementlearningusingcompositionalrepresentationsforperforminginstructions AT maggsven deepreinforcementlearningusingcompositionalrepresentationsforperforminginstructions AT webercornelius deepreinforcementlearningusingcompositionalrepresentationsforperforminginstructions AT wermterstefan deepreinforcementlearningusingcompositionalrepresentationsforperforminginstructions AT fudi deepreinforcementlearningusingcompositionalrepresentationsforperforminginstructions |