Robot Navigation in Crowded Environments: A Reinforcement Learning Approach
For a mobile robot, navigation in a densely crowded space can be a challenging and sometimes impossible task, especially with traditional techniques. In this paper, we present a framework to train neural controllers for differential drive mobile robots that must safely navigate a crowded environment...
| Hlavní autoři: | , , , , , |
|---|---|
| Médium: | Článek |
| Jazyk: | English |
| Vydáno: |
MDPI AG
2023-02-01
|
| Edice: | Machines |
| Témata: | |
| On-line přístup: | https://www.mdpi.com/2075-1702/11/2/268 |
| _version_ | 1827756737481932800 |
|---|---|
| author | Matteo Caruso Enrico Regolin Federico Julian Camerota Verdù Stefano Alberto Russo Luca Bortolussi Stefano Seriani |
| author_facet | Matteo Caruso Enrico Regolin Federico Julian Camerota Verdù Stefano Alberto Russo Luca Bortolussi Stefano Seriani |
| author_sort | Matteo Caruso |
| collection | DOAJ |
| description | For a mobile robot, navigation in a densely crowded space can be a challenging and sometimes impossible task, especially with traditional techniques. In this paper, we present a framework to train neural controllers for differential drive mobile robots that must safely navigate a crowded environment while trying to reach a target location. To learn the robot’s policy, we train a convolutional neural network using two Reinforcement Learning algorithms, <i>Deep Q-Networks</i> (DQN) and <i>Asynchronous Advantage Actor Critic</i> (A3C) and develop a training pipeline that allows to scale the process to several compute nodes. We show that the asynchronous training procedure in A3C can be leveraged to quickly train neural controllers and test them on a real robot in a crowded environment. |
| first_indexed | 2024-03-11T08:31:27Z |
| format | Article |
| id | doaj.art-56ee191bc95647dbba51b1351523982c |
| institution | Directory Open Access Journal |
| issn | 2075-1702 |
| language | English |
| last_indexed | 2024-03-11T08:31:27Z |
| publishDate | 2023-02-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Machines |
| spelling | doaj.art-56ee191bc95647dbba51b1351523982c2023-11-16T21:46:15ZengMDPI AGMachines2075-17022023-02-0111226810.3390/machines11020268Robot Navigation in Crowded Environments: A Reinforcement Learning ApproachMatteo Caruso0Enrico Regolin1Federico Julian Camerota Verdù2Stefano Alberto Russo3Luca Bortolussi4Stefano Seriani5Department of Engineering and Architecture, University of Trieste, Via A. Valerio 6/1, 34127 Trieste, ItalyDepartment of Matematics and Geoscience, University of Trieste, Via Edoardo Weiss 2, 34128 Trieste, ItalyDepartment of Matematics and Geoscience, University of Trieste, Via Edoardo Weiss 2, 34128 Trieste, ItalyDepartment of Matematics and Geoscience, University of Trieste, Via Edoardo Weiss 2, 34128 Trieste, ItalyDepartment of Matematics and Geoscience, University of Trieste, Via Edoardo Weiss 2, 34128 Trieste, ItalyDepartment of Engineering and Architecture, University of Trieste, Via A. Valerio 6/1, 34127 Trieste, ItalyFor a mobile robot, navigation in a densely crowded space can be a challenging and sometimes impossible task, especially with traditional techniques. In this paper, we present a framework to train neural controllers for differential drive mobile robots that must safely navigate a crowded environment while trying to reach a target location. To learn the robot’s policy, we train a convolutional neural network using two Reinforcement Learning algorithms, <i>Deep Q-Networks</i> (DQN) and <i>Asynchronous Advantage Actor Critic</i> (A3C) and develop a training pipeline that allows to scale the process to several compute nodes. We show that the asynchronous training procedure in A3C can be leveraged to quickly train neural controllers and test them on a real robot in a crowded environment.https://www.mdpi.com/2075-1702/11/2/268mobile roboticsneural networkscontrol systemsreinforcement learningcrowd navigation |
| spellingShingle | Matteo Caruso Enrico Regolin Federico Julian Camerota Verdù Stefano Alberto Russo Luca Bortolussi Stefano Seriani Robot Navigation in Crowded Environments: A Reinforcement Learning Approach Machines mobile robotics neural networks control systems reinforcement learning crowd navigation |
| title | Robot Navigation in Crowded Environments: A Reinforcement Learning Approach |
| title_full | Robot Navigation in Crowded Environments: A Reinforcement Learning Approach |
| title_fullStr | Robot Navigation in Crowded Environments: A Reinforcement Learning Approach |
| title_full_unstemmed | Robot Navigation in Crowded Environments: A Reinforcement Learning Approach |
| title_short | Robot Navigation in Crowded Environments: A Reinforcement Learning Approach |
| title_sort | robot navigation in crowded environments a reinforcement learning approach |
| topic | mobile robotics neural networks control systems reinforcement learning crowd navigation |
| url | https://www.mdpi.com/2075-1702/11/2/268 |
| work_keys_str_mv | AT matteocaruso robotnavigationincrowdedenvironmentsareinforcementlearningapproach AT enricoregolin robotnavigationincrowdedenvironmentsareinforcementlearningapproach AT federicojuliancamerotaverdu robotnavigationincrowdedenvironmentsareinforcementlearningapproach AT stefanoalbertorusso robotnavigationincrowdedenvironmentsareinforcementlearningapproach AT lucabortolussi robotnavigationincrowdedenvironmentsareinforcementlearningapproach AT stefanoseriani robotnavigationincrowdedenvironmentsareinforcementlearningapproach |