Memory-driven deep-reinforcement learning for autonomous robot navigation in partially observable environments
Service robots with autonomous navigational capabilities play a critical role in dynamic contexts where safe and collision-free human interactions are important. However, the unpredictable nature of human behavior, the prevalence of occlusions and the lack of complete environmental perception due to...
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Elsevier
2025-02-01
|
| Series: | Engineering Science and Technology, an International Journal |
| Subjects: | |
| Online Access: | http://www.sciencedirect.com/science/article/pii/S2215098624003288 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1832087626246520832 |
|---|---|
| author | Estrella Montero Nabih Pico Mitra Ghergherehchi Ho Seung Song |
| author_facet | Estrella Montero Nabih Pico Mitra Ghergherehchi Ho Seung Song |
| author_sort | Estrella Montero |
| collection | DOAJ |
| description | Service robots with autonomous navigational capabilities play a critical role in dynamic contexts where safe and collision-free human interactions are important. However, the unpredictable nature of human behavior, the prevalence of occlusions and the lack of complete environmental perception due to sensor limitations can severely restrict effective robot navigation. We propose a memory-driven algorithm that employs deep reinforcement learning to enable collision-free proactive navigation in partially observable environments. The proposed method takes the relative states of humans within a limited FoV and sensor range as input into the neural network. The model employs a bidirectional gated recurrent unit as a temporal function to strategically incorporate the previous context of input sequences and facilitate the assimilation of the observations. This approach allows the model to assign greater attention to intricate human–robot relations, allowing a better understanding of the ever-changing dynamics within an environment. Simulations and experimental outcomes validate the efficacy of the policy-based navigation approach. It achieves superior collision avoidance performance compared to representative existing methods and exhibits efficient navigation by incorporating the limitations of sensors during training. |
| format | Article |
| id | doaj-art-ee5979ccd5404a969b64a3172f1e407f |
| institution | Kabale University |
| issn | 2215-0986 |
| language | English |
| publishDate | 2025-02-01 |
| publisher | Elsevier |
| record_format | Article |
| series | Engineering Science and Technology, an International Journal |
| spelling | doaj-art-ee5979ccd5404a969b64a3172f1e407f2025-02-06T05:11:51ZengElsevierEngineering Science and Technology, an International Journal2215-09862025-02-0162101942Memory-driven deep-reinforcement learning for autonomous robot navigation in partially observable environmentsEstrella Montero0Nabih Pico1Mitra Ghergherehchi2Ho Seung Song3Sungkyunkwan University, Department of Electrical and Computer Engineering, Natural Sciences Campus, Suwon 16419, South KoreaSungkyunkwan University, Department of Mechanical Engineering, Natural Sciences Campus, Suwon, South Korea; Escuela Superior Politecnica del Litoral, ESPOL, Facultad de Ingenieria en Electricidad y Computacion, Guayaquil 09-01-5863, EcuadorSungkyunkwan University, Department of Electrical and Computer Engineering, Natural Sciences Campus, Suwon 16419, South Korea; Corresponding authors.Catholic Kwandong University, Department of Electronic Engineering, Gangwon-do, South Korea; Corresponding authors.Service robots with autonomous navigational capabilities play a critical role in dynamic contexts where safe and collision-free human interactions are important. However, the unpredictable nature of human behavior, the prevalence of occlusions and the lack of complete environmental perception due to sensor limitations can severely restrict effective robot navigation. We propose a memory-driven algorithm that employs deep reinforcement learning to enable collision-free proactive navigation in partially observable environments. The proposed method takes the relative states of humans within a limited FoV and sensor range as input into the neural network. The model employs a bidirectional gated recurrent unit as a temporal function to strategically incorporate the previous context of input sequences and facilitate the assimilation of the observations. This approach allows the model to assign greater attention to intricate human–robot relations, allowing a better understanding of the ever-changing dynamics within an environment. Simulations and experimental outcomes validate the efficacy of the policy-based navigation approach. It achieves superior collision avoidance performance compared to representative existing methods and exhibits efficient navigation by incorporating the limitations of sensors during training.http://www.sciencedirect.com/science/article/pii/S2215098624003288Autonomous robotsReinforcement learningCollision avoidanceHuman–robot interaction |
| spellingShingle | Estrella Montero Nabih Pico Mitra Ghergherehchi Ho Seung Song Memory-driven deep-reinforcement learning for autonomous robot navigation in partially observable environments Engineering Science and Technology, an International Journal Autonomous robots Reinforcement learning Collision avoidance Human–robot interaction |
| title | Memory-driven deep-reinforcement learning for autonomous robot navigation in partially observable environments |
| title_full | Memory-driven deep-reinforcement learning for autonomous robot navigation in partially observable environments |
| title_fullStr | Memory-driven deep-reinforcement learning for autonomous robot navigation in partially observable environments |
| title_full_unstemmed | Memory-driven deep-reinforcement learning for autonomous robot navigation in partially observable environments |
| title_short | Memory-driven deep-reinforcement learning for autonomous robot navigation in partially observable environments |
| title_sort | memory driven deep reinforcement learning for autonomous robot navigation in partially observable environments |
| topic | Autonomous robots Reinforcement learning Collision avoidance Human–robot interaction |
| url | http://www.sciencedirect.com/science/article/pii/S2215098624003288 |
| work_keys_str_mv | AT estrellamontero memorydrivendeepreinforcementlearningforautonomousrobotnavigationinpartiallyobservableenvironments AT nabihpico memorydrivendeepreinforcementlearningforautonomousrobotnavigationinpartiallyobservableenvironments AT mitraghergherehchi memorydrivendeepreinforcementlearningforautonomousrobotnavigationinpartiallyobservableenvironments AT hoseungsong memorydrivendeepreinforcementlearningforautonomousrobotnavigationinpartiallyobservableenvironments |