Multi-Agent Meta Reinforcement Learning for Reliable and Low-Latency Distributed Inference in Resource-Constrained UAV Swarms
<p dir="ltr">The integration of unmanned aerial vehicles (UAVs) in the Industrial Internet of Things (IIoT) for smart city applications has been gaining significant attention. UAV swarms are increasingly employed to monitor ground-based IIoT devices in smart cities, offering valuable...
محفوظ في:
| المؤلف الرئيسي: | |
|---|---|
| مؤلفون آخرون: | , , , , |
| منشور في: |
2025
|
| الموضوعات: | |
| الوسوم: |
إضافة وسم
لا توجد وسوم, كن أول من يضع وسما على هذه التسجيلة!
|
| _version_ | 1864513533146824704 |
|---|---|
| author | Marwan Dhuheir (19170898) |
| author2 | Aiman Erbad (14150589) Bechir Hamdaoui (22565315) Samir Brahim Belhaouari (9427347) Mohsen Guizani (12580291) Thang X. Vu (22565318) |
| author2_role | author author author author author |
| author_facet | Marwan Dhuheir (19170898) Aiman Erbad (14150589) Bechir Hamdaoui (22565315) Samir Brahim Belhaouari (9427347) Mohsen Guizani (12580291) Thang X. Vu (22565318) |
| author_role | author |
| dc.creator.none.fl_str_mv | Marwan Dhuheir (19170898) Aiman Erbad (14150589) Bechir Hamdaoui (22565315) Samir Brahim Belhaouari (9427347) Mohsen Guizani (12580291) Thang X. Vu (22565318) |
| dc.date.none.fl_str_mv | 2025-05-19T12:00:00Z |
| dc.identifier.none.fl_str_mv | 10.1109/access.2025.3572036 |
| dc.relation.none.fl_str_mv | https://figshare.com/articles/journal_contribution/Multi-Agent_Meta_Reinforcement_Learning_for_Reliable_and_Low-Latency_Distributed_Inference_in_Resource-Constrained_UAV_Swarms/30541205 |
| dc.rights.none.fl_str_mv | CC BY 4.0 info:eu-repo/semantics/openAccess |
| dc.subject.none.fl_str_mv | Information and computing sciences Artificial intelligence Distributed computing and systems software Machine learning Meta-reinforcement learning Industrial Internet of Things distributed resource optimization UAV swarms energy harvesting Autonomous aerial vehicles Reliability Resource management Data communication Surveillance Vectors Optimization Real-time systems Energy consumption |
| dc.title.none.fl_str_mv | Multi-Agent Meta Reinforcement Learning for Reliable and Low-Latency Distributed Inference in Resource-Constrained UAV Swarms |
| dc.type.none.fl_str_mv | Text Journal contribution info:eu-repo/semantics/publishedVersion text contribution to journal |
| description | <p dir="ltr">The integration of unmanned aerial vehicles (UAVs) in the Industrial Internet of Things (IIoT) for smart city applications has been gaining significant attention. UAV swarms are increasingly employed to monitor ground-based IIoT devices in smart cities, offering valuable support to situation-awareness IoT applications, such as surveillance, traffic management, and emergency response. A key requirement in these applications is minimizing the latency of data processing, particularly for time-sensitive tasks like image classification of IIoT device data. Due to resource limitations, UAVs often rely on online task offloading to remote machines, but this can be inefficient due to unstable connections, constrained resources, and high latency. Distributed inference enabled via swarms of collaborative UAVs presents a promising solution by partitioning tasks among UAVs based on their available resources, allowing for more efficient, collaborative processing. However, the IIoT inference distribution raises challenges in ensuring reliable data transmission with minimal latency while respecting the practical UAVs’ constraints. To address these issues, we formulate the problem of CNN layer distribution and UAV trajectory planning (LDTP) as an optimization problem to improve latency, reliability, and resource usage. Given the complexity of the LDTP solution for managing online requests, we propose a real-time, lightweight solution using multi-agent meta-reinforcement learning. Our approach is tested on CNN networks and benchmarked against state-of-the-art conventional reinforcement learning algorithms. Extensive simulations show that our model outperforms competitive methods by around 29% in terms of latency and around 23% in terms of transmission power improvements while delivering results comparable to the traditional LDTP optimization solution by around 9% in terms of latency.</p><h2>Other Information</h2><p dir="ltr">Published in: IEEE Access<br>License: <a href="https://creativecommons.org/licenses/by/4.0/deed.en" target="_blank">https://creativecommons.org/licenses/by/4.0/</a><br>See article on publisher's website: <a href="https://dx.doi.org/10.1109/access.2025.3572036" target="_blank">https://dx.doi.org/10.1109/access.2025.3572036</a></p> |
| eu_rights_str_mv | openAccess |
| id | Manara2_753e4d73281060adca163ee5fc2b398f |
| identifier_str_mv | 10.1109/access.2025.3572036 |
| network_acronym_str | Manara2 |
| network_name_str | Manara2 |
| oai_identifier_str | oai:figshare.com:article/30541205 |
| publishDate | 2025 |
| repository.mail.fl_str_mv | |
| repository.name.fl_str_mv | |
| repository_id_str | |
| rights_invalid_str_mv | CC BY 4.0 |
| spelling | Multi-Agent Meta Reinforcement Learning for Reliable and Low-Latency Distributed Inference in Resource-Constrained UAV SwarmsMarwan Dhuheir (19170898)Aiman Erbad (14150589)Bechir Hamdaoui (22565315)Samir Brahim Belhaouari (9427347)Mohsen Guizani (12580291)Thang X. Vu (22565318)Information and computing sciencesArtificial intelligenceDistributed computing and systems softwareMachine learningMeta-reinforcement learningIndustrial Internet of Thingsdistributed resource optimizationUAV swarmsenergy harvestingAutonomous aerial vehiclesReliabilityResource managementData communicationSurveillanceVectorsOptimizationReal-time systemsEnergy consumption<p dir="ltr">The integration of unmanned aerial vehicles (UAVs) in the Industrial Internet of Things (IIoT) for smart city applications has been gaining significant attention. UAV swarms are increasingly employed to monitor ground-based IIoT devices in smart cities, offering valuable support to situation-awareness IoT applications, such as surveillance, traffic management, and emergency response. A key requirement in these applications is minimizing the latency of data processing, particularly for time-sensitive tasks like image classification of IIoT device data. Due to resource limitations, UAVs often rely on online task offloading to remote machines, but this can be inefficient due to unstable connections, constrained resources, and high latency. Distributed inference enabled via swarms of collaborative UAVs presents a promising solution by partitioning tasks among UAVs based on their available resources, allowing for more efficient, collaborative processing. However, the IIoT inference distribution raises challenges in ensuring reliable data transmission with minimal latency while respecting the practical UAVs’ constraints. To address these issues, we formulate the problem of CNN layer distribution and UAV trajectory planning (LDTP) as an optimization problem to improve latency, reliability, and resource usage. Given the complexity of the LDTP solution for managing online requests, we propose a real-time, lightweight solution using multi-agent meta-reinforcement learning. Our approach is tested on CNN networks and benchmarked against state-of-the-art conventional reinforcement learning algorithms. Extensive simulations show that our model outperforms competitive methods by around 29% in terms of latency and around 23% in terms of transmission power improvements while delivering results comparable to the traditional LDTP optimization solution by around 9% in terms of latency.</p><h2>Other Information</h2><p dir="ltr">Published in: IEEE Access<br>License: <a href="https://creativecommons.org/licenses/by/4.0/deed.en" target="_blank">https://creativecommons.org/licenses/by/4.0/</a><br>See article on publisher's website: <a href="https://dx.doi.org/10.1109/access.2025.3572036" target="_blank">https://dx.doi.org/10.1109/access.2025.3572036</a></p>2025-05-19T12:00:00ZTextJournal contributioninfo:eu-repo/semantics/publishedVersiontextcontribution to journal10.1109/access.2025.3572036https://figshare.com/articles/journal_contribution/Multi-Agent_Meta_Reinforcement_Learning_for_Reliable_and_Low-Latency_Distributed_Inference_in_Resource-Constrained_UAV_Swarms/30541205CC BY 4.0info:eu-repo/semantics/openAccessoai:figshare.com:article/305412052025-05-19T12:00:00Z |
| spellingShingle | Multi-Agent Meta Reinforcement Learning for Reliable and Low-Latency Distributed Inference in Resource-Constrained UAV Swarms Marwan Dhuheir (19170898) Information and computing sciences Artificial intelligence Distributed computing and systems software Machine learning Meta-reinforcement learning Industrial Internet of Things distributed resource optimization UAV swarms energy harvesting Autonomous aerial vehicles Reliability Resource management Data communication Surveillance Vectors Optimization Real-time systems Energy consumption |
| status_str | publishedVersion |
| title | Multi-Agent Meta Reinforcement Learning for Reliable and Low-Latency Distributed Inference in Resource-Constrained UAV Swarms |
| title_full | Multi-Agent Meta Reinforcement Learning for Reliable and Low-Latency Distributed Inference in Resource-Constrained UAV Swarms |
| title_fullStr | Multi-Agent Meta Reinforcement Learning for Reliable and Low-Latency Distributed Inference in Resource-Constrained UAV Swarms |
| title_full_unstemmed | Multi-Agent Meta Reinforcement Learning for Reliable and Low-Latency Distributed Inference in Resource-Constrained UAV Swarms |
| title_short | Multi-Agent Meta Reinforcement Learning for Reliable and Low-Latency Distributed Inference in Resource-Constrained UAV Swarms |
| title_sort | Multi-Agent Meta Reinforcement Learning for Reliable and Low-Latency Distributed Inference in Resource-Constrained UAV Swarms |
| topic | Information and computing sciences Artificial intelligence Distributed computing and systems software Machine learning Meta-reinforcement learning Industrial Internet of Things distributed resource optimization UAV swarms energy harvesting Autonomous aerial vehicles Reliability Resource management Data communication Surveillance Vectors Optimization Real-time systems Energy consumption |