A Heuristic Deep Q Learning for Offloading in Edge Devices in 5 g Networks

The 5G Wireless Environments have huge data transmission; therefore, there is an increase in the requests for computational tasks from Intelligent Wireless Mobile Nodes. This computational capability leads to high reliability and low latency in a 5G network. Mobile edge computing (MEC) allows end sy...

Celý popis

Uloženo v:
Podrobná bibliografie
Vydáno v:Journal of grid computing Ročník 21; číslo 3; s. 37
Hlavní autoři: Dong, YanRu, Alwakeel, Ahmed M., Alwakeel, Mohammed M., Alharbi, Lubna A., Althubiti, Sara A
Médium: Journal Article
Jazyk:angličtina
Vydáno: Dordrecht Springer Netherlands 01.09.2023
Springer Nature B.V
Témata:
5G mobile communication
Algorithms
Computation offloading
Computer Science
Data transmission
Deep learning
Edge computing
Energy consumption
Machine learning
Management of Computing and Information Systems
Mobile computing
Multiagent systems
Network latency
Nodes
Processor Architectures
Task scheduling
User Interfaces and Human Computer Interaction
Wireless networks
Partial offloading
Double deep Q network
Mobile edge computing
Karush-Kuhn-Tucker
Wireless node
ISSN:1570-7873, 1572-9184
On-line přístup:Získat plný text
Tagy: Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
Popis
Shrnutí:The 5G Wireless Environments have huge data transmission; therefore, there is an increase in the requests for computational tasks from Intelligent Wireless Mobile Nodes. This computational capability leads to high reliability and low latency in a 5G network. Mobile edge computing (MEC) allows end systems with constrained computing capacity to handle computationally demanding tasks and offer accurate alternatives. The MEC server’s physical position is nearer to WN than other servers, which satisfies the demands for low latency and excellent dependability. To overcome the issues of existing work, such as low latency, offloading and task scheduling, the proposed method provides efficient results. In this work for job scheduling, Multi-agent Collaborative Deep Reinforcement Learning based Scheduling Algorithm with a double deep Q network (DQN) is used in the MEC system. To minimize the total Latency in Wireless Nodes, it uses Karush-Kuhn-Tucker (KKT) approach. This approach provides the optimum solutions to the partial and complete offloading tasks. The double deep Q network (DQN) reduces energy consumption and offers better convergence Between the Wireless Nodes. Compared to traditional algorithms like DeMDRL and BiDRL, the proposed MDRL-DDQN demonstrates superior performance.
Bibliografie:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:1570-7873
1572-9184
DOI:10.1007/s10723-023-09667-w