Skip to main content
AkademIndex

Products

For developers

AkademBasesoonOpen API for the ecosystem
Latin
Article

Deep Augmentation In 5G Scheduling Design Acquiring Knowledge from Concept to Application

Ammar Hameed ShnainThe Islamic University of Al Diwaniyah,College of Technical Engineering,Department of Computers Techniques Engineering,Al Diwaniyah,IraqHeera PatwalGraphic Era Hill University,Department of Computer Science & Engineering,Bhimtal,India,263132Urinbaev SharofiddinTashkent Institute of Irrigation and Agricultural Mechanization Engineers" National Research University,Department of Physics and Chemistry,Tashkent,UzbekistanAmarjeet KaurChandigarh Engineering College, Chandigarh Group of Colleges,Department of Electrical and Communication Engineering,Mohali,Punjab,India,140307Manish GuptaGLA University,Department of Electronics and Communication Engineering,MathuraNiharika VarshneyGraphic Era Deemed to be University,Department of Electrical Engineering,Dehradun,India
2024en
ABI

Abstract

In this paper, we presented a knowledge base assisted deep reinforcement learning (DRL) solution for wireless programming of time-sensitive delivery in 5G cellular networks. Because the deep reliable policy grades mapping from queue and channel states (similar to a Q-learning) to scheduling actions, they may be used as an approximation of good schedule policies with parameterized weights. Unfortunately, the great is Thankfully a modest DD process that comes together with extreme consider for quality-of-service and performs as poorly when targeting this same unpredictable live 5G mobile networks it was supposed to run on. We propose an abstract directional light model to solve these problems, and then present a DRL decision algorithm built on the theoretical framework of wireless communications. We also offer an e-induction and training framework, that after being initialised offline with K-DDPG allows scheduler to be adapted in real-time due the imbalance between off-line (calculation) rewards agreed upon start-up point vs. reality anomaly present during a live system operating time (Utilization of High-Fix action). Simulation results demonstrate that our technique outperforms state-of-the-art schedules in service quality by orders of magnitude and drastically reduces the DDPG iteration time. Our method also reaches faster (in minutes of online fine-tuning consensus) the same higher initial QoS as reported in Table I, for off-line starting than random initialization [16].

Topics

Identifiers

Citations and references

Cited by 026 references
Metrics — AkademScholar · Coming soon