Journal
IEEE COMMUNICATIONS LETTERS
Volume 26, Issue 3, Pages 701-705Publisher
IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC
DOI: 10.1109/LCOMM.2021.3137809
Keywords
Switches; Reinforcement learning; Prediction algorithms; Optimization; Load modeling; Delays; Data models; Spectrum handoff; hybrid priority queuing model; DQN; reinforcement learning; transfer learning
Categories
Funding
- Natural Science Foundation of Shanghai [19ZR1455200]
Ask authors/readers for more resources
In this paper, a new hybrid priority queuing model is proposed and a deep Q-network-based algorithm is designed to optimize the latency performance of spectrum handoffs in cognitive radio networks. The transfer learning method is introduced to accelerate the learning process, and simulation results show that the proposed method outperforms conventional reinforcement learning-based approaches in terms of latency performance.
In cognitive radio networks (CRNs), spectrum handoff techniques help the interrupted secondary user (SU) vacate the licensed channel and seek for another suitable channel to resume its unfinished transmission. However, multiple interruptions from primary users and various latency requirements impose enormous challenges to spectrum handoffs. To this end, we propose a new hybrid priority queuing model for predictive spectrum handoffs and derive the closed-form expression for the extended data delivery time (latency performance), and then a deep Q-network (DQN)-based algorithm is designed to minimize the transmission latency for SUs. Furthermore, the transfer learning method is also introduced in our spectrum handoff algorithm to accelerate the learning process in which a newly added SU can obtain the initial loss function from its nearest neighbor. Simulation results show that the proposed spectrum handoff method outperforms the conventional approaches based on reinforcement learning in terms of the latency performance.
Authors
I am an author on this paper
Click your name to claim this paper and add it to your profile.
Reviews
Recommended
No Data Available