In this article, we consider a practical scenario for secure wireless-powered communication in the presence of imperfect channel state information (CSI) with simultaneous energy harvesting, in which it is required to keep information secret from an untrusted energy receiver allowed only to harvest energy from the transmitted signals. We aim to find the robust transmit power control (TPC) strategy to maximize the secrecy rate whilst ensuring the spectral efficiency of transceiver pairs and the amount of energy harvested by the energy receiver, even when the CSI is inaccurate. To deal with the nonconvexity of the formulated optimization problem, we first derive a suboptimal form of TPC in an iterative manner by adopting dual methods. In order to overcome the drawbacks of the conventional optimization-based approach regarding the suboptimality of performance and requiring long computation time, we devise a deep learning (DL)-assisted TPC as an alternative means of deriving the TPC. In the considered DL-assisted TPC, a deep neural network (DNN) is trained to compensate for the distortion caused by channel errors in an unsupervised manner. More specifically, artificially distorted CSI, which reflects the difference between actual and estimated CSI, is fed into the DNN during training and used to update the weights and biases of the proposed DNN using a bounded loss function, which allows a robust TPC strategy to be approximated by the DNN. Simulation results reveal the robustness of the proposed DL-assisted TPC against channel errors, such that it achieves a near-optimal performance with a lower computation time, even when the CSI is incorrect.
|Number of pages||13|
|Journal||IEEE Internet of Things Journal|
|Publication status||Published - 2022 Jul 1|
Bibliographical noteFunding Information:
This work was supported in part by the Singapore University of Technology and Design (SUTD) Growth Plan Grant for AI, and in part by the National Research Foundation of Korea (NRF) grant funded by the Korea Government (MSIT) under Grant 2021R1A2C4002024 and Grant 2021R1F1A1046932.
© 2014 IEEE.
All Science Journal Classification (ASJC) codes
- Signal Processing
- Information Systems
- Hardware and Architecture
- Computer Science Applications
- Computer Networks and Communications