Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Teacher-Student Learning based Low Complexity Relay Selection in Wireless Powered Communications

Aysun Gurur Onalan, Berkay Kopru, Sinem Coleri

TL;DR

This work tackles joint relay selection, scheduling, and power control in multi-source-multi-relay Wireless Powered Communication Networks under nonlinear energy harvesting. It develops a two-step framework: NL-POWMU for optimal scheduling/power given relay choices, and CNN-based relay selection with two architectures (SC-NET and SKIN-NET) to reduce online complexity. To further cut runtime without sacrificing performance, it introduces teacher-student learning (STU-SC-NET) guided by a novel architecture-search mechanism (DASA) and a distillation loss combining hard labels and teacher soft outputs. Empirical results show significant runtime reductions and competitive near-optimal performance compared to state-of-the-art iterative methods, with SKIN-NET offering memory efficiency and STU-SC-NET delivering the best runtime. The approach enables scalable, real-time relay management in WPCNs, and points to future enhancements via reinforcement learning and active data selection to reduce offline training costs.

Abstract

Radio Frequency Energy Harvesting (RF-EH) networks are key enablers of massive Internet-of-things by providing controllable and long-distance energy transfer to energy-limited devices. Relays, helping either energy or information transfer, have been demonstrated to significantly improve the performance of these networks. This paper studies the joint relay selection, scheduling, and power control problem in multiple-source-multiple-relay RF-EH networks under nonlinear EH conditions. We first obtain the optimal solution to the scheduling and power control problem for the given relay selection. Then, the relay selection problem is formulated as a classification problem, for which two convolutional neural network (CNN) based architectures are proposed. While the first architecture employs conventional 2D convolution blocks and benefits from skip connections between layers; the second architecture replaces them with inception blocks, to decrease trainable parameter size without sacrificing accuracy for memory-constrained applications. To decrease the runtime complexity further, teacher-student learning is employed such that the teacher network is larger, and the student is a smaller size CNN-based architecture distilling the teacher's knowledge. A novel dichotomous search-based algorithm is employed to determine the best architecture for the student network. Our simulation results demonstrate that the proposed solutions provide lower complexity than the state-of-art iterative approaches without compromising optimality.

Teacher-Student Learning based Low Complexity Relay Selection in Wireless Powered Communications

TL;DR

This work tackles joint relay selection, scheduling, and power control in multi-source-multi-relay Wireless Powered Communication Networks under nonlinear energy harvesting. It develops a two-step framework: NL-POWMU for optimal scheduling/power given relay choices, and CNN-based relay selection with two architectures (SC-NET and SKIN-NET) to reduce online complexity. To further cut runtime without sacrificing performance, it introduces teacher-student learning (STU-SC-NET) guided by a novel architecture-search mechanism (DASA) and a distillation loss combining hard labels and teacher soft outputs. Empirical results show significant runtime reductions and competitive near-optimal performance compared to state-of-the-art iterative methods, with SKIN-NET offering memory efficiency and STU-SC-NET delivering the best runtime. The approach enables scalable, real-time relay management in WPCNs, and points to future enhancements via reinforcement learning and active data selection to reduce offline training costs.

Abstract

Radio Frequency Energy Harvesting (RF-EH) networks are key enablers of massive Internet-of-things by providing controllable and long-distance energy transfer to energy-limited devices. Relays, helping either energy or information transfer, have been demonstrated to significantly improve the performance of these networks. This paper studies the joint relay selection, scheduling, and power control problem in multiple-source-multiple-relay RF-EH networks under nonlinear EH conditions. We first obtain the optimal solution to the scheduling and power control problem for the given relay selection. Then, the relay selection problem is formulated as a classification problem, for which two convolutional neural network (CNN) based architectures are proposed. While the first architecture employs conventional 2D convolution blocks and benefits from skip connections between layers; the second architecture replaces them with inception blocks, to decrease trainable parameter size without sacrificing accuracy for memory-constrained applications. To decrease the runtime complexity further, teacher-student learning is employed such that the teacher network is larger, and the student is a smaller size CNN-based architecture distilling the teacher's knowledge. A novel dichotomous search-based algorithm is employed to determine the best architecture for the student network. Our simulation results demonstrate that the proposed solutions provide lower complexity than the state-of-art iterative approaches without compromising optimality.
Paper Structure (17 sections, 2 theorems, 17 equations, 6 figures, 3 tables, 3 algorithms)

This paper contains 17 sections, 2 theorems, 17 equations, 6 figures, 3 tables, 3 algorithms.

Table of Contents

  1. Introduction
  2. System Model
  3. Channel Model
  4. Tranmission Model
  5. Downlink Energy Harvesting
  6. Uplink Information Transfer
  7. Problem Formulation and Solution Strategy
  8. Scheduling and Power Control under Non-Linear EH
  9. Teacher-Student Learning based Relay Selection Framework
  10. SC-NET Architecture
  11. SKIN-NET Architecture
  12. Teacher-Student Learning
  13. Performance Evaluation
  14. Data Generation
  15. Deep Learning Performance
  16. ...and 2 more sections

Key Result

Lemma 1

Let $\overline{\tau_{S_1}}(\overline{\tau_0})$ be the optimal solution to Problem (eq:subproblm1APNS) for $\tau_0 = \overline{\tau_0}$. The optimal IT duration $\overline{\tau_{S_1}}$ is expressed by if Otherwise, $\overline{\tau_{S_1}}(\overline{\tau_0})$ is derived as a solution of the nonlinear equation where

Figures (6)

  • Figure 1: Time Frame
  • Figure 2: a) Training and b) Prediction processes of deep learning model
  • Figure 3: a) Proposed Network Architecture with convolution blocks b) Proposed Network Architecture with inception blocks c) Details of inception block
  • Figure 4: Cross Entropy Loss vs. a) number of layers b) number of trainable parameters in a 3-source-2-relay network.
  • Figure 5: a) Training and b) Validation Losses throughout the iterations of DASA given in Algorithm \ref{['alg:search_arch']} in a 3-source-2-relay network.
  • ...and 1 more figures

Theorems & Definitions (4)

  • Lemma 1
  • proof
  • Lemma 2
  • proof