Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Model-Driven Sensing-Node Selection and Power Allocation for Tracking Maneuvering Targets in Perceptive Mobile Networks

Lei Xie, Hengtao He, Shenghui Song, Yonina C. Eldar

TL;DR

This work addresses joint SN selection and power allocation for tracking maneuvering targets in PMNs by framing it as PCRLB minimization. It advances a model-driven solution that unfolds an MM-ADMM SN-selection algorithm into a Deep Alternating Network (DAN) with momentum-based acceleration, together with a fixed-point water-filling power allocator within an alternating-optimization framework. The proposed method achieves near-exhaustive performance with substantially reduced computational cost, and provides convergence guarantees for the neural unfolded architecture. The practical impact lies in enabling real-time, high-accuracy sensing in dense cellular networks where latency and scalability are critical.

Abstract

Maneuvering target tracking will be an important service of future wireless networks to assist innovative applications such as intelligent transportation. However, tracking maneuvering targets by cellular networks faces many challenges. For example, the dense network and high-speed targets make the selection of the sensing nodes (SNs) and the associated power allocation very challenging. Existing methods demonstrated engaging performance, but with high computational complexity. In this paper, we propose a model-driven deep learning (DL)-based approach for SN selection. To this end, we first propose an iterative SN selection method by jointly exploiting the majorization-minimization (MM) framework and the alternating direction method of multipliers (ADMM). Then, we unfold the iterative algorithm as a deep neural network and prove its convergence. The proposed method achieves lower computational complexity, because the number of layers is less than the number of iterations required by the original algorithm, and each layer only involves simple matrix-vector additions/multiplications. Finally, we propose an efficient power allocation method based on fixed point (FP) water filling and solve the joint SN selection and power allocation problem under the alternative optimization framework. Simulation results show that the proposed method achieves better performance than the conventional optimization-based methods with much lower computational complexity.

Model-Driven Sensing-Node Selection and Power Allocation for Tracking Maneuvering Targets in Perceptive Mobile Networks

TL;DR

This work addresses joint SN selection and power allocation for tracking maneuvering targets in PMNs by framing it as PCRLB minimization. It advances a model-driven solution that unfolds an MM-ADMM SN-selection algorithm into a Deep Alternating Network (DAN) with momentum-based acceleration, together with a fixed-point water-filling power allocator within an alternating-optimization framework. The proposed method achieves near-exhaustive performance with substantially reduced computational cost, and provides convergence guarantees for the neural unfolded architecture. The practical impact lies in enabling real-time, high-accuracy sensing in dense cellular networks where latency and scalability are critical.

Abstract

Maneuvering target tracking will be an important service of future wireless networks to assist innovative applications such as intelligent transportation. However, tracking maneuvering targets by cellular networks faces many challenges. For example, the dense network and high-speed targets make the selection of the sensing nodes (SNs) and the associated power allocation very challenging. Existing methods demonstrated engaging performance, but with high computational complexity. In this paper, we propose a model-driven deep learning (DL)-based approach for SN selection. To this end, we first propose an iterative SN selection method by jointly exploiting the majorization-minimization (MM) framework and the alternating direction method of multipliers (ADMM). Then, we unfold the iterative algorithm as a deep neural network and prove its convergence. The proposed method achieves lower computational complexity, because the number of layers is less than the number of iterations required by the original algorithm, and each layer only involves simple matrix-vector additions/multiplications. Finally, we propose an efficient power allocation method based on fixed point (FP) water filling and solve the joint SN selection and power allocation problem under the alternative optimization framework. Simulation results show that the proposed method achieves better performance than the conventional optimization-based methods with much lower computational complexity.
Paper Structure (26 sections, 1 theorem, 73 equations, 8 figures, 3 tables)

This paper contains 26 sections, 1 theorem, 73 equations, 8 figures, 3 tables.

Table of Contents

  1. Introduction
  2. System Model and Problem Formulation
  3. Target Motion Model
  4. Signal Model
  5. Posterior Cramér-Rao Lower Bound
  6. Problem Formulation
  7. Model-Driven DL-based Sensing Node Selection and Power Allocation Scheme
  8. MM-ADMM based Sensing Node Selection
  9. MM framework for solving (\ref{['prob01u']})
  10. ADMM-based method for solving (\ref{['prob1Gu']})
  11. Deep-Alternative-Network: DL-Based Sensing Node Selection
  12. Convergence of DAN
  13. Transmit Power Allocation for Multiple Targets
  14. Simulation
  15. Computational Cost
  16. ...and 11 more sections

Key Result

Theorem 1

Assume that, for all $l\in[2,L]$, $\phi_{l,i}^{-1} \leq \phi_{l-1,i}^{-1}$. The regret is bounded by where $C_1 = \frac{\sqrt{1-\beta_{2}}D_{u,1} D_{\Delta}}{\alpha_{1}^{-}(1-\sqrt{\beta_2})(1-\beta_{1})}$ and $C_2$ is defined by (regbound), given at the top of this page.

Figures (8)

  • Figure 1: Illustration of the system.
  • Figure 2: Illustration of DAN.
  • Figure 3: Cost function over the number of iterations/layers.
  • Figure 4: Average RMSE versus the total power budget $P$.
  • Figure 5: SN selection result by DAN at 4 consecutive frames. (a) Frame 2; (b) Frame 4; (c) Frame 6; (d) Frame 8;
  • ...and 3 more figures

Theorems & Definitions (4)

  • Remark 1
  • Remark 2
  • Theorem 1
  • Remark 3