Tone Reservation-Based PAPR Reduction Using Manifold Optimization for OFDM-ISAC Systems

Getuar Rexhepi,Kuranage Roche Rayan Ranasinghe,Giuseppe Thadeu Freitas de Abreu,David González G

2024-09-24

Abstract:We consider the peak-to-average power ratio (PAPR) reduction challenge of orthogonal frequency division multiplexing (OFDM) systems utilizing tone reservation (TR) under a sensing-enabling constraint, such that the signals placed in the reserved tones (RTs) can be exploited for Integrated Sensing and Communication (ISAC). To that end, the problem is first cast as an unconstrained manifold optimization problem, and then solved via an iterative projected gradient descent algorithm assisted by an approximation of the infinity norm. Simulation results show that the proposed method, while maintaining a level of PAPR reduction similar to state of the art (SotA), not only has lower computational complexity but also outperforms the alternatives in terms of sensing performance.

Signal Processing

What problem does this paper attempt to address?

The problem that this paper attempts to solve is: in an Orthogonal Frequency - Division Multiplexing (OFDM) system, how to effectively reduce the Peak - to - Average Power Ratio (PAPR) while retaining the sensing function. Specifically, the paper proposes a Tone Reservation (TR) method based on manifold optimization to achieve PAPR reduction and ensure that the reserved tones can be used for sensing tasks in the Integrated Sensing and Communication (ISAC) system.

### Problem Background
The OFDM system has advantages such as resistance to multipath fading and reduction of Inter - Symbol Interference (ISI), but its high PAPR is a significant disadvantage. High PAPR will lead to a decrease in the Signal - to - Quantization - Noise Ratio (SQNR) and reduce the efficiency of the power amplifier at the transmitter. Therefore, reducing PAPR is one of the key issues in improving the performance of the OFDM system.

### Solution
The solutions proposed in the paper include the following aspects:

1. **Problem Modeling**:
- Transform the PAPR minimization problem into an unconstrained manifold optimization problem.
- Introduce complex signals on the unit circle as reserved tones to ensure that these tones can be used for sensing tasks.

2. **Optimization Algorithm**:
- Use the Projected Gradient Descent algorithm to solve the optimization problem.
- By approximating the infinity norm (\(\|x\|_\infty\)) and calculating the gradient, effective solution of the optimization problem is achieved.

3. **Performance Evaluation**:
- Verify the performance of the proposed method in terms of PAPR reduction through simulation. The results show that this method not only has low computational complexity but also outperforms existing methods in sensing performance.
- Compared with the optimal Quadratically Constrained Quadratic Programming (QCQP) method, the proposed method performs equally well or better in terms of PAPR reduction and sensing performance.

### Key Formulas
- PAPR Definition:
\[
\text{PAPR}(x) \triangleq \frac{\|x\|_\infty^2}{\frac{1}{N} \|x\|_2^2}
\]
where \(\|x\|_\infty\) and \(\|x\|_2\) represent the \(\ell_\infty\) norm and \(\ell_2\) norm of vector \(x\), respectively.

- Optimization Problem:
\[
\minimize_{r \in \mathbb{C}^{N_r \times 1}} \|d + F_H^R r\|_\infty^2,
\]
subject to:
\[
|r_i| = 1, \quad i = 0, \ldots, N_r - 1.
\]

- Gradient Calculation:
\[
\frac{\partial}{\partial i} \|x\|_p = \left( \frac{|x_i|^{p - 2}x_i}{\|x\|_p^{p - 1}} \right),
\]
where \(x = d + F_H^R r\).

### Summary
By introducing the manifold optimization framework and the Projected Gradient Descent algorithm, the paper successfully solves the problem of reducing the PAPR of the OFDM system while retaining the sensing function. The experimental results show that this method not only performs well in terms of PAPR reduction but also has obvious advantages in computational complexity and sensing performance.