CSI-Free Position Optimization for Movable Antenna Communication Systems: A Black-Box Optimization Approach

Xianlong Zeng,Jun Fang,Bin Wang,Boyu Ning,Hongbin Li

2024-08-09

Abstract:Movable antenna (MA) is a new technology which leverages local movement of antennas to improve channel qualities and enhance the communication performance. Nevertheless, to fully realize the potential of MA systems, complete channel state information (CSI) between the transmitter-MA and the receiver-MA is required, which involves estimating a large number of channel parameters and incurs an excessive amount of training overhead. To address this challenge, in this paper, we propose a CSI-free MA position optimization method. The basic idea is to treat position optimization as a black-box optimization problem and calculate the gradient of the unknown objective function using zeroth-order (ZO) gradient approximation techniques. Simulation results show that the proposed ZO-based method, through adaptively adjusting the position of the MA, can achieve a favorable signal-to-noise-ratio (SNR) using a smaller number of position measurements than the CSI-based approach. Such a merit makes the proposed algorithm more adaptable to fast-changing propagation channels.

Signal Processing

What problem does this paper attempt to address?

The problem that this paper attempts to solve is: in the Movable Antenna (MA) communication system, how to optimize the position of MA without complete Channel State Information (CSI). Specifically, traditional CSI methods require a large amount of channel parameter estimation and training overhead, which is especially impractical in rapidly changing propagation environments. Therefore, this paper proposes a black - box optimization method based on Zeroth - Order (ZO) gradient approximation technology to optimize the position of MA, thereby reducing the number of required channel measurements and improving the signal reception quality.

### Problem Background

1. **Advantages of Movable Antennas**
- Movable antennas can move within a specified area through mechanical controllers and actuators to obtain better channel quality.
- This additional spatial degree of freedom can significantly improve communication performance, especially in Multiple - Input Multiple - Output (MIMO) systems.

2. **Challenges of Traditional CSI Methods**
- In order to achieve the optimal MA position, it is necessary to obtain the complete channel response of the entire movable area between the transmitter and the receiver.
- This involves a large amount of channel parameter estimation and training overhead, especially in rich - scattering environments where the number of channel parameters can be very large.

### Paper Solutions

1. **CSI - free Position Optimization Method**
- Consider the position optimization problem as a black - box optimization problem.
- Use the zeroth - order gradient approximation technique to estimate the gradient of the objective function without explicitly obtaining the channel parameters.
- Specific steps include:
- Calculate a new position based on the received signal.
- Move the MA to the new position for the next measurement.
- Repeat the above process until convergence.

2. **Algorithm Advantages**
- **High Sample Efficiency**: Compared with traditional CSI methods, this method only requires a smaller number of channel measurements to find a near - optimal MA position.
- **Strong Robustness to Noise**: Since it does not need to explicitly obtain channel parameters, this method has higher robustness to noise in measurements.

### Models and Formulas

1. **System Model**
- Consider a point - to - point narrow - band communication system, where the receiving end uses a single movable antenna and the transmitting end uses an antenna at a fixed position.
- The received signal expression is:
\[
y(r)=\sqrt{P}h(r)s + n
\]
where \(P\) is the transmit power, \(h(r)\) is the channel response, \(s\) is the transmit signal, and \(n\) is the additive white Gaussian noise.

2. **Channel Response Model**
- The channel response \(h(r)\) can be expressed as the superposition of multipath components:
\[
h(r)=\sum_{l = 1}^{L}b_l e^{-j\frac{2\pi}{\lambda}\rho_l^r(x_r,y_r)}
\]
where \(b_l\) is the complex coefficient of the \(l\) - th path, and \(\rho_l^r(x_r,y_r)\) is the signal propagation distance change of the \(l\) - th path.

3. **Optimization Objective**
- Maximize the signal - to - noise ratio (SNR) at the receiving end:
\[
\gamma(r)=\frac{|h(r)|^2P}{\sigma^2}
\]
where \(\sigma^2\) is the noise variance.

4. **Zeroth - Order Gradient Approximation**
- Use the central difference method to estimate the gradient:
\[
\hat{\nabla}f(x)=\frac{d}{2\mu}[f(x+\mu u)-f(x-\mu u)]u
\]
where \(u\) is a random unit vector and \(\mu\) is a small step size.

### Experimental Results