Junichiro Hagiwara,Toshihiko Nishimura,Takanori Sato,Yasutaka Ogawa,Takeo Ohgane
Abstract:Multiple-input multiple-output (MIMO) technology is essential for the optimal functioning of next-generation wireless networks; however, enhancing its signal-detection performance for improved spectral efficiency is challenging. Here, we propose an approach that transforms the discrete MIMO detection problem into a continuous problem while leveraging the efficient Hamiltonian Monte Carlo algorithm. For this continuous framework, we employ a mixture of t-distributions as the prior distribution. To improve the performance in the coded case further, we treat the likelihood's temperature parameter as a random variable and address its optimization. This treatment leads to the adoption of a horseshoe density for the likelihood. Theoretical analysis and extensive simulations demonstrate that our method achieves near-optimal detection performance while maintaining polynomial computational complexity. This MIMO detection technique can accelerate the development of 6G mobile communication systems.
What problem does this paper attempt to address?
### Problems the paper attempts to solve
The paper aims to solve the problem of improving the signal detection performance in multi - input multi - output (MIMO) systems. Specifically, the paper proposes a new method that transforms the discrete MIMO signal detection problem into a continuous one and uses the efficient Hamiltonian Monte Carlo (HMC) algorithm to solve it. Through this method, the paper hopes to achieve near - optimal signal detection performance while maintaining polynomial computational complexity.
### Main contributions
1. **Problem transformation**: Transform the discrete MIMO signal detection problem into a continuous problem so as to utilize efficient algorithms in the continuous domain.
2. **Prior distribution selection**: Adopt the mixture t - distribution as the prior distribution to improve the exploration efficiency of the transmitted signal points.
3. **Temperature parameter optimization**: Consider the temperature parameter of the likelihood function as a random variable and improve the detection performance in the coded case through automatic optimization.
4. **Theoretical analysis and simulation**: Verify the effectiveness of the proposed method through theoretical analysis and a large number of simulation experiments, and prove that it can achieve near - optimal detection performance while the computational complexity is controllable.
5. **Application in 6G communication systems**: This method is expected to accelerate the development of 6G mobile communication systems and improve the spectral efficiency of wireless transmission.
### Method overview
1. **System model**:
- Consider a full - flow MIMO configuration, which contains \( N \) transmit antennas and \( M \) receive antennas.
- The system relationship is described by the following equation:
\[
y = Hu + w, \quad w \sim \mathcal{CN}(0, \sigma_w^2 I)
\]
where \( y \) is the received symbol vector, \( H \) is the channel matrix, \( u \) is the transmitted symbol vector, \( w \) is the noise vector, and \(\sigma_w^2\) is the noise variance.
2. **Posterior distribution**:
- Using Bayes' theorem, the posterior distribution can be expressed as:
\[
p(u|y) \propto p(y|u) p(u)
\]
- where \( p(y|u) \) is the likelihood function and \( p(u) \) is the prior distribution.
3. **Likelihood function**:
- The likelihood function is defined as:
\[
p(y|u) = \mathcal{N}(y; Hu, \sigma_w^2 I)
\]
4. **Prior distribution**:
- Adopt the mixture t - distribution as the prior distribution:
\[
p_1(u) = \prod_{n = 1}^{2N} \frac{1}{q} \left( T(u_n; a_1, \sigma, \nu)+\cdots+T(u_n; a_q, \sigma, \nu) \right)
\]
- where \( T \) represents the t - distribution density, and \( \sigma \) and \( \nu \) are adjustable parameters.
5. **Numerical approximation of the posterior distribution**:
- Use the HMC algorithm to perform the numerical approximation of the posterior distribution. The HMC algorithm solves the local optimum problem and improves the sampling efficiency by introducing the auxiliary momentum variable \( r \) and the virtual time \( \tau \).
6. **Extension in the coded case**:
- In the coded case, consider that the temperature parameter \( \sigma_w \) of the likelihood function is multiplied by a positive coefficient \( \lambda_n \):
\[
\sigma_w\rightarrow\lambda_n\sigma_w
\]
- Consider \( \lambda_n \) as an independent random variable and assume that it follows the Cauchy distribution \(\lambda_n\sim C^+(0, \sigma_{\text{Cauchy}})\)