MIMO信号检测系统量子算法的优化及运用英文小论文 第2页
This is a two-dimensional function which has only one global minimize when in the entire analysis domain. Although it is single-peak function, it is morbid and difficult to global optimization.
(2) Coldstein-Price function
Experimental control parameters set as Table 1. The test results show in Figure 1 (left for the De Jong function, right for Coldstein-Price). Experiments show that: the quantum genetic algorithm which proposed in this paper is better than the classical genetic algorithm in convergence and closer to the target in final convergence value.
Table 1 Experimental Control Parameters
Fig.1. Convergence curve of test function
3. MIMO Signal Detection Scheme based on Quantum Optimization Algorithm
3.1 MIMO signal detection system
Consider the MIMO system of transmitting antennas and receive antennas in Additive White Gaussian Noise (AWGN) channels, so the input and output relationship can be described as （6）
where y——The detection signal received by the receiving antenna. Dimension：N*1;
x——Transmission vector. Dimension：M*1;
H——Channel gain matrix. Dimension：M*N;
n——Gaussian white noise of zero mean.
The receiver's task is to detect transmitting signal from x.
3.2 MIMO signal detection scheme based on Quantum Genetic Algorithm
Fig.2. MIMO signal detection scheme based on QGA
The signal detection of MIMO system based on Quantum Genetic Algorithm (QGA) is designed as Figure 2. In the receiver, each antenna receives linear superposition signals send from different antennas and through the MIMO channel, and the Quantum Genetic Algorithm proposed in Section 2.2 as the detection algorithm in detection part. Then the information bit stream is recovered by series /parallel transition after demodulating the parallel data stream.
In order to achieve a good performance of MIMO signal detection based on QGA, we make the main parameters of the QGA algorithm design as follows:
(1)The denotation and measurement process of the Q-bit individual are same to the description before. After measuring the Q-bit will only take chromosome gene for the 0 to -1, taking gene 1 fixed.
(2)Population initialization: In simulation, the qubit gene number of quantum chromosome is equal to the number of transmitting antennas. The rest qubit gene of quantum chromosome in initialization population are initialized as .
(3)The fitness function is used to assess the stand or fall of the quality of each chromosome, and it is non-negative. Based on the maximum likelihood rule: （7）
The objective function for optimal MIMO detector is: （8）
Assumption the objective function value is maximum value when is . Because is unable to be ensured that it is nonnegative, so the fitness function for QGA based MIMO detection is designed as: （9）
Where, is 0.05 in simulation.
(4) The algorithm is terminated when the number of iterations is equal to G, which is the generation number.
To understand the detection performance of Quantum Genetic Algorithm (QGA), we will compare it with the classical Genetic Algorithm (GA) and Minimum Mean Square Error (MMSE) algorithm. Experimental conditions set as follows: The simulations were performed in a M=N=4 environment and adopt the BPSK/QPSK modulation. On the channel, adopt AWGN and power control was applied and we assumed perfect knowledge of channel parameters and the channel matrix H at T = 1000 symbols per cycle remains the same.
Algorithm control parameters as follows:A: Population number is equal to the number of transmitting antennas, the maximum genetic algebra is 20, crossover probability is 0.9, and mutation probability is 0.05;
QGA: Population number is equal to the number of transmitting antennas, the maximum genetic algebra is 20, with all cross-interference, mutation probability is 0.05, and the rotation angle of Q-gates is ;
The test result is depicted in Figure 3(left for BPSK modulation, right for QPSK modulation). Results show that: whether BPSK or QPSK modulation, QGA performance better than GA as the SNR increases.
Fig.3. Detection performance of QGA when M=N=4
3.3 MIMO signal detection scheme based on neural network optimized by QGA
The implementation of neural network algorithm is the signal from the input layer forward propagation upon hidden layer, and finally to the output layer, each layer only affects the state of next layer, thus establishing global nonlinear relationship between the input and the output layer. However, selecting a different starting point in network training may be getting different extreme points, it is difficult to guarantee the obtained extreme points is global optimal solution. So, it is need to find an algorithm with global search capability to determine the global extreme value range in order to overcome the lack of neural network. Then using neural network can prevent the local minimum point effectively, the experimenter can be satisfied with the optimal solution. Genetic algorithm is a multi-point multi-path searching algorithm with global search capability, and Quantum Genetic Algorithm (QGA) has characteristics such as small population size without affecting the performance of algorithm、develop and explore ability、fast convergence rate, for representing the chromosome as quantum bits encoded, completing the evolution as quantum gates. The experimental results in section 2.3 and 3.2 show that QGA and the MIMO detection based on QGA performance better than GA. The RBF neural network of K means clustering algorithm optimized by QGA is proposed in this paper.
The test results of QGA as test input of RBF neural network in proposed MIMO detector.
The detector optimization process is divided into two stages: QGA in a wide range of global "rough search" and the neural network of local "fine search." QGA is used to find a better search result for global searching in the solution space, then this result as the initial value of neural networks which find the global optimal solution.
Fig.4. Signal detection scheme of MIMO system based on neural network optimized by QGA
The proposed combination optimal scheme is depicted in Figure 4. To understand the detection performance of proposed scheme, we will compare it with QGA、MMSE and MIMO detection scheme based on neural network.
The simulations were performed in a M=N=4/M=N=8 environment and adopt the QPSK modulation. The control parameters of QGA in experiment set as section 3.2, the training data length of clustering RBF network is 160, the length of test sample data is 10240, and overlapping constant of hidden node is 1.0.
The results show in Figure 5 (left for M=N=4, right for M=N=8). Results show that: Whether M=N=4 or M=N=8, the RBF network of clustering algorithm optimized by QGA obtained better detection performance as QGA detection provide a better initial value for neural network, avoiding error detection code for selecting the initial value randomly.
Figure 5 The detection performance of QGA-RBF when QPSK modulation
4. MIMO-OFDM Signal Detection Scheme based on Quantum Optimization Algorithm
4.1 MIMO-OFDM signal detection system
The received signal of MIMO system has serious inter-symbol interference in frequency selective fading channel, so, the technology correspond to the frequency selective channel must be applied. Orthogonal Frequency Division Multiplexing (OFDM) is one of multi-carrier and narrow-band transmission technologies; it can be effective against frequency selective fading and ICI for subcarriers orthogonally. MIMO-OFDM[6] which constitutes of MIMO and OFDM technologies can overcome frequency selective of MIMO channel and achieve high utilization of frequency bandwidth has good development prospects[7,8].
Fig.6. Baseband system diagram of MIMO-OFDM
Non-coding MIMO-OFDM systems based spatial multiplexing is depicted in Figure 6. Setting up transmitter antennas and receiver antennas, channel between the transmit antennas and the receive antennas is multipath Rayleigh fading channel, OFDM subcarrier
上一页 [1] [2] [3] 下一页
MIMO信号检测系统量子算法的优化及运用英文小论文 第2页下载如图片无法显示或论文不完整，请联系qq752018766
