Constellation shaping optimization method based on genetic algorithm

Abstract

The invention provides a constellation shaping optimization method and system based on a genetic algorithm, and belongs to the technical field of optical fiber communication. The constellation point layout is coded into chromosomes, and the constellation point distribution is iteratively optimized by using selection, crossover and mutation operations of a genetic algorithm, so that the problems of high calculation complexity, low convergence speed and limited adaptability in the existing constellation shaping technology are solved. The transmitting end generates an optimized constellation through coding and fitness evaluation, and the receiving end improves the signal quality in combination with an equalization technology. According to the invention, the spectrum efficiency and anti-noise capability of the system are obviously improved, the bit error rate is reduced, and the method is suitable for 5G, 6G and other high-speed communication scenes.

Description

Technical Field

[0001] The present invention belongs to the technical field of optical fiber communication, and in particular relates to a constellation shaping optimization method and system based on a genetic algorithm, which is used to improve the spectrum efficiency and anti-noise capability of an optical communication system. Background Art

[0002] With the development of 5G and 6G technologies, high-speed coherent optical communication systems are placing increasingly stringent demands on spectral efficiency and noise immunity. Traditional rectangular QAM constellations, due to their regular distribution, do not adequately match channel characteristics, making signals susceptible to noise interference and resulting in high bit error rates. Among existing constellation shaping technologies, probabilistic shaping reduces average energy by adjusting the probability distribution of constellation points, but this has high computational complexity and slow convergence. Geometric shaping improves the minimum Euclidean distance by optimizing the positions of constellation points, but this relies on complex algorithms and has limited adaptability. Therefore, there is an urgent need for an efficient and adaptive constellation shaping method to address the high computational complexity, slow convergence, and lack of adaptability of existing technologies. Summary of the Invention

[0003] The present invention takes high-speed coherent optical communication systems as its application object and iteratively optimizes the constellation point layout through a genetic algorithm to achieve the following goals: reducing computational complexity, increasing convergence speed, and enhancing the system's adaptability to different channel conditions.

[0004] S1. Constellation point layout coding:

[0005] For the 16QAM constellation, the coordinates of the 16 constellation points are encoded as a complete chromosome, and multiple chromosomes are randomly generated to form the initial population;

[0006] S2. Fitness function design:

[0007] The minimum Euclidean distance is used as the core indicator, and the fitness function is defined as the square of the minimum Euclidean distance. The Euclidean distance between any two points in the constellation point set corresponding to each chromosome in the population is calculated, and the minimum value is taken and squared as the fitness value. The larger the value, the better the constellation layout. The core indicator of the fitness function is the minimum Euclidean distance of the constellation point set, and its mathematical expression is:

[0008] Fitness=(min i≠j d(x i , x j )) 2

[0009] Among them, d(x i , x j ) is the constellation point x i with x jBy calculating the distance between any two points and taking the square of the minimum value, the geometric shaping target is converted into a quantifiable fitness value. A larger value indicates a larger minimum interval between constellation points and a stronger anti-noise capability.

[0010] S3. Select an operation:

[0011] The roulette wheel selection method is used to calculate the selection probability based on the chromosome fitness value. The higher the fitness of the chromosome, the greater the probability of being selected, ensuring the retention and transmission of excellent genes. The formula for calculating the roulette wheel selection probability is:

[0012]

[0013] Among them, P i is the selection probability of the i-th chromosome, Fitness i is the fitness value, and N is the population size. This formula makes the probability of high-fitness chromosomes being selected proportional to their fitness, ensuring that high-quality genes have a higher probability of entering the parent generation, realizing the evolutionary mechanism of "survival of the fittest."

[0014] S4, cross operation:

[0015] Randomly select parent chromosomes with a set crossover probability (such as 0.8) and perform single-point crossover: randomly select a crossover point in the chromosome, exchange the gene segments of the two parent chromosomes after the crossover point, and generate two daughter chromosomes to increase population diversity;

[0016] S5. Mutation operation:

[0017] Flip the chromosome gene bit (0 to 1 or 1 to 0) with a low mutation probability (such as 0.01), and make slight adjustments to the constellation point coordinates to prevent the algorithm from falling into local optimality;

[0018] S6. Iterative optimization:

[0019] Repeat S2-S5, and calculate the fitness value in each generation. If the fitness does not improve significantly for several consecutive generations or reaches the maximum number of iterations, the iteration is terminated and the optimal constellation layout is output. The iterative process introduces an adaptive step adjustment mechanism to dynamically adjust the step length according to the current number of iterations t. The formula is:

[0020]

[0021] Among them, step max and step min The maximum and minimum variable step lengths are respectively, and MaxIter is the preset maximum number of iterations. This mechanism uses a large step size to explore new layouts in the early stages of iteration, and a smaller step size for refined search in the later stages, balancing global exploration and local development capabilities, and improving the algorithm's convergence efficiency.

[0022] The present invention uses the global search capability of the genetic algorithm to dynamically optimize the distribution of constellation points. Compared with traditional methods, the computational complexity is reduced by 40%, the convergence speed is increased by 50%, and it can adapt to different channel conditions, significantly improving system performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 This is a flow chart of the genetic algorithm of the present invention, showing the complete process from initialization to output of the optimal constellation; DETAILED DESCRIPTION

[0024] Taking 16QAM constellation optimization as an example, the specific steps are as follows:

[0025] S1. Encoding module: quantize the coordinates of each constellation point (I, Q) into 8-bit binary numbers. The 16 points with a total of 16×16=256 bits constitute a chromosome, generating 50 initial chromosomes.

[0026] S2, evaluation module: Calculate the Euclidean distances of the 16 points in each chromosome, take the minimum value and square it as the fitness. For example, if the minimum distance of the initial population is 2.0, the fitness is 4.0.

[0027] S3, selection module: Roulette wheel selection is used, the first 20 chromosomes with the highest fitness directly enter the parent generation, and the rest are supplemented through roulette wheel selection.

[0028] S4, crossover module: Perform single-point crossover on the parent chromosome, such as crossover at position 128, to generate new offspring.

[0029] S5, mutation module: flip the gene position of each daughter chromosome with a probability of 0.01, for example, the coordinates of a point (3,3) mutate to (3,3.1).

[0030] S6, control module: set the maximum number of iterations to 200 generations, terminate when the fitness growth is less than 1% for 50 consecutive generations, and output the optimal constellation layout.

[0031] Through the above-described implementation, tests conducted under a Gaussian channel show that the optimized constellation diagram reduces the bit error rate by 35% and improves spectral efficiency by 25% compared to traditional 16QAM, effectively verifying the superiority of the technical solution. This invention utilizes a genetic algorithm to achieve efficient optimization of the constellation point layout, combined with adaptive step size adjustment and multi-technique integration, to address the shortcomings of existing technologies and possesses broad engineering application value.

Claims

1. A constellation shaping optimization method based on genetic algorithm, characterized in that: The following steps are involved: S1. Constellation point layout encoding: The coordinates (I, Q) of the constellation points in the complex plane are quantized and encoded into binary strings to construct an initial population containing multiple chromosomes. The chromosomes carry the complete constellation point layout information. S2. Fitness function design: Based on the initial population generated in step S1, the fitness function is designed using the minimum Euclidean distance or bit error rate (BER). The Euclidean distance between any two points in the constellation point layout corresponding to each chromosome is calculated. The minimum value is taken and squared as the fitness value of the chromosome, which is used to quantify the quality of the constellation layout. S3, selection operation: According to the fitness value calculated in step S2, high fitness chromosomes are selected from the population as parents through roulette wheel selection or tournament selection method to ensure the retention and transmission of excellent genes; S4, crossover operation: Perform a single-point or multi-point crossover operation on the parent chromosome selected in step S3 with a preset crossover probability, generating offspring chromosomes carrying parental characteristics by exchanging gene segments, thereby increasing population diversity; S5, mutation operation: flip the offspring chromosome gene bits generated in step S4 with a preset mutation probability, make slight adjustments to the constellation point coordinates, and introduce a new constellation point layout to avoid local optimality; S6. Iterative optimization: Repeat steps S2-S5 to generate a new generation of population and evaluate its fitness. Dynamically optimize the variable step length through the adaptive step size adjustment mechanism until the termination conditions such as no significant improvement in fitness over multiple generations or reaching the maximum number of iterations are met, and the optimal constellation layout is output.

2. The constellation shaping optimization method based on genetic algorithm according to claim 1, characterized in that: The fitness function S2 is defined as the square of the minimum Euclidean distance, and the formula is: Fitness=(min i≠j d(x i ,x j )) 2 Among them, d(x i , x j ) is the constellation point x i with x j The Euclidean distance of .

3. The constellation shaping optimization method based on genetic algorithm according to claim 1, characterized in that: The formula for calculating the probability of S3 roulette selection is: Among them, P i is the selection probability of the i-th chromosome, Fitness i is the fitness value, and N is the population size.

4. The constellation shaping optimization method based on genetic algorithm according to claim 1, characterized in that: S6 includes an adaptive step size adjustment mechanism, which dynamically adjusts the step size according to the number of iterations. The formula is:

5. A constellation shaping optimization system based on genetic algorithm, characterized in that: include: Encoding module: used to encode constellation point coordinates into chromosomes and initialize the population; Evaluation module: used to calculate the fitness value of chromosome; Selection module: used to perform selection operations to screen parent chromosomes; Crossover module: used to perform crossover operations to generate offspring chromosomes; Mutation module: used to perform mutation operations to introduce new layouts; Control module: used to control the iteration process, determine the termination conditions and output the optimal constellation.

Images