Quantum heuristic combined optimization solving method and system

Abstract

The invention relates to the technical field of computers, in particular to a quantum heuristic combinatorial optimization solution method and system, and the method comprises the steps: generating an initial population through quantum probability distribution, carrying out the population updating through quantum behavior simulation, and calculating the quality score of a solution in combination with multi-target quality evaluation. Through quantum crossover mutation operation, efficient parallel search is realized, the time complexity is reduced to O (nlogn), and the convergence speed is increased by 40%-60%. And meanwhile, the global search capability is enhanced by utilizing quantum characteristics, local optimum is effectively avoided, the solution quality is averagely improved by 30%, and the solution efficiency and the global search performance are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and particularly to a quantum-inspired combinatorial optimization solving method and system thereof, which is applicable to solving various complex combinatorial optimization problems. Background Art

[0002] Combinatorial optimization problems widely exist in various fields of the real world, such as path planning, resource scheduling, network design, etc. With the continuous increase in the problem scale and complexity, traditional optimization algorithms such as greedy algorithms, dynamic programming, and heuristic algorithms often face challenges such as low computational efficiency and easy to fall into local optima when dealing with large-scale combinatorial optimization problems.

[0003] In recent years, quantum computing, as a new computing paradigm, has shown great potential. Its parallelism and probabilistic characteristics provide new ideas for solving complex optimization problems. However, the practical application of real quantum computers still faces many technical challenges. Quantum-inspired algorithms provide a new way to solve combinatorial optimization problems by simulating the core characteristics of quantum computing, such as quantum superposition and quantum entanglement, on classical computers.

[0004] Existing quantum-inspired algorithms mainly include quantum genetic algorithms, quantum particle swarm algorithms, etc. However, these methods still have problems such as low efficiency and unstable convergence when dealing with high-dimensional and multi-constrained combinatorial optimization problems. Especially under complex constraint conditions, existing methods are difficult to effectively balance the global exploration and local search capabilities, resulting in limited algorithm performance.

[0005] Therefore, there is an urgent need for a new solving method that can fully combine the advantages of quantum computing and classical optimization algorithms to improve the solving efficiency and solution quality of combinatorial optimization problems. Summary of the Invention

[0006] The object of the present invention is to provide a quantum-inspired combinatorial optimization solving method and system thereof, aiming to solve the problems of low efficiency and easy to fall into local optima in existing combinatorial optimization algorithms when dealing with high-dimensional and multi-constrained problems.

[0007] The present invention proposes a quantum-inspired combinatorial optimization solving method, including:

[0008] A step of generating an initial population, generating an initial population by using a quantum probability distribution method according to the settings of the optimization problem;

[0009] A step of updating the population, updating the initial population based on a quantum behavior simulation method;

[0010] A step of quality evaluation, introducing a multi-objective quality evaluation method to calculate the quality score of each solution in the updated population;

[0011] Cross - mutation step: Based on the quantum - behavior simulation method and the mass fraction, perform probabilistic crossover and mutation on the updated population;

[0012] Termination judgment step: Judge whether the loop termination condition is reached. If so, output the optimal solution; otherwise, return to the population update step.

[0013] Preferably, the step of generating the initial population specifically includes:

[0014] Transform the optimization problem into a standard form, set the problem dimension, and determine the crossover probability and mutation probability;

[0015] Generate the first solution through quantum - behavior simulation, set the initial population in quantum - state representation, and calculate the wave function as a linear superposition of the solution space of the initial problem;

[0016] Select the quantum entanglement coefficient, perform quantum - state interference operations on a single random quantum state and the quantum entanglement basis, and obtain the complete initial population through quantum - behavior simulation observation.

[0017] Preferably, the population update step specifically includes:

[0018] Map each solution in the population to a particle in the solution space, and the particle state is composed of coefficients;

[0019] Update the particle state through quantum - state evolution and the rotation gate of the solution space;

[0020] Select a particle set from the initial solution set and select the characteristics corresponding to the target solution in the particle set;

[0021] Obtain the states of all next - step particles through the quantum entanglement state;

[0022] Update each quantum state through the rotation gate of the solution space to obtain the updated particle state.

[0023] Preferably, the mass evaluation step specifically includes:

[0024] Determine the minimum cost or benefit range of individuals in the problem solution space;

[0025] Calculate the weight value of each evaluation index;

[0026] Score - calculate the multi - objective quality of the initial population to generate a comprehensive quality score;

[0027] Sort the solutions in the population based on the mass fraction.

[0028] Preferably, the probabilistic crossover in the cross - mutation step specifically includes:

[0029] Two particles in the parent generation generate new particles through particle - to - particle crossover operations;

[0030] Perform an inter-particle crossover operation on the particle set in the parent population to generate a child particle set;

[0031] Among them, the crossover operation performs information exchange based on the randomly selected cut point position.

[0032] Preferably, the probability mutation in the crossover and mutation step specifically includes:

[0033] Perform a mutation operation on each solution in the population. After mutation, a new solution is obtained, which forms the child population;

[0034] Among them, the mutation probability is proportional to the quality evaluation score;

[0035] The mutation operation is realized by quantum random walk to enhance the exploration ability of the search space.

[0036] Preferably, the termination judgment step specifically includes:

[0037] Set the loop termination condition as the threshold of the population quality score;

[0038] When the termination condition is reached, output the optimal solution, and evaluate the quality of the solution by outputting the multi-objective evaluation score;

[0039] According to the quality score threshold, dynamically adjust the search strategy;

[0040] When the quality score is high, it indicates that the current search direction is effective, and the updated step size and weight value are small, otherwise they are large.

[0041] Preferably, it further includes:

[0042] Implement parallel computing through the quantum behavior simulation method, and generate the initial particle set by observing the quantum behavior simulation;

[0043] Evolve and observe the particle set to generate new particles;

[0044] For each observation result, select the characteristics of the target solution from the solution space of the current population;

[0045] Observe the evolution of the solution space through quantum behavior simulation, observe the particles in the population, and obtain a new quantum state.

[0046] Preferably, it further includes:

[0047] Introduce an adaptive quantum amplitude adjustment mechanism to balance local search and global exploration;

[0048] Among them, the quantum amplitude adjustment of the particle is controlled by an adaptive factor, and the adaptive factor is dynamically calculated according to the difference between the local optimal energy and the global optimal point;

[0049] Based on the quantum memory effect, historical search information is incorporated into the current search strategy to avoid repeated exploration of known regions.

[0050] A quantum-inspired combinatorial optimization solution system, comprising:

[0051] An initial generation module for generating an initial population using a quantum probability distribution method according to the settings of the optimization problem;

[0052] An update module for updating the initial population based on a quantum behavior simulation method;

[0053] A quality assessment module for calculating the quality score of each solution in the updated population by introducing a multi-objective quality assessment method;

[0054] A crossover and mutation module for performing probabilistic crossover and mutation on the updated population based on the quantum behavior simulation method and the quality score;

[0055] A termination judgment module for judging whether the loop termination condition is reached. If so, the optimal solution is output; otherwise, it returns to the update module for the next loop.

[0056] By introducing a quantum behavior simulation method, combining a multi-objective quality assessment mechanism and an adaptive quantum amplitude adjustment technology, the present invention constructs a novel combinatorial optimization solution framework. This method has the following beneficial effects:

[0057] 1. Improve the solution efficiency: Through quantum behavior simulation for parallel search, compared with traditional methods, the time complexity is reduced from O(n²) to O(nlogn), and the convergence speed is increased by about 40% - 60%;

[0058] 2. Enhance the global search ability: Based on the quantum probability distribution and quantum memory effect, it effectively avoids falling into local optima, and the quality of the solutions is on average improved by 30% under the same computing resources;

[0059] 3. Improve the multi-objective optimization ability: Through the multi-objective quality assessment mechanism, it can find the best balance among multiple conflicting objectives and is applicable to optimization problems under complex constraint conditions;

[0060] 4. Enhance the algorithm adaptability: Based on the adaptive quantum amplitude adjustment mechanism, the algorithm can dynamically adjust the search strategy according to the search progress and adapt to different types and scales of optimization problems. Brief Description of the Drawings

[0061] Figure 1 is a flowchart of the quantum-inspired combinatorial optimization solution method of the present invention;

[0062] Figure 2 is a detailed flowchart of the initial population generation step of the present invention;

[0063] Figure 3 This is the detailed flowchart of the population update step of the present invention;

[0064] Figure 4 This is the detailed flowchart of the multi-objective quality assessment step of the present invention;

[0065] Figure 5 This is the detailed flowchart of the crossover and mutation step of the present invention;

[0066] Figure 6 This is the detailed flowchart of the termination judgment step of the present invention;

[0067] Figure 7 This is the structural block diagram of the quantum-inspired combinatorial optimization solution system of the present invention. Specific embodiments

[0068] Please refer to the appendix Figure 1-7 , and the present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments. However, the embodiments of the present invention are not limited thereto.

[0069] Refer to Figure 1 , the present invention provides a quantum-inspired combinatorial optimization solution method, including the following steps: generating an initial population step S1, population update step S2, quality assessment step S3, crossover and mutation step S4, and termination judgment step S5.

[0070] In the step S1 of generating an initial population, according to the settings of the optimization problem, a quantum probability distribution method is used to generate an initial population. Specifically, first, the optimization problem is transformed into a standard form, the problem dimension is set, and the crossover probability and mutation probability are determined. Then, the first solution is generated through quantum behavior simulation, the initial population is set to be represented in a quantum state, and the wave function is calculated as a linear superposition of the solution space of the initial problem. Next, the quantum entanglement coefficient is selected, and quantum state interference operations are performed on a single random quantum state and the quantum entanglement basis, and a complete initial population is obtained through quantum behavior simulation observation.

[0071] Preferably, as Figure 2As shown, the step S1 of generating the initial population may include three sub-steps: a standardization sub-step S11, a first solution generation sub-step S12, and a population completion sub-step S13. In the standardization sub-step S11, the optimization problem is transformed into a standard form, the dimension d of the problem is set. When generating the initial population, each solution contains d variables, and the variable values are integers between 0 and d - 1. At the same time, the crossover probability Pc and the mutation probability Pm are set. In one embodiment, the value range of Pc is 0.6 - 0.9, preferably 0.8, because a higher crossover probability helps to maintain population diversity; the value range of Pm is 0.01 - 0.1, preferably 0.05, because a lower mutation probability can avoid destroying existing high-quality solutions.

[0072] In the first solution generation sub-step S12, the first solution is generated by means of quantum behavior simulation. The initial population is set as The particle represents the solution space of the initial problem, and the solution observed from the particle is . Calculate the wave function of the observed problem by quantum behavior simulation as , where is the linear superposition of the standard basis of the initial problem solution space, represents the initial basis of the observed problem.

[0073] In the population completion sub-step S13, the quantum entanglement coefficient is selected, and quantum state interference operations are performed on a single random quantum state and the quantum entanglement basis to generate new particles. Its expression is:

[0074] ,

[0075] where, represents the entanglement basis of the observed problem, represents a single particle, represents the number of single particles participating in the interference, represents the probability of interference, represents the order of the particles participating in the interference. By observing the solution of the quantum entanglement superposition through the quantum behavior simulation method, the initial population is obtained, where is a single solution obtained by quantum behavior simulation, is 's linear superposition.

[0076] In the population update step S2, the initial population is updated based on the quantum behavior simulation method. As Figure 3 shown, this step specifically includes: a mapping sub-step S21, a selection sub-step S22, an observation sub-step S23, and a rotation update sub-step S24.

[0077] In the mapping sub-step S21, the initial population parameters are updated, and each solution xj in the population is mapped to a particle in the solution space. The state of the particle is composed of coefficients. The particle state is updated by evolving the quantum state through the rotation gate in the solution space. The initial particle is mapped to , the particle state after the -th step update is:

[0078] ,

[0079] where, is the initial basis of the initial problem, is the basis of the -th step, is the initial particle quantum state, is the particle quantum state of the -th step. When , is the set of all possible solution sets of the initial problem, where is the dimension, is the variable length, is the probability value between [0, 1]. is the particle characteristic of the solution at the -th step, is the quantum basis of the initial problem, is the particle characteristic of the solution at the -th step. The characteristic of the solution at the -th step is the solution set of the initial problem.

[0080] In the selection sub-step S22, a particle set is selected from the initial solution set. In the particle set, the characteristic of the solution at the -th step corresponding to the target solution is selected. The basis of the -th step is , the particle quantum state of the solution at the -th step is . The states of all particles at the -th step are obtained through the quantum entanglement state. The calculation expression is:

[0081] ,

[0082] where, represents the evolution function from the solution at time to the solution at time . is the initial state of particle , is the quantum state of particle , is the quantum state of particle transformed from the initial state after evolution to the state at time . Parameters representing the initial state is the evolution step of the particle.

[0083] In the observation sub-step S23, the quantum state evolution is observed by superimposing quantum states, the particle characteristics of the solution at time k are obtained, and the particle characteristics are transformed into the problem space to obtain the corresponding solution, which constitutes , represents the particle characteristics mapped into the solution space. Generate the solution transformed from the particle characteristics into the problem space, which constitutes . Map the wave function of the solution at time k to , and observe the evolution of the quantum state through quantum behavior simulation to obtain the corresponding quantum state evolution, represents the probability of observing the quantum state evolution, and the particle state of the solution at the next moment is obtained.

[0084] In the rotation update sub-step S24, each quantum state is updated through the rotation gate in the solution space to obtain the particle state of the solution at the moment, and the calculation formula is:

[0085] ,

[0086] Among them, is the rotation gate of the solution space, which is used to accelerate the convergence process of the problem; is the rotation angle of the solution space, is the characteristic of the solution at the moment, represents the characteristic of the solution at the 1 moment, represents the basis of the solution space at the moment. By evolving the quantum state, update the solution set , where is the particle characteristic, is the mapping of the observed quantum state in the solution space at the moment, is the mapping of the observed particle quantum state in the solution space at the moment, is the particle quantum state of the solution at the moment, is the observed particle. By repeating steps S21 to S24, update the initial population, and update the initial population to obtain a new population .

[0087] In the quality evaluation step S3, a multi-objective quality evaluation method is introduced to calculate the quality score of each solution in the updated population. As Figure 4 shown, this step includes a determination range sub-step S31 and a calculation score sub-step S32.

[0088] In the determining range sub-step S31, the minimum cost or benefit range of individuals in the problem solution space is determined. The problem solution space contains the minimum cost or benefit λ of the individuals, and the individual cost or benefit of the optimal solution in the k-th iteration is represented by λ_k. The problem solutions are represented from the minimum cost / benefit to the maximum cost / benefit as .

[0089] In the calculating score sub-step S32, the multi-objective quality scores of the initial population are evaluated, and the calculation formula is:

[0090] ,

[0091] wherein, represents the weight of each evaluation index, represents the number of indexes, , represents the number of optimal solutions of the problem solution under the -th index, represents the optimal solution of the problem solution under the -th index. In one embodiment, the value of the weight can be set according to the specific problem. Usually, the weight of the main optimization objective can be set to 0.6 - 0.8, and the weight of the secondary objective can be set to 0.2 - 0.4.

[0092] In the crossover and mutation step S4, based on the quantum behavior simulation method and the said quality scores, probabilistic crossover and mutation are implemented on the updated population. As Figure 5 shown, this step includes a crossover sub-step S41 and a mutation sub-step S42.

[0093] In the crossover sub-step S41, two particles in the parent generation generate new particles through the inter-particle crossover operation; the particle set in the parent population is subjected to the inter-particle crossover operation to generate the offspring particle set; wherein, is the -th solution in the solution space, is the -th solution in the solution space, is a random integer value between, where is the length of the problem solution, is the -th qubit of the quantum state of the particle in the offspring population. The crossover operation performs information exchange based on the randomly selected cut point position, and the expression is:

[0094] ,

[0095] In the mutation sub-step S42, a mutation operation is performed on each solution in the population, and new solutions are obtained after mutation to form the offspring population Among them, the mutation probability is , and the calculation formula is expressed as:

[0096] ,

[0097] Among them, represents quantum random walk, represents the solution characteristics of the offspring particles ; represents the solution characteristics of the parent particles ; The mutation probability is proportional to the quality evaluation score. Preferably, can be set as the product of the basic mutation probability and the quality score, that is , where is the basic mutation probability, usually set to 0.01 - 0.1.

[0098] In the termination judgment step S5, it is judged whether the loop termination condition is reached. If so, the optimal solution is output; otherwise, return to the population update step S2 for the next loop. As Figure 6 shown, this step includes a condition judgment sub-step S51, a strategy adjustment sub-step S52, and a result output sub-step S53.

[0099] In the condition judgment sub-step S51, the loop termination condition is set as the threshold of the population quality score; when the termination condition is reached, the optimal solution is output, and the quality of the solution is evaluated by outputting the multi-objective evaluation score; the termination condition is expressed as:

[0100] ,

[0101] Among them, represents the number of solutions in the optimal solution set under the th index, represents the number of solutions in the optimal solution set under the th index, represents the threshold under the th index, represents the number of indexes. Preferably, the threshold can be set to 0.8 - 0.95, indicating that the iteration is terminated when the solutions in the optimal solution set approach or reach the theoretical optimum under each index.

[0102] In the strategy adjustment sub-step S52, according to the quality score threshold, the search strategy is dynamically adjusted; a high quality score indicates that the current search direction is effective, and the update step size and weight are small, otherwise they are large; the formula for calculating the step size is expressed as:

[0103] ,

[0104] Among them, represents the iteration step size, represents the number of iterations, is the population size for quantum behavior simulation, represents the threshold, is a random number between [0, 1]; when the quality score is lower than the threshold, the search range is expanded by adaptively iterating the search step size, otherwise, the search range is reduced. In one embodiment, the initial step size can be set to 0.1 - 0.5, and the threshold can be set to 0.7 - 0.9. As the iteration progresses, the step size gradually decreases, which helps the algorithm to converge.

[0105] In the result output sub-step S53, the optimal solution is evaluated through the multi-objective evaluation score; the judgment criterion for the optimal solution is:

[0106] ,

[0107] where is the quality score of the th solution.

[0108] The present invention also provides a quantum-inspired parallel computing mechanism, which realizes parallel computing through quantum behavior simulation. Specifically, it includes: generating an initial particle set through observing quantum behavior simulation; evolving and observing the particle set to generate new particles; for each observation result, selecting the characteristics of the target solution from the solution space of the current population; observing the evolution of the solution space through quantum behavior simulation, observing the particles in the population to obtain new quantum states, and performing crossover and mutation on the particle set in the observation result to generate new particles and; repeating the crossover and mutation operations on the particle set to obtain the finally searched population and.

[0109] In addition, the present invention also introduces an adaptive quantum amplitude adjustment mechanism to balance local search and global exploration. Among them, the quantum amplitude adjustment of particles is controlled by an adaptive factor, and the adaptive factor is dynamically calculated according to the difference between the local optimal energy and the global optimal point; based on the quantum memory effect, historical search information is incorporated into the current search strategy to avoid repeated exploration of known regions. The quantum amplitude adjustment expression is:

[0110] ,

[0111] where is the quantum amplitude of particle , represents the conjugate of the quantum amplitude of is the adaptive factor of particle , which is used to control the range of the quantum amplitude adjusted by particle , and its calculation formula is:

[0112] ,

[0113] wherein represents the local optimal energy of the particle , represents the difference between the global optimal points of the particle , represents the adaptive fitness function of the particle . In one embodiment, the value range of the adaptive factor is usually 0.01 - 0.2. A smaller value helps with fine local search, and a larger value helps with global exploration.

[0114] Referring to Figure 7 , the present invention also provides a quantum - inspired combinatorial optimization solving system, including: an initial generation module 1, an update module 2, a quality evaluation module 3, a crossover and mutation module 4, and a termination judgment module 5.

[0115] The initial generation module 1 is used to generate an initial population by using a quantum probability distribution method according to the settings of the optimization problem. More specifically, the initial generation module 1 includes a normalization unit 11, a first solution generation unit 12, and a population completion unit 13, which respectively correspond to implementing the normalization sub - step S11, the first solution generation sub - step S12, and the population completion sub - step S13 in the foregoing method.

[0116] The update module 2 is used to update the initial population based on a quantum behavior simulation method. More specifically, the update module 2 includes a mapping unit 21, a selection unit 22, an observation unit 23, and a rotation update unit 24, which respectively correspond to implementing the mapping sub - step S21, the selection sub - step S22, the observation sub - step S23, and the rotation update sub - step S24 in the foregoing method.

[0117] The quality evaluation module 3 is used to introduce a multi - objective quality evaluation method to calculate the quality scores of each solution in the updated population. More specifically, the quality evaluation module 3 includes a determination range unit 31 and a score calculation unit 32, which respectively correspond to implementing the determination range sub - step S31 and the score calculation sub - step S32 in the foregoing method.

[0118] The crossover and mutation module 4 is used to perform probabilistic crossover and mutation on the updated population based on the quantum behavior simulation method and the quality scores. More specifically, the crossover and mutation module 4 includes a crossover unit 41 and a mutation unit 42, which respectively correspond to implementing the crossover sub - step S41 and the mutation sub - step S42 in the foregoing method.

[0119] The termination judgment module 5 is used to judge whether the loop termination condition is reached. If so, the optimal solution is output; otherwise, it returns to the update module 2 for the next loop. More specifically, the termination judgment module 5 includes a condition judgment unit 51, a strategy adjustment unit 52, and a result output unit 53, which respectively implement the condition judgment sub-step S51, the strategy adjustment sub-step S52, and the result output sub-step S53 in the foregoing method.

[0120] Preferably, the system further includes a parallel computing module 6 and an adaptive adjustment module 7. The parallel computing module 6 is used to implement parallel computing through the quantum behavior simulation method. The adaptive adjustment module 7 is used to introduce an adaptive quantum amplitude adjustment mechanism to balance local search and global exploration.

[0121] The quantum-inspired combinatorial optimization solution method and system of the present invention construct an efficient and flexible optimization solution framework by integrating the advantages of quantum computing and the practicality of classical optimization algorithms. This method performs excellently in practical applications. For example, in a traveling salesman problem test with 1000 cities, compared with the traditional genetic algorithm, the convergence speed is increased by about 50%, and the quality of the final solution is improved by about 25%. In the resource scheduling problem under complex constraint conditions, compared with the particle swarm algorithm, the feasibility of the solution is increased by about 35%, and the overall calculation time is reduced by about 40%.

[0122] The above is only a preferred embodiment of the present invention and does not impose any form of limitation on the present invention. Although the present invention has been disclosed above in a preferred embodiment, it is not intended to limit the present invention. Any person skilled in the art can make many possible changes and modifications to the technical solution of the present invention within the scope of the technical solution of the present invention, or modify it into an equivalent embodiment with equivalent changes. Therefore, any simple modification, equivalent change, and modification made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention shall fall within the scope of the protection of the technical solution of the present invention.

[0123] The technical solution and its application of the present invention will be further described below in conjunction with specific embodiments.

[0124] Embodiment 1: Solving the Traveling Salesman Problem

[0125] This embodiment takes the classical traveling salesman problem (TSP) as an example to show the application of the quantum-inspired combinatorial optimization solution method of the present invention.

[0126] First, in the step S1 of generating the initial population, the TSP problem is transformed into a standard form. For a TSP problem with 50 cities, the problem dimension is set , and each solution is represented as a permutation of the city visit order. The crossover probability is set , Mutation probability , Initial population size .

[0127] The first solution is generated by using the quantum behavior simulation method as follows:

[0128] 1. Map the TSP problem to the quantum space, where each city corresponds to a position state

[0129] 2. Initialize the quantum state , representing the uniform superposition of all city visit sequences

[0130] 3. Calculate the initial wave function , where Then select the quantum entanglement coefficient , and perform an interference operation on a single random quantum state and the quantum entanglement basis. In this embodiment, the number of single particles participating in the interference is set , Interference probability . Through the observation of the quantum behavior simulation, the initial population is obtained.

[0131] In the population update step S2, the following operations are performed:

[0132] 1. Map each solution in the initial population to a particle in the solution space

[0133] 2. Set the rotation angle θ = π / 18 and construct the rotation gate R(θ)

[0134] 3. Select a particle set from the initial solution set and select the characteristics corresponding to the target solution

[0135] 4. Update the particle state through the quantum entanglement state

[0136] 5. Update the quantum state through the rotation gate in the solution space

[0137] Preferably, in this embodiment, the evolution function adopts a linear mapping form:

[0138] ,

[0139] where is the evolution parameter, and its value is 0.6, representing the weight ratio of the new and old states.

[0140] After the update, a new population is obtained. For the TSP problem, each solution represents a feasible travel path.

[0141] In the quality evaluation step S3, for the TSP problem, the evaluation index is set as the total travel distance. In the problem solution space, the shortest path length is denoted as λ, and the shortest path length in the current iteration is denoted as In this embodiment, normalization processing is adopted to simplify the quality evaluation score calculation formula to:

[0142] ,

[0143] where is the total path length corresponding to the th solution. This scoring method ensures that the shorter the path, the higher the quality score.

[0144] In the crossover and mutation step S4, the crossover operation is performed as follows:

[0145] 1. Randomly select two parent solutions and ;

[0146] 2. Randomly select a cut point , with the range being [1, 49];

[0147] 3. For the cities before , keep the original order;

[0148] 4. For the cities after , obtain them by crossover while avoiding duplicate cities;

[0149] Specifically, this embodiment adopts the partially mapped crossover (PMX) method, which is particularly suitable for the TSP problem and can ensure the feasibility of the offspring solutions.

[0150] The mutation operation is implemented by quantum random walk:

[0151] 1. Randomly select two positions and ;

[0152] 2. Exchange the cities at these two positions;

[0153] 3. The mutation probability is proportional to the quality score, and the calculation formula is ;

[0154] In the termination judgment step S5, the termination condition is set as follows: the improvement amplitude of the optimal solution is less than 0.1% for 50 consecutive iterations, or the maximum number of iterations reaches 500 times. For the TSP problem, a single index , the threshold is set to , indicating that the iteration can be terminated when the optimal solution is close to 95% of the theoretical optimal value.

[0155] During the iteration process, the search step size is dynamically adjusted:

[0156] ,

[0157] where \(l\) is the current iteration number, is a random number between \([0, 1]\).

[0158] Through experiments, this method can converge to a near-optimal solution after an average of 200 iterations for the 50-city TSP problem, showing significant advantages over the traditional genetic algorithm (average 350 iterations) and the particle swarm algorithm (average 300 iterations). The final path length is on average 15% shorter than that of the traditional method.

[0159] Example 2: Resource Allocation Problem

[0160] This example is applied to the resource allocation problem, specifically for allocating types of resources to tasks to maximize the overall benefit.

[0161] In the step S1 of generating the initial population, set the problem dimension , and the variable value represents the amount of resources allocated to each task. In this example, , , set the crossover probability , the mutation probability , and the initial population size is 80.

[0162] Generate the initial solution using the quantum superposition state. By setting different amplitude distributions, make the initial solution bias towards the feasible region. Set the quantum entanglement coefficient , the number of interfering single particles , and the interference probability .

[0163] In the step S2 of population update, this example adopts the adaptive rotation angle strategy:

[0164] ,

[0165] where , is the quality fraction of the current solution, and are the minimum and maximum quality fractions in the current population respectively. This strategy enables solutions with lower quality to obtain a larger rotation angle, facilitating rapid convergence towards a better region.

[0166] In the step S3 of quality evaluation, the resource allocation problem involves multiple metrics: total benefit, resource utilization rate, and task completion rate. Set the weights to be , , respectively, which reflects the relative importance of the three metrics.

[0167] The formula for calculating the quality fraction is:

[0168] ,

[0169] where , and represent the total revenue, resource utilization rate, and task completion rate of the th solution respectively, and , and represent the maximum values of the corresponding indicators respectively.

[0170] In the crossover and mutation step S4, perform uniform crossover operation:

[0171] 1. Randomly generate a binary mask with the same length as the solution;

[0172] 2. When the mask bit is 1, the offspring 1 takes the corresponding bit of the parent 1, and the offspring 2 takes the corresponding bit of the parent 2;

[0173] 3. When the mask bit is 0, the offspring 1 takes the corresponding bit of the parent 2, and the offspring 2 takes the corresponding bit of the parent 1;

[0174] The mutation operation adopts Gaussian perturbation:

[0175] ,

[0176] where represents a Gaussian random number with a mean of 0 and a standard deviation of , and is set to 0.1. The higher the quality score, the smaller the perturbation, which helps to retain the characteristics of high-quality solutions.

[0177] In the termination judgment step S5, set the multi-index threshold to , indicating that it can be terminated when the total revenue reaches 90% of the theoretical maximum value, the resource utilization rate reaches 85%, and the task completion rate reaches 80%. The maximum number of iterations is set to 300.

[0178] After experiments, this method shows excellent performance in the resource allocation problem. Compared with the traditional method, the resource utilization rate is increased by about 18%, and the total revenue is increased by about 22%.

[0179] Example 3: Network Routing Optimization Problem

[0180] This example is applied to the network routing optimization problem, and the goal is to minimize the total delay and load imbalance while meeting the network traffic demand.

[0181] In the step S1 of generating the initial population, the network routing problem is represented as a matrix, where each row represents a source-destination pair and each column represents possible path selections. In this embodiment, it is assumed that the network has 20 nodes, 30 links, and 15 traffic demands to be routed, and each demand has 3 alternative paths. Therefore, the problem dimension d = 15. The crossover probability Pc = 0.85, the mutation probability Pm = 0.06, and the size of the initial population is 120.

[0182] To improve the quality of the initial population, this embodiment adopts a quantum bias strategy, that is, different probability amplitudes are set for the initial quantum state, so that the states more likely to generate high-quality solutions obtain higher probabilities:

[0183] ,

[0184] where Pre-calculated based on simple heuristic rules, which reflects the initial priorities of each path.

[0185] In the population update step S2, this embodiment introduces the quantum memory effect, enabling the particles to remember their historical optimal positions:

[0186] ,

[0187] where is the historical optimal quantum state of the particle , is the memory factor, set to 0.3.

[0188] In the quality evaluation step S3, the network routing optimization problem involves three metrics: total delay, load balance, and throughput. The weights are set to be , , .

[0189] The quality score is calculated using a non-linear mapping:

[0190] ,

[0191] where 、 and respectively represent the total delay, load imbalance, and throughput of the th solution, and are sensitivity coefficients, making the score more sensitive to delay and load balance.

[0192] In the crossover and mutation step S4, a path exchange crossover operation is adopted:

[0193] 1. Randomly select routing demands;

[0194] 2. Swap the path selections of these requirements in two parent solutions;

[0195] The mutation operation adopts a multi-level mutation strategy:

[0196] 1. Execute a mild mutation with a probability of 0.7: Randomly select a requirement and change its path selection;

[0197] 2. Execute a moderate mutation with a probability of 0.2: Randomly select 2 - 3 requirements and change their path selections;

[0198] 3. Execute a strong mutation with a probability of 0.1: Randomly select 4 - 5 requirements and change their path selections;

[0199] The relationship between the mutation probability and the quality score is , the higher the quality, the lower the mutation probability.

[0200] In the termination judgment step S5, set an adaptive termination condition: Terminate when the average improvement rate of the optimal solution is lower than 0.5% in 30 consecutive iterations, or when the maximum number of iterations 400 is reached.

[0201] This embodiment particularly introduces a quantum amplitude adaptive adjustment mechanism, and its calculation formula is:

[0202] ,

[0203] where and are respectively the local optimal energy and the global optimal energy of particle , is the current number of iterations, is the maximum number of iterations. This adaptive mechanism makes the algorithm tend to global exploration in the initial stage and local optimization in the later stage.

[0204] The experimental results show that this method has achieved remarkable results in the network routing optimization problem: the average delay is reduced by 25%, the load balance is improved by 30%, the throughput is increased by 15%, and the overall network performance is significantly improved.

[0205] Example 4: Parallel Computing Acceleration

[0206] This embodiment demonstrates the specific implementation and performance improvement of the parallel computing module 6 in the present invention. When dealing with large-scale optimization problems, parallel computing can significantly improve the algorithm efficiency.

[0207] The parallel strategy based on quantum behavior simulation is as follows:

[0208] 1. Divide the initial population into k subpopulations, and each subpopulation is assigned to a different computing core;

[0209] 2. Each sub - population evolves independently through calculation, including operations of update, evaluation, crossover, and mutation;

[0210] 3. Every certain number of iterations (e.g., 10 times), information is exchanged among sub - populations to share their respective optimal solutions;

[0211] 4. According to the shared information, adjust the search directions and strategies of each sub - population;

[0212] This embodiment is implemented on a computer with an 8 - core processor, and the population is divided into 8 sub - populations. Experiments show that, compared with serial computing, the parallel strategy shortens the computing time by about 75%, and due to the diversity of the search directions of sub - populations, the quality of the final solution is also improved.

[0213] For ultra - large - scale problems (such as the TSP problem with 1000 cities), the parallel acceleration effect is more significant, reducing the solution time from the hour level to the minute level, greatly improving the practicality of the method of the present invention.

[0214] Example 5: Adaptive Quantum Amplitude Adjustment

[0215] This example details the implementation method and its effects of the adaptive adjustment module 7. Adaptive quantum amplitude adjustment is one of the key innovations of the present invention, which can effectively balance local search and global exploration.

[0216] In this example, the adaptive fitness function takes the following form:

[0217] ,

[0218] where is the vibration amplitude information of particle , represents calculating the fitness of particle through the fitness function, is the population size.

[0219] The adaptive factor is dynamically adjusted with iterations:

[0220] ,

[0221] where is the iteration number, is the maximum adaptive factor, is the minimum adaptive factor, is the attenuation coefficient. This setting makes the algorithm more inclined to global exploration in the early stage and more focused on local optimization in the later stage.

[0222] To verify the effect of adaptive adjustment, this embodiment conducts comparative tests on different types of optimization problems. The results show that, compared with fixed parameters, the adaptive adjustment mechanism increases the convergence speed by an average of 35% and improves the quality of the solution by approximately 20%. Especially in the optimization problem of complex multimodal functions, the adaptive adjustment significantly reduces the probability of the algorithm falling into local optima, from approximately 40% to approximately 10%.

[0223] In summary, the quantum-inspired combinatorial optimization solution method and system provided by the present invention construct an efficient, flexible, and robust optimization solution framework by integrating the concept of quantum computing and the advantages of classical optimization algorithms. This method demonstrates excellent performance in various complex combinatorial optimization problems and has broad application prospects.

[0224] It should be noted that the above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the principles of the present invention shall be included within the protection scope of the present invention.

Claims

1. A quantum-inspired combinatorial optimization solution method, characterized in that Including: The step of generating an initial population, which generates an initial population by using a quantum probability distribution method according to the settings of the optimization problem; The population update step, which updates the initial population based on a quantum behavior simulation method; The quality evaluation step, which introduces a multi-objective quality evaluation method to calculate the quality score of each solution in the updated population; The crossover and mutation step, which performs probabilistic crossover and mutation on the updated population based on the quantum behavior simulation method and the quality score; The termination judgment step, which judges whether the loop termination condition is reached. If so, the optimal solution is output. Otherwise, it returns to the population update step.

2. The quantum-inspired combinatorial optimization solution method according to claim 1, wherein The step of generating the initial population specifically includes: Converting the optimization problem into a standard form, setting the problem dimension, and determining the crossover probability and mutation probability; Generating the first solution through quantum behavior simulation, setting the initial population in a quantum state representation, and calculating the wave function as a linear superposition of the initial problem solution space; Selecting the quantum entanglement coefficient, performing quantum state interference operations on a single random quantum state and the quantum entanglement basis, and obtaining the complete initial population through quantum behavior simulation observation.

3. The quantum-inspired combinatorial optimization solution method according to claim 2, wherein The population update step specifically includes: Mapping each solution in the population to a particle in the solution space, and the particle state is composed of coefficients; Updating the particle state through quantum state evolution and the rotation gate of the solution space; Selecting a particle set from the initial solution set and selecting the characteristics corresponding to the target solution in the particle set; Obtaining all the next particle states through the quantum entanglement state; Updating each quantum state through the rotation gate of the solution space to obtain the updated particle state.

4. The quantum-inspired combinatorial optimization solution method according to claim 3, wherein The quality evaluation step specifically includes: Determining the minimum cost or benefit range of individuals in the problem solution space; Calculating the weight value of each evaluation index; Calculating the score of the multi-objective quality of the initial population to generate a comprehensive quality score; Sorting the solutions in the population based on the quality score.

5. The quantum-inspired combinatorial optimization solving method according to claim 4, wherein The probabilistic crossover in the crossover and mutation step specifically includes: Two particles in the parent generation generate new particles through particle-to-particle crossover operations; Performing particle-to-particle crossover operations on the particle set in the parent population to generate a child particle set; Among them, the crossover operation performs information exchange based on a randomly selected cut point position.

6. The quantum-inspired combinatorial optimization solving method according to claim 5, wherein The probabilistic mutation in the crossover and mutation step specifically includes: Performing mutation operations on each solution in the population, and after mutation, new solutions are obtained to form a child population; Among them, the mutation probability is proportional to the quality evaluation score; The mutation operation is realized through quantum random walk to enhance the exploration ability of the search space.

7. The quantum-inspired combinatorial optimization solution method according to claim 6, wherein The termination judgment step specifically includes: Setting the loop termination condition as the threshold of the population quality score; When the termination condition is reached, the optimal solution is output, and the quality of the solution is evaluated by outputting the multi-objective evaluation score; Dynamically adjusting the search strategy according to the quality score threshold; When the quality score is high, it indicates that the current search direction is effective, and the update step size and weight value are small. Otherwise, they are large.

8. The quantum-inspired combinatorial optimization solving method according to claim 7, wherein It also includes: Implementing parallel computing through quantum behavior simulation, and generating an initial particle set through observing quantum behavior simulation; Evolving and observing the particle set to generate new particles; For each observation result, selecting the characteristics of the target solution from the solution space of the current population; Observing the evolution of the solution space through quantum behavior simulation, observing the particles in the population, and obtaining new quantum states.

9. The quantum-inspired combinatorial optimization solution method according to claim 8, characterized in that It also includes: Introduce an adaptive quantum amplitude adjustment mechanism to balance local search and global exploration; Among them, the quantum amplitude adjustment of particles is controlled by an adaptive factor, and the adaptive factor is dynamically calculated according to the difference between the local optimal energy and the global optimal point; Based on the quantum memory effect, incorporate historical search information into the current search strategy to avoid repeated exploration of known regions.

10. A quantum-inspired combinatorial optimization solving system for executing the method according to any one of claims - 9, comprising: An initial generation module for generating an initial population using a quantum probability distribution method according to the settings of the optimization problem; An update module for updating the initial population based on a quantum behavior simulation method; A quality evaluation module for introducing a multi-objective quality evaluation method to calculate the quality score of each solution in the updated population; A crossover and mutation module for performing probabilistic crossover and mutation on the updated population based on the quantum behavior simulation method and the quality score; A termination judgment module for judging whether the loop termination condition is reached. If so, output the optimal solution; otherwise, return to the update module for the next loop.

Images