A network update driven adaptive genetic clustering optimization method, device and medium

Abstract

The application provides a network updating driven adaptive genetic clustering optimization method, device and medium, and belongs to the field of unmanned underwater vehicle cluster cooperative control.The method takes the current position and residual energy of the AUV as input, adopts individual coding composed of cluster head vector I and membership vector M, calculates the relative position change according to the node motion state at each network update, adaptively sets the mutation rate, and performs multi-generation genetic optimization within a single network update.The fitness function comprehensively considers the communication distance, cluster head energy, cluster size uniformity and cluster head energy balance;in the selection stage, the elite reservation and roulette are combined;in the crossover stage, the dynamic crossover rate is adopted with the generation number decreasing;in the mutation stage, the exchange mutation and constraint repair mechanism are adopted;at the same time, the excellent individuals of the last network update are inherited and random individuals are injected, so that the solving efficiency and stability under the dynamic topology are improved.

Description

Technical Field

[0001] This invention relates to the fields of underwater mobile communication network optimization, unmanned underwater vehicle (UUV) swarm collaborative control, and intelligent optimization algorithm application technology, and in particular to a network update-driven adaptive genetic clustering optimization method, device, and medium for dynamic mobile AUV swarms. Background Technology

[0002] Autonomous Underwater Vehicles (AUVs) have wide applications in scenarios such as ocean observation, underwater communication relay, collaborative detection, target search, and distributed operations. With the expansion of mission areas and the limitations of single-node perception and communication capabilities, forming mobile self-organizing networks of multiple AUVs and organizing and managing them through clustering has become an important technical approach to improve networking efficiency and mission coordination.

[0003] In existing clustering methods, common static clustering or one-time optimization schemes typically assume that node positions remain unchanged over a long period of time, or complete cluster head selection and member allocation only under a single network topology. When AUVs move continuously, causing frequent changes in network topology, the clustering scheme obtained at the previous moment quickly becomes invalid, easily leading to problems such as unsuitable cluster head positions, significantly increased intra-cluster communication distances, unbalanced cluster sizes, and uneven energy consumption.

[0004] Genetic algorithms, due to their good global search capabilities, are often used for cluster head selection, path planning, or resource allocation optimization. However, traditional genetic algorithms still have significant shortcomings in dynamic network environments: First, the crossover and mutation rates often use fixed parameters, making it difficult to adaptively adjust the search intensity according to the degree of network topology change; second, there is a lack of an effective mechanism for the inheritance of good individuals between adjacent network update times, resulting in almost starting from scratch each time the solution is recalculated; third, invalid individuals are easily generated after crossover and mutation, such as the number of cluster heads not meeting the constraints, ordinary members pointing to invalid cluster heads, or the cluster size exceeding the threshold, thereby reducing the effective search efficiency.

[0005] Therefore, there is an urgent need for a genetic clustering optimization scheme that can be tightly coupled with the dynamic network update process, so that the genetic solution is no longer limited to one-time offline calculation under static topology, but can be continuously updated, adaptively adjusted and constraint consistency repaired as the AUV cluster position changes, thereby improving the real-time clustering performance in dynamic underwater network scenarios. Summary of the Invention

[0006] The purpose of this invention is to address the problems of fixed genetic clustering parameters, insufficient response to topology changes, and a large number of invalid individuals in existing technologies for dynamic mobile networks. This invention proposes a network update-driven adaptive genetic clustering optimization method, device, and medium for dynamic mobile AUV clusters.

[0007] This invention is achieved through the following technical solution: This invention proposes a network update-driven adaptive genetic clustering optimization method for dynamic mobile AUV clusters, the method comprising: Step 1: Obtain the position parameters and remaining energy parameters of N AUV nodes at the current network update time, and construct an individual code consisting of cluster head vector I and membership vector M, where I is used to characterize whether each AUV is a cluster head, and M is used to characterize the cluster head to which a non-cluster head AUV belongs; Step 2: Generate the initial population, or build the current population based on the superior individuals from the previous network update when it is not the first network update; Step 3: Calculate the relative position change Δ of the current AUV position with respect to the reference position during the network update phase. rel and based on the Δ rel Adaptively determine the mutation rate corresponding to the current network update; Step 4: Perform multi-generational genetic optimization within a single network update, which includes fitness evaluation, selection, crossover, mutation, and repair; Step 5: Output the optimal clustering scheme under the current network update, and update the reference position or reference distance matrix until the preset number of network updates is reached, and output the final clustering result.

[0008] Furthermore, the fitness function F of the individual is constructed as follows: in, This represents the average communication distance from a regular node to its cluster head. This represents the average remaining energy at the cluster head. The standard deviation of the size of each cluster node, The standard deviation of the remaining energy of each cluster head is represented. , , , These are the weighting coefficients.

[0009] Furthermore, the selection operation employs a combination of elite retention and roulette wheel selection, where a predetermined proportion of individuals with the highest fitness are selected as elite individuals and directly enter the next generation, while the remaining individuals are selected based on roulette wheel probabilities.

[0010] Furthermore, the crossover operation employs single-point crossover and a dynamic crossover rate that decreases with the number of generations. The dynamic crossover rate satisfies: Where T is the current algebra, The maximum number of generations within a single network update. The minimum threshold for the crossover rate. This represents the initial crossover rate.

[0011] Furthermore, the relative position change Δ rel The dynamic variability rate is determined by the difference between the pairwise distance matrix of the current AUV and the pairwise distance matrix of the reference AUV, and is determined as follows: : , in, Indicates the network number Next update w is the initial mutation rate. rel Weights for relative position changes. and These are the lower and upper limits of the mutation rate, respectively, and clip(·) represents the clipping function. This represents the change in relative position. The pairwise distance matrix of AUVs under the reference network topology.

[0012] Furthermore, the mutation operation employs exchange mutation, which swaps the cluster head identifiers and membership relationships at two different gene loci in an individual, and performs constraint repair after the exchange.

[0013] Furthermore, the repair includes: Adjust the number of cluster heads to the preset number of cluster heads K; Correct invalid memberships to memberships to the nearest valid cluster head; The number of nodes is lower than the lower limit N. min Clusters can expand by borrowing nodes from other clusters; and the number of nodes can be increased beyond the upper limit N. max Nodes in the cluster are reassigned to a target cluster that meets the capacity requirements and has a better distance.

[0014] Furthermore, during non-first network updates, the current population is constructed in the following way: some superior individuals from the previous network update are retained, the best individuals from the previous network update are injected into the current population, and random individuals are added according to a preset refresh ratio, so as to balance the use of historical experience and population diversity.

[0015] The present invention also proposes an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the network update-driven adaptive genetic clustering optimization method for dynamic mobile AUV clusters.

[0016] The present invention also proposes a computer-readable storage medium for storing computer instructions, which, when executed by a processor, implement the steps of the network update-driven adaptive genetic clustering optimization method for dynamic mobile AUV clusters.

[0017] The beneficial effects of this invention are: 1. This invention abstracts the dynamic movement process of AUV into a two-layer optimization framework of "outer network update - inner genetic solution", which can continuously output available clustering schemes during network topology changes, which is significantly better than the method of solving only a single static topology at once.

[0018] 2. The variation rate of this invention is expressed as a relative position change Δ rel As the driving variable, it can automatically adjust the search intensity according to the degree of topology change, so that the algorithm has both the ability to respond quickly to drastic network changes and the ability to track slowly changing networks stably.

[0019] 3. This invention adopts a dual-vector encoding method consisting of cluster head vector I and membership vector M, which has a clear structure and is convenient for directly expressing cluster head selection and member affiliation; combined with elite retention, roulette wheel selection, single-point crossover, exchange mutation and constraint repair, it can improve the proportion of effective individuals and search efficiency.

[0020] 4. This invention integrates average communication distance, cluster head remaining energy, cluster size balance, and cluster head energy balance into a unified fitness evaluation model, which can obtain clustering results that better meet the needs of dynamic mobile underwater networking.

[0021] 5. This invention improves the robustness and practical value of the algorithm in dynamic networking scenarios by maintaining the continuity of the solution while avoiding premature population convergence through the excellent individual inheritance and random individual injection mechanism across network updates. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0023] Fig. 1 Flowchart of a network update-driven adaptive genetic clustering optimization method for dynamic mobile AUV clusters; Fig. 2 This is a graph of network evolution metrics, including fitness, update time, crossover rate, and mutation rate. Detailed Implementation

[0024] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0025] This invention proposes a network update-driven adaptive genetic clustering optimization method, device, and medium for dynamic mobile AUV clusters. At each network update time, the current position and remaining energy of all AUV nodes are first obtained, and an individual code consisting of a cluster head vector I and a membership vector M is constructed. Then, the population is evaluated using a fitness function that considers the average communication distance within the cluster, the average remaining energy of the cluster heads, cluster size balance, and cluster head energy balance. Next, a multi-generation genetic search is performed within a single network update, where the crossover rate decreases with each generation, and the mutation rate is dynamically adjusted according to the degree of change in relative network position. Finally, constraints are repaired on the individuals after the genetic operation, and the superior individuals from the current network update are inherited to continue solving at the next network update time.

[0026] Specifically, in combination Figs. 1-2 This invention proposes a network update-driven adaptive genetic clustering optimization method for dynamic mobile AUV clusters, the method comprising: Step 1: Obtain the position parameters and remaining energy parameters of N AUV nodes at the current network update time, and construct an individual code consisting of cluster head vector I and membership vector M, where I is used to characterize whether each AUV is a cluster head, and M is used to characterize the cluster head to which a non-cluster head AUV belongs; Step 2: Generate the initial population, or build the current population based on the superior individuals from the previous network update when it is not the first network update; Step 3: Calculate the relative position change Δ of the current AUV position with respect to the reference position during the network update phase. rel and based on the Δ rel Adaptively determine the mutation rate corresponding to the current network update; Step 4: Perform multi-generational genetic optimization within a single network update, which includes fitness evaluation, selection, crossover, mutation, and repair; Step 5: Output the optimal clustering scheme under the current network update, and update the reference position or reference distance matrix until the preset number of network updates is reached, and output the final clustering result.

[0027] Furthermore, the fitness function F of the individual is constructed as follows: in, This represents the average communication distance from a regular node to its cluster head. This represents the average remaining energy at the cluster head. The standard deviation of the size of each cluster node, The standard deviation of the remaining energy of each cluster head is represented. , , , These are the weighting coefficients. In the fitness function, the weighting coefficients are... Characterize the target for communication distance, in order to Characterizing cluster head energy targets, in Characterizing the cluster size equilibrium objective, with Characterize the cluster head energy balance target, so that the solved clustering scheme takes into account communication efficiency, energy utilization and load balance.

[0028] Furthermore, the selection operation employs a combination of elite retention and roulette wheel selection, where a predetermined proportion of individuals with the highest fitness are selected as elite individuals and directly enter the next generation, while the remaining individuals are selected based on roulette wheel probabilities.

[0029] Furthermore, the crossover operation employs single-point crossover and a dynamic crossover rate that decreases with the number of generations. The dynamic crossover rate satisfies: Where T is the current algebra, The maximum number of generations within a single network update. The minimum threshold for the crossover rate. The initial crossover rate is used. A dynamic decreasing strategy is employed for the crossover rate to maintain strong recombination capabilities in the early stages of the search, while gradually converging in the later stages to stabilize a good solution.

[0030] Furthermore, the relative position change Δ rel The dynamic variability rate is determined by the difference between the pairwise distance matrix of the current AUV and the pairwise distance matrix of the reference AUV, and is determined as follows: : , in, Indicates the network number Next update w is the initial mutation rate. rel Weights for relative position changes. and These are the lower and upper limits of the mutation rate, respectively, and clip(·) represents the clipping function. This represents the change in relative position. Let be the pairwise distance matrix of AUVs in the reference network topology. The mutation rate does not depend solely on the number of generations, but rather on the change in relative position Δ of the current network at the time of update relative to the reference topology. rel The greater the network change, the higher the mutation rate, thus enhancing the algorithm's ability to escape the local optima of the old topology; when the network change is small, the mutation rate is reduced to avoid excessive damage to the existing good structure.

[0031] Where Δ rel The relative position change is given by the difference between the current distance matrix and the reference distance matrix:

[0032] in Indicates the first The distance matrix at the next update This represents the reference distance matrix.

[0033] Furthermore, the mutation operation employs exchange mutation, swapping the cluster head identifiers and membership relationships at two different gene loci within an individual, and performing constraint repair after the exchange. The constraint repair includes three types of operations: cluster head number repair, membership relationship repair, and cluster size repair, used to ensure that the individuals obtained after the genetic operation always satisfy the fixed cluster head number, effective cluster head pointers, and upper and lower bounds on the cluster size.

[0034] Furthermore, the repair includes: Adjust the number of cluster heads to the preset number of cluster heads K; Correct invalid memberships to memberships to the nearest valid cluster head; The number of nodes is lower than the lower limit N. min Clusters can expand by borrowing nodes from other clusters; and the number of nodes can be increased beyond the upper limit N. max Nodes in the cluster are reassigned to a target cluster that meets the capacity requirements and has a better distance.

[0035] Furthermore, during non-first network updates, the current population is constructed in the following way: some superior individuals from the previous network update are retained, the best individuals from the previous network update are injected into the current population, and random individuals are added according to a preset refresh ratio, so as to balance the use of historical experience and population diversity.

[0036] This invention proposes a network update-driven adaptive genetic clustering optimization method, device, and medium for dynamic mobile AUV clusters. The method takes the current position and remaining energy of the AUV as input, employs an individual encoding composed of a cluster head vector I and a membership vector M, and calculates the relative position change Δ based on the node's motion state during each network update. rel The mutation rate is then adaptively set, and multi-generation genetic optimization is performed within a single network update. The fitness function integrates communication distance, cluster head energy, cluster size uniformity, and cluster head energy balance. The selection phase combines elite retention with roulette wheel selection. The crossover phase employs a dynamic crossover rate that decreases with each generation. The mutation phase utilizes exchange mutation and constraint repair mechanisms. Simultaneously, by inheriting superior individuals from the previous network update and injecting random individuals, the solution efficiency and stability under dynamic topology are improved. This method is applicable to real-time clustering, cluster head selection, and network optimization scenarios in mobile underwater communication networks.

[0037] The following is in conjunction with the appendix Figs. 1-2The preferred embodiments of the present invention will be described herein. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0038] Example 1: This invention proposes a network update-driven adaptive genetic clustering optimization method for dynamic mobile AUV clusters. This embodiment adopts a two-layer solution framework of "outer layer network update and inner layer genetic optimization". The outer layer is used to describe the network topology changes of AUV nodes during continuous movement, and the inner layer is used to perform genetic optimization on the current clustering result at each network update.

[0039] Let the total number of AUV nodes be N, the preset number of cluster heads be K, and the minimum number of nodes per cluster be N. min The maximum number of nodes in a single cluster is N. max The number of genetic iterations within a single network update is T The number of network updates is U, and the population size is P. The current position of the i-th AUV at the time of the u-th network update is denoted as […]. The corresponding remaining energy is denoted as The positions of all AUVs can be written as a position matrix. .

[0040] Step 1: Individual Encoding and Initial Population Construction In this embodiment, a single candidate clustering scheme is represented by a dual-vector encoding method as X={I,M}. Where I=[I1,I2,...,I...] N Let be the cluster head vector, satisfying I i ∈{0,1}; when I i =1 indicates that the i-th AUV is selected as the cluster head, when I i =0 indicates that the i-th AUV is a normal member. M=[M1,M2,...,M...] N ] is the membership vector; when I i When =1, let M i =-1; when I i When =0, M i =h indicates that the i-th AUV belongs to the cluster head node with index h. Therefore, the cluster head set H = {i | I} can be obtained. i =1}, and satisfy |H|=K.

[0041] For the h-th cluster head, its set of nodes within the cluster is denoted as C. h ={h}∪{i | I i =0, M i =h}, the corresponding cluster size is denoted as n. h =|C hTo ensure that the obtained codes can be used for subsequent fitness assessment and genetic evolution, this embodiment requires that any feasible individual satisfy the following constraints: (1) Fixed cluster head number constraint:

[0042] (2) Effective membership constraint: If I i =0, then M i ∈H; if I i =1, then M i =-1; (3) Cluster size constraint: For any h∈H, N min ≤ n h ≤ N max N min and N max These represent the upper and lower limits of the cluster size, respectively.

[0043] At the first network update time u=1, the initial population is generated by the random individual generation module. Then, constraint repair is performed on each individual to ensure that all individuals satisfy the above three constraints.

[0044] Step 2: Constructing the fitness function To simultaneously consider communication cost, cluster head energy, and cluster balance, this embodiment constructs the following comprehensive fitness function: (1) in, , , , For the preset weighting coefficients, and + + + =1. The terms in the formula are defined as follows: (1) Average communication distance D avg The average communication cost between a regular member node and its cluster head can be expressed as: (2) (2) Average residual energy E of the cluster head h The energy base used to measure the forwarding and management tasks currently undertaken by the cluster head can be expressed as: (3) (3) Cluster size standard deviation The measure of the balance in the number of nodes across clusters can be expressed as: (4) (4) Standard deviation of residual energy in cluster head The degree of dispersion of remaining energy among different cluster heads can be expressed as: (5) From equations (1) to (5), it can be seen that the larger the fitness value, the shorter the average distance from the ordinary node to the cluster head, the higher the average energy of the cluster head, the more balanced the cluster size, and the more balanced the energy distribution of the cluster head. Therefore, the corresponding clustering scheme is better.

[0045] Step 3: Outer Network Update Process At the start of the u-th network update, if u=1, then the current reference position matrix is ​​taken from the initial position matrix P. ref =P (1) And cause the relative position to change If u>1, then each node is first advanced from the previous network update time to the current time according to the AUV motion model.

[0046] In a preferred implementation, the AUV cluster center is located at the unit direction vector. Above speed After translation, the cluster center in the s-th discrete motion step can be represented as: (6) in, Let be the cluster center position during the s-th motion step in the u-th network update process. Let the initial position of the i-th AUV relative to the cluster center be denoted as . After considering the smoothing perturbation, the candidate position of the i-th AUV in this motion step can be written as: (7) in, A periodic perturbation term related to the time step and node number is used to simulate local fluctuations during the smooth movement of the formation. Furthermore, when the distance between any two nodes is less than the minimum safe distance d... min At this time, a repulsion correction term is introduced to adjust the candidate positions, thereby ensuring that nodes do not cluster too closely together.

[0047] Obtain the position matrix P at the current network update time. (u) Next, calculate the relative positional changes between the current topology and the reference topology. First, define the pairwise distance matrix D. (u) =[ ] (N×N) and reference distance matrix D ref =[ ] (N×N) ,in: (8) Therefore, the change in relative position is defined as: (9) At the same time, calculate the reference average pairwise distance: (10) Equation (9) reflects the overall degree of change of the current network topology relative to the reference topology; The larger the value, the more significant the change in network structure.

[0048] Step 4: Dynamic Mutation Rate and Current Population Preparation This embodiment does not use a fixed mutation rate, but rather adaptively adjusts the mutation rate at the u-th network update time based on the relative position change obtained from equation (9). Dynamic mutation rate Represented as: (11) in, Indicates the network number Next update w is the initial mutation rate. rel Weights for relative position changes. and Here, represents the lower and upper limits of the mutation rate, respectively, and clip(·) represents the clipping function. From equation (11), it can be seen that when the current network differs significantly from the reference network, The corresponding increase is made to improve the search's ability to escape local optima in the old topology; when the current network changes little, The mutation rate is kept at a low level to protect the existing excellent structure. This dynamic mutation rate calculation method is consistent with the design in the source code that "updates the mutation rate according to the network update time rather than according to the small displacement of a single generation".

[0049] For population preparation at the current network update time, if u=1, then the initial population that has been randomly initialized and repaired is directly used. If u>1, then for the population after the last network update... Perform inheritance-based construction. Specifically, let the random refresh ratio be... Then the number of individuals retained is: (12) First, obtain the population from the previous time step through a selection operation. First, select the better individual; second, select the locally best individual from the previous time step. Inject into the current population; finally, fill the remaining positions by randomly generating and repairing individuals, thus forming the initial population at the current network update time. This mechanism is able to maintain population diversity while utilizing historically favorable structures.

[0050] Step 5: Inner Layer Genetic Optimization Process At the u-th network update time, with Perform genetic optimization for the initial population in generations T. For the t-th generation population... First, calculate the fitness value one by one. And obtain the contemporary optimal fitness: (13) Simultaneously record the current average fitness: (14) like If the fitness is greater than the global historical best, then update the global best individual and its corresponding node position.

[0051] Step 5.1 Select Operation The selection phase employs a combination of "elite retention + roulette wheel" method. Let the elite ratio be... Then the number of elite individuals is First, sort by fitness from highest to lowest, and keep the top results. First, select one elite individual; second, sample the remaining individuals according to the probability of roulette, where the probability of the q-th individual being selected is: (15) This yields the selected population. This step ensures that high-performing individuals are not compromised while maintaining opportunities for non-elite individuals to compete.

[0052] Step 5.2 Dynamic Crossover Operation For the selected population Individuals are randomly shuffled and paired up. The crossover rate is not fixed but decreases gradually with each generation within a single network update, expressed as: (16) in, This is the initial crossover rate. For each pair of parent individuals X... a ={I a M a} and X b ={I b M b Randomly generate intersection points. ,satisfy Then, single-point crossover is performed on the cluster head vector and membership vector respectively to obtain two offspring: (17) (18) If the random number rand∈[0,1] is less than If the crossover is successful, then equations (17) and (18) are executed to complete the crossover; otherwise, the parent generation is directly retained. After the crossover is completed, constraint repair is immediately performed on the generated child generation.

[0053] Step 5.3 Exchange Mutation Operation For the crossover population The current network update mutation rate is given by equation (11). Check each random number individually to see if it has mutated. If the random number rand∈[0,1] is less than... Then from Two different positions a and b are randomly selected in the individual's code, and the two positions are swapped and mutated: (19) Where I' and M' represent the cluster head vector and membership vector after the crossover, respectively. After the exchange, for ordinary member nodes whose original membership becomes invalid due to the change in cluster head status, their membership is updated according to the principle of "most recent valid cluster head first"; if a node is exchanged to become a cluster head, its membership value M is directly set. i =-1. Then perform constraint repair.

[0054] Step Six: Constraint Repair Process Because crossover and mutation may cause individuals to no longer meet requirements such as the fixed number of cluster heads, effective membership relationships, and upper and lower limits of cluster size, a repair operation is performed after each crossover and mutation. The repair process includes at least the following three sub-steps: (1) Cluster head number repair: If the current total number of cluster heads is greater than K, then the excess cluster heads are removed from the current cluster head node; if it is less than K, then new cluster heads are selected from non-cluster head nodes until the fixed cluster head number constraint is met.

[0055] (2) Membership repair: For any ordinary member node i, if M i H is then reassigned to the nearest valid cluster head, i.e.: (20) For any cluster head node h, force M to... h =-1.

[0056] (3) Cluster size repair: For any cluster head h, if n h <N min If n h >N max If this is not possible, some member nodes in the cluster will be reassigned to other cluster heads that meet the capacity requirements and are closer in distance. After this step, it is guaranteed that all clusters meet the upper and lower limits of cluster size constraints.

[0057] After the repair is completed, the next generation population is obtained . After t iterates from 0 to T - 1, the local optimal individual at the u-th network update time is denoted as , and the corresponding optimal fitness is denoted as: (21) Step Seven: End of Network Update and Update of Reference State When the inner genetic optimization at the u-th network update time ends, the current local optimal individual is output as the optimal clustering scheme at this network time, and its fitness value is recorded . At the same time, the current position matrix P (u) is updated to the reference position matrix for the next network update time, that is, let P ref ←P (u) , and the reference distance matrix D ref is recalculated according to this reference position for use in calculating the relative position change and dynamic mutation rate using equations (9) and (11) during the next network update

[0058] If u < U, return to the outer network update process, first perform the smooth movement of the AUV, and then enter the genetic optimization at the (u + 1)-th network update time; if u = U, the algorithm terminates, and the optimal individual of the last network update is output and its fitness value , and the corresponding cluster head set and the membership relationship of ordinary members to the cluster heads are given

[0059] Example 2: This preferred embodiment provides an electronic device, which includes a processor and a memory. A computer program is stored in the memory, and the computer program is loaded and executed by the processor to implement any one of the above-mentioned network update-driven adaptive genetic clustering optimization methods for dynamic mobile AUV clusters. This computer device can achieve the beneficial effects of the network update-driven adaptive genetic clustering optimization method for dynamic mobile AUV clusters provided by the embodiments of the present invention. For details, see the previous embodiments and will not be repeated here

[0060] Example 3: Those of ordinary skill in the art can understand that all or part of the steps in the above-mentioned various methods can be completed by instructions or by controlling related hardware through instructions. The instructions can be stored in a computer-readable storage medium and loaded and executed by a processor

[0061] To this end, embodiments of the present invention provide a computer-readable storage medium storing a computer program, which is loaded and executed by a processor to implement the steps of any embodiment of the network update-driven adaptive genetic clustering optimization method for dynamic mobile AUV clusters.

[0062] The memory in this application embodiment can be volatile memory or non-volatile memory, or it can include both volatile and non-volatile memory. The non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. The volatile memory can be random access memory (RAM), which is used as an external cache. By way of example, but not limitation, many forms of RAM are available, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDRSDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous linked dynamic random access memory (SLDRAM), and direct rambus RAM (DR RAM). It should be noted that the memory used in the methods described in this invention is intended to include, but is not limited to, these and any other suitable types of memory.

[0063] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions. When the computer instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium accessible to a computer or a data storage device such as a server or data center that integrates one or more available media. The available media may be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., high-density digital video discs (DVDs)), or semiconductor media (e.g., solid-state disks (SSDs)).

[0064] In implementation, each step of the above method can be completed by integrated logic circuits in the processor's hardware or by instructions in software. The steps of the method disclosed in the embodiments of this application can be directly implemented by a hardware processor, or by a combination of hardware and software modules in the processor. The software modules can reside in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. This storage medium is located in memory, and the processor reads information from the memory and, in conjunction with its hardware, completes the steps of the above method. To avoid repetition, detailed descriptions are omitted here.

[0065] It should be noted that the processor in the embodiments of this application can be an integrated circuit chip with signal processing capabilities. During implementation, each step of the above method embodiments can be completed by the integrated logic circuitry in the processor's hardware or by instructions in software form. The processor can be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this application can be directly embodied as execution by a hardware decoding processor, or as a combination of hardware and software modules in the decoding processor. The software modules can be located in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. This storage medium is located in memory, and the processor reads the information in the memory and, in conjunction with its hardware, completes the steps of the above methods.

[0066] The foregoing has provided a detailed description of the network update-driven adaptive genetic clustering optimization method, device, and medium for dynamic mobile AUV clusters proposed in this invention. Specific examples have been used to illustrate the principles and implementation methods of this invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of this invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this invention. Therefore, the content of this specification should not be construed as a limitation of this invention.

Claims

1. A network update-driven adaptive genetic clustering optimization method, characterized in that, The method includes: Step 1: Obtain the position parameters and remaining energy parameters of N AUV nodes at the current network update time, and construct an individual code consisting of cluster head vector I and membership vector M, where I is used to characterize whether each AUV is a cluster head, and M is used to characterize the cluster head to which a non-cluster head AUV belongs; Step 2: Generate the initial population, or build the current population based on the superior individuals from the previous network update when it is not the first network update; Step 3: Calculate the relative position change Δ of the current AUV position with respect to the reference position during the network update phase. rel and based on the Δ rel Adaptively determine the mutation rate corresponding to the current network update; Step 4: Perform multi-generational genetic optimization within a single network update, which includes fitness evaluation, selection, crossover, mutation, and repair; Step 5: Output the optimal clustering scheme under the current network update, and update the reference position or reference distance matrix until the preset number of network updates is reached, and output the final clustering result.

2. The method according to claim 1, characterized in that, The fitness function F of the individual is constructed as follows: in, This represents the average communication distance from a regular node to its cluster head. This represents the average remaining energy at the cluster head. The standard deviation of the size of each cluster node, The standard deviation of the remaining energy of each cluster head is represented. , , , These are the weighting coefficients.

3. The method according to claim 1, characterized in that, The selection operation employs a combination of elite retention and roulette wheel selection, where individuals with the highest fitness, representing a predetermined proportion, directly enter the next generation as elite individuals, while the remaining individuals are selected based on roulette wheel probabilities.

4. The method according to claim 1, characterized in that, The crossover operation employs single-point crossover and a dynamic crossover rate that decreases with the number of generations. The dynamic crossover rate satisfies: Where T is the current algebra, The maximum number of generations within a single network update. The minimum threshold for the crossover rate. This represents the initial crossover rate.

5. The method according to claim 1, characterized in that, The relative position change Δ rel The dynamic variability rate is determined by the difference between the pairwise distance matrix of the current AUV and the pairwise distance matrix of the reference AUV, and is determined as follows: : , in, Indicates the network number Next update w is the initial mutation rate. rel Weights for relative position changes. and These are the lower and upper limits of the mutation rate, respectively, and clip(·) represents the clipping function. This represents the change in relative position. The pairwise distance matrix of AUVs under the reference network topology.

6. The method according to claim 1, characterized in that, The mutation operation employs exchange mutation, which swaps the cluster head identifiers and membership relationships at two different gene loci in an individual, and performs constraint repair after the exchange.

7. The method according to claim 1, characterized in that, The repair includes: Adjust the number of cluster heads to the preset number of cluster heads K; Correct invalid memberships to memberships to the nearest valid cluster head; The number of nodes is lower than the lower limit N. min Clusters can expand by borrowing nodes from other clusters; and the number of nodes can be increased beyond the upper limit N. max Nodes in the cluster are reassigned to a target cluster that meets the capacity requirements and has a better distance.

8. The method according to claim 1, characterized in that, During non-first network updates, the current population is constructed in the following way: some superior individuals from the previous network update are retained, the best individuals from the previous network update are injected into the current population, and random individuals are added according to a preset refresh ratio, so as to balance the use of historical experience and population diversity.

9. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1-8.

10. A computer-readable storage medium for storing computer instructions, characterized in that, When the computer instructions are executed by the processor, they implement the steps of the method according to any one of claims 1-8.

Images