Adaptive optimization method for positioning communication base station based on improved fish swarm algorithm

Abstract

The application discloses a communication base station positioning adaptive optimization method based on an improved fish swarm algorithm, and comprises the following steps: collecting target communication area planning basic data, preprocessing to generate communication planning data set; splitting a candidate base station positioning scheme into coverage, interference, load and constraint search domains, extracting domain features and combining to generate a multi-domain search vector; encoding the candidate base station positioning scheme into a fish swarm individual position vector and constructing an initial fish swarm population; storing the historical single-domain superior scheme according to the multi-domain search vector to generate a multi-domain independent memory pool; calculating the correlation difference of each search domain and constructing an inter-domain conflict matrix to determine the conflict type and conflict strength; extracting and recombining the single-domain superior features to correct the fish swarm individual position vector; performing hierarchical search on the target communication area and iteratively updating to output the optimal base station deployment result. The application improves the communication base station positioning optimization precision, and improves the conflict processing capability and the complex scene adaptability.

Description

Technical Field

[0001] This invention relates to the field of communication positioning optimization technology, and in particular to an adaptive optimization method for communication base station positioning based on an improved fish swarm algorithm. Background Technology

[0002] With the continuous expansion of mobile communication networks, base station location optimization has become a crucial aspect of network planning. The deployment location of base stations directly impacts regional coverage quality, inter-site interference levels, service carrying capacity, and construction efficiency. Existing base station location methods largely rely on manual experience, rule-based selection, or single-objective optimization, determining the deployment plan after conducting coverage analysis, interference analysis, and capacity assessment of candidate sites. While these methods can accomplish basic planning tasks when the network scale is small and the scenario relationships are simple, in target communication areas with dense user distribution, significant differences in service demands, and complex constraints, the base station location process involves the coupled influence of multiple factors such as coverage, interference, load, and constraints. Traditional processing methods struggle to simultaneously address multiple planning objectives.

[0003] Swarm intelligence optimization methods have been introduced into the field of communication base station site selection and positioning optimization. The fish swarm algorithm, with its global search capability, has been used for candidate site optimization. However, most existing applications of the fish swarm algorithm reduce the base station positioning problem to a single fitness search problem, guiding individual fish to update their positions through a unified evaluation result. This approach compresses coverage benefits, interference suppression, load allocation, and deployment constraints into the same search channel, and historical iteration results are mostly saved as the overall optimal result, lacking independent expression and accumulation of different search objectives. Faced with communication planning tasks where multiple objectives have significant conflicts, the single fitness-driven approach easily allows local dominance information to be masked by the overall result, leading to convergence in the update direction of individual fish, making the search process prone to local convergence, and making it difficult to perform differentiated corrections for the main conflict relationship. Furthermore, the differentiated needs between regions are not fully reflected in the unified update process. Based on these problems, an adaptive optimization method for communication base station positioning is needed that can perform domain-specific representation, conflict identification, historical experience accumulation, and hierarchical collaborative search of coverage, interference, load, and constraints. Therefore, how to provide an adaptive optimization method for communication base station positioning based on an improved fish swarm algorithm is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0004] One objective of this invention is to propose an adaptive optimization method for communication base station positioning based on an improved fish swarm algorithm. This invention employs multi-domain decoupling search, inter-domain conflict identification, single-domain advantage memory accumulation, and hierarchical sub-fish swarm collaborative update mechanism. It describes in detail the implementation process of communication base station positioning optimization and has the advantages of high positioning accuracy, strong conflict handling capability, and good adaptability to complex scenarios.

[0005] The adaptive optimization method for communication base station positioning based on the improved fish swarm algorithm according to an embodiment of the present invention includes the following steps:

[0006] Collect basic data for the target communication area planning and preprocess it to generate a communication planning dataset.

[0007] Each candidate base station positioning scheme is divided into a coverage search domain, an interference search domain, a load search domain, and a constraint search domain. Domain features are extracted and combined to generate a multi-domain search vector.

[0008] The candidate base station positioning scheme is encoded into the individual fish population position vector, the initial fish population is constructed, and the multi-domain search vector is associated with each individual fish population.

[0009] Based on the multi-domain search vectors corresponding to each individual fish group, the single-domain dominant solutions in the historical iterations are stored separately according to the search domain, generating a multi-domain independent memory pool.

[0010] Based on the multi-domain search vector calculation, the correlation difference between each search domain is calculated, an inter-domain conflict matrix is ​​constructed, and the conflict type and conflict intensity of the candidate base station positioning scheme for each individual fish group are determined.

[0011] Based on the conflict type and intensity, the fish group individuals are controlled to execute the corresponding split update path. Different single-domain advantageous features are extracted from the multi-domain independent memory pool and recombined to generate a composite guiding vector. The composite guiding vector is used to correct the position vector of the fish group individuals, changing the unified update to conflict-driven split update. The single-domain advantageous features are recombined to guide the correction of the current position, thus establishing a heterogeneous search mechanism oriented towards conflict state.

[0012] The target communication area is divided, and a hierarchical sub-fish swarm cooperative search is performed on the updated candidate base station positioning scheme. The positioning results are fed back to the update process, and the process is iterated until the termination condition is met, and the optimal base station deployment result is output.

[0013] Optionally, the preprocessing specifically includes: parsing the fields of the target communication area planning basic data, and sequentially performing coordinate unification, data cleaning, spatial grid division, location mapping, unit unification and normalization processing on the parsed target communication area planning basic data, and then aggregating and encapsulating the data to generate a communication planning dataset.

[0014] Optionally, the process of generating the multi-domain search vector includes:

[0015] For each candidate base station positioning scheme in the communication planning dataset, the candidate site location, existing base station location, user distribution location, service demand distribution, and constraint condition distribution are read. Centered on the candidate site location, the correspondence between the candidate site location and the user distribution location is divided into the coverage search domain, the correspondence between the candidate site location and the existing base station location is divided into the interference search domain, the correspondence between the candidate site location and the service demand distribution is divided into the load search domain, and the correspondence between the candidate site location and the constraint condition distribution is divided into the constraint search domain. Within the coverage search domain, interference search domain, load search domain, and constraint search domain, the correspondence calculation is performed according to the candidate site order to generate the coverage domain feature sequence, interference domain feature sequence, load domain feature sequence, and constraint domain feature sequence.

[0016] The four feature sequences are aligned and arranged according to the candidate site order under the same candidate base station positioning scheme, keeping the boundaries of each search domain independent, generating multi-domain search vectors, and writing them into the fish swarm individual position vector association table of the corresponding candidate base station positioning scheme to complete the correspondence binding between multi-domain search vectors and fish swarm individual position vectors.

[0017] Optionally, the process of constructing the initial fish population includes:

[0018] According to the candidate site order in the candidate base station positioning scheme, the positions of each candidate site are read sequentially; the spatial coordinates of the candidate site positions are used as position encoding units and concatenated according to the candidate site order to generate the individual fish position vectors corresponding to the candidate base station positioning scheme; based on all candidate base station positioning schemes in the communication planning dataset, the individual fish position vectors are arranged in a set to construct the initial fish population.

[0019] According to the one-to-one correspondence between each candidate base station positioning scheme and the multi-domain search vector, the multi-domain search vector corresponding to each candidate base station positioning scheme is read; the multi-domain search vector is written to the associated position of the fish individual with the same sequence number, using the arrangement order of the fish individual position vector in the initial fish population as the index order; the correspondence verification is performed on each fish individual after writing the multi-domain search vector to complete the association between the multi-domain search vector and each fish individual.

[0020] Optionally, the process of generating a multi-domain independent memory pool includes:

[0021] The fish swarm individual position vectors and corresponding multi-domain search vectors from each historical iteration are read in iterative order. Based on the search domain division results in the multi-domain search vectors, the candidate base station positioning schemes corresponding to each fish swarm individual are divided into coverage search domain segments, interference search domain segments, load search domain segments, and constraint search domain segments. Within the same historical iteration, single-domain comparisons are performed on the coverage search domain segments, interference search domain segments, load search domain segments, and constraint search domain segments to select the single-domain advantageous segments in each search domain. The selected single-domain advantageous segments are bound and recorded with the corresponding historical iteration sequence number, population sequence number, candidate base station positioning scheme identifier, and search domain identifier. According to the search domain identifier, the single-domain advantageous segments that have been bound and recorded are appended to the corresponding independent memory area, keeping the content written in previous historical iterations within the same search domain from being deleted. After each append, the single-domain advantageous segments in the corresponding independent memory area are arranged sequentially according to the historical iteration sequence number and population sequence number to form a continuously accumulated historical experience sequence under the same search domain. The coverage independent memory area, interference independent memory area, load independent memory area, and constraint independent memory area are combined to generate a multi-domain independent memory pool.

[0022] Optionally, the process of constructing the inter-domain conflict matrix includes:

[0023] Read the multi-domain search vectors corresponding to each individual fish in the population group sequentially according to their population numbers. Using the arrangement of the covering search domain, interference search domain, load search domain, and constraint search domain in the multi-domain search vectors of the same individual fish as a unified correspondence benchmark, perform position alignment on the domain vector segments of any two search domains, calculate the difference results item by item according to the corresponding positions, and generate inter-domain difference sequences between each search domain. Perform segment aggregation on each inter-domain difference sequence to obtain the inter-domain difference values ​​between the covering search domain and the interference search domain, between the covering search domain and the load search domain, between the covering search domain and the constraint search domain, between the interference search domain and the load search domain, between the interference search domain and the constraint search domain, and between the load search domain and the constraint search domain. Using the pairwise correspondence of the search domains as the matrix row and column indices, write the inter-domain difference values ​​into the corresponding matrix positions, making the matrix symmetrical. To generate an inter-domain conflict matrix, the inter-domain difference values ​​of all matrix positions in the inter-domain conflict matrix are kept consistent. The inter-domain difference values ​​of all matrix positions in the inter-domain conflict matrix are sorted according to their size. The search domain combination corresponding to the first matrix position in the sorting result is taken as the primary conflict domain pair, and the search domain combination corresponding to the second matrix position in the sorting result is taken as the secondary conflict domain pair. The search domain combination corresponding to the primary conflict domain pair is determined as the conflict type of the current fish population. When the primary and secondary conflict domain pairs contain the same search domain, the search domain combination corresponding to the primary conflict domain pair and the non-repeating search domain combination in the secondary conflict domain pair are determined as the composite conflict type. The inter-domain difference values ​​corresponding to the primary conflict domain pair are read, and combined with the inter-domain difference values ​​corresponding to the secondary conflict domain pairs and the distribution of inter-domain difference values ​​of the remaining matrix positions in the inter-domain conflict matrix, the conflict concentration is calculated to generate the conflict intensity under the current conflict type.

[0024] Optionally, the process of generating the composite guiding vector includes:

[0025] Based on the conflict type and intensity corresponding to the current fish population, the target search domain to be invoked is determined. Single-domain dominant fragments are read from the corresponding independent memory regions of the multi-domain independent memory pool according to the target search domain, and the retrieval is performed sequentially according to the historical iteration number and population number. Single-domain dominant features consistent with the candidate site order in the current fish population's position vector are extracted from the sequential retrieval results. The extracted single-domain dominant features corresponding to different target search domains are aligned according to the candidate site order, maintaining the independence of each target search domain boundary. Based on the conflict type corresponding to the current fish population, the aligned single-domain dominant features are segmented and recombined according to the order of primary conflict domain pairing and secondary conflict domain pairing, generating a composite feature sequence. The composite feature sequence is written into the guide vector construction position according to the arrangement order of the current fish population's position vector, generating a composite guide vector.

[0026] Optionally, the process of establishing a heterogeneous search mechanism oriented towards conflict states includes:

[0027] Based on the conflict type, the position vectors of the current fish swarm individuals are divided into a main conflict domain update segment, a secondary conflict domain update segment, and a non-conflict domain preservation segment. The guide segment corresponding to the main conflict domain update segment in the composite guide vector is written into the main conflict domain update segment, and the guide segment corresponding to the secondary conflict domain update segment in the composite guide vector is written into the secondary conflict domain update segment, while keeping the position content of the non-conflict domain preservation segment unchanged. The correction magnitudes of the main conflict domain update segment and the secondary conflict domain update segment are determined based on the conflict intensity, ensuring that the correction magnitude of the main conflict domain update segment is greater than that of the secondary conflict domain update segment. A principal correction is performed on the main conflict domain update segment after the guide segment is written, and a principal correction is performed on the secondary conflict domain update segment after the guide segment is written. The new segment performs auxiliary correction, backfilling the corrected main conflict domain update segment, secondary conflict domain update segment, and non-conflict domain preservation segment to the current fish population's position vector in the original position order, generating a split update position vector; the split update position vector performs candidate site order consistency check and search domain boundary consistency check. When the check passes, the split update position vector is determined as the corrected fish population's position vector. When the check fails, the composite guide vector is reread according to the original population number, and the segment writing and position backfilling are repeated; the corrected fish population's position vector is written into the hierarchical sub-fish population cooperative search process, and fish populations under different conflict types perform differentiated position correction according to the corresponding update segment.

[0028] Optionally, the process of generating the optimal base station deployment result includes:

[0029] Based on the updated candidate base station positioning scheme, the target communication area is divided into hotspot search areas, bearer search areas, and edge search areas according to user distribution density, service demand distribution, and regional boundary locations. Based on the regional division results of the hotspot search areas, bearer search areas, and edge search areas, position segments corresponding to the areas are extracted from the corrected individual position vectors of the fish swarms to construct hotspot sub-fish swarms, bearer sub-fish swarms, and edge sub-fish swarms respectively. Within the hotspot sub-fish swarms, position correction is performed by prioritizing the composite guiding vector segment corresponding to the load search domain, according to the position segment order within the hotspot search area. Within the bearer sub-fish swarms, position correction is performed by prioritizing the composite guiding vector segments corresponding to the coverage search domain and interference search domain, according to the position segment order within the bearer search area. Within the edge sub-fish swarms, position correction is performed by prioritizing the composite guiding vector segments corresponding to the coverage search domain and constraint search domain, according to the position segment order within the edge search area. This generates the updated sub-fish swarm results for each search area. The updated sub-fish swarm results for each search area are then backfilled into the corrected area according to the spatial position order within the original target communication area. The corresponding region location in the individual fish location vector is used to generate a cooperative localization result. Using the cooperative localization result as input, the domain features corresponding to the coverage search domain, interference search domain, load search domain, and constraint search domain are re-extracted according to the candidate site order. These features are then recombined according to the original search domain boundaries to generate an updated multi-domain search vector. Based on the updated multi-domain search vector, the inter-domain difference values ​​between each search domain are recalculated and written back according to the original matrix positions to generate an updated inter-domain conflict matrix. Based on the updated multi-domain search vector and the updated inter-domain conflict matrix, single-domain advantageous segments in each search domain are re-selected and appended to the corresponding independent memory area according to the search domain to generate an updated multi-domain independent memory pool. It is then determined whether the updated multi-domain search vector, the updated inter-domain conflict matrix, and the updated multi-domain independent memory pool meet the preset termination conditions. If the preset termination conditions are met, the current cooperative localization result is determined as the optimal base station deployment result. If the preset termination conditions are not met, the updated multi-domain search vector, the updated inter-domain conflict matrix, and the updated multi-domain independent memory pool are written into the next round of individual fish update process to continue iterating.

[0030] The beneficial effects of this invention are:

[0031] (1) The present invention changes the communication base station positioning process from a single fitness search to a multi-domain decoupled search process with a coverage search domain, an interference search domain, a load search domain and a constraint search domain. By using multi-domain search vectors to maintain the independent representation relationship of each search domain, the differences between different optimization objectives in the base station positioning process can be identified and participate in subsequent updates. This improves the problem that local effective information is easily obscured after multiple planning factors are uniformly compressed in the prior art, and improves the positioning accuracy and search targeting in complex communication scenarios.

[0032] (2) The present invention constructs an inter-domain conflict matrix, identifies the correlation differences between each search domain, performs conflict-driven split update according to the conflict type and conflict intensity, and then combines the single-domain advantage features in the multi-domain independent memory pool to reorganize and generate a composite guiding vector, and performs directional correction on the position vector of individual fish, so that individual fish no longer update in a uniform way, enhances the response capability to the main conflict relationship, reduces the probability of the search process falling into single path convergence, and improves the conflict handling capability and search stability in the base station positioning optimization process;

[0033] (3) The present invention divides the target communication area into hot spot search area, carrier search area and edge search area, respectively constructs corresponding sub-fish groups to perform hierarchical collaborative search, and feeds back the collaborative positioning results to the update process of multi-domain search vector, inter-domain conflict matrix and multi-domain independent memory pool, so that the positioning optimization in different areas can be independently corrected and backfilled according to the differences in regional needs, which improves the problem that the unified search method in the prior art is difficult to take into account the differentiated needs of hot spot area, regular carrier area and edge area, and improves the adaptive optimization capability and deployment result quality in complex target communication areas. Attached Figure Description

[0034] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0035] Figure 1 The flowchart shows the adaptive optimization method for communication base station positioning based on the improved fish swarm algorithm proposed in this invention.

[0036] Figure 2 This is a schematic diagram of the multi-domain search vector construction process proposed in this invention;

[0037] Figure 3 This is a flowchart of the conflict-driven split update and hierarchical sub-fish swarm cooperative search process proposed in this invention. Detailed Implementation

[0038] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0039] refer to Figures 1-3 An adaptive optimization method for communication base station positioning based on an improved fish swarm algorithm includes the following steps:

[0040] Collect basic data for the target communication area planning and preprocess it to generate a communication planning dataset.

[0041] Each candidate base station positioning scheme is divided into a coverage search domain, an interference search domain, a load search domain, and a constraint search domain. Domain features are extracted and combined to generate multi-domain search vectors. By changing the single fitness search to a multi-domain decoupled search, a parallel representation relationship between coverage, interference, load, and constraints is established.

[0042] The candidate base station positioning scheme is encoded into the individual fish population position vector, the initial fish population is constructed, and the multi-domain search vector is associated with each individual fish population.

[0043] Based on the multi-domain search vectors corresponding to each individual fish group, the single-domain advantage schemes in the historical iterations are stored separately according to the search domain, generating a multi-domain independent memory pool. By changing the global optimal retention to single-domain advantage classification retention, a historical experience accumulation mechanism for different search domains is established.

[0044] Based on multi-domain search vector calculation, the correlation difference between each search domain is calculated, an inter-domain conflict matrix is ​​constructed, the conflict type and conflict intensity of the candidate base station positioning scheme corresponding to each individual fish group are determined, and inter-domain conflict determination is introduced before the fish group is updated to establish a conflict identification mechanism between search domains.

[0045] Based on the conflict type and intensity, the fish group individuals are controlled to execute the corresponding split update path. Different single-domain advantageous features are extracted from the multi-domain independent memory pool and recombined to generate a composite guiding vector. The composite guiding vector is used to correct the position vector of the fish group individuals, changing the unified update to conflict-driven split update. The single-domain advantageous features are recombined to guide the correction of the current position, thus establishing a heterogeneous search mechanism oriented towards conflict state.

[0046] The target communication area is divided into hotspot search area, bearer search area and edge search area. Hierarchical sub-fish swarm cooperative search is performed on the updated candidate base station positioning scheme. The positioning results are fed back to the update process of multi-domain search vector, inter-domain conflict matrix and multi-domain independent memory pool. The process is iterated until the termination condition is met and the optimal base station deployment result is output.

[0047] In this embodiment, the preprocessing specifically includes: parsing the fields of the target communication area planning basic data, and sequentially performing coordinate unification, data cleaning, spatial grid division, location mapping, unit unification and normalization on the target communication area planning basic data after field parsing, and then aggregating and encapsulating it to generate a communication planning dataset; the target communication area planning basic data includes geographical location data, site distribution data, base station distribution data, user distribution data, service requirement data and constraint data.

[0048] In this embodiment, the process of generating multi-domain search vectors includes:

[0049] For each candidate base station positioning scheme in the communication planning dataset, the candidate site location, existing base station location, user distribution location, service demand distribution, and constraint condition distribution are read. Centered on the candidate site location, the correspondence between the candidate site location and the user distribution location is divided into the coverage search domain, the correspondence between the candidate site location and the existing base station location is divided into the interference search domain, the correspondence between the candidate site location and the service demand distribution is divided into the load search domain, and the correspondence between the candidate site location and the constraint condition distribution is divided into the constraint search domain. Within the coverage search domain, interference search domain, load search domain, and constraint search domain, the correspondence calculation is performed according to the candidate site order to generate the coverage domain feature sequence, interference domain feature sequence, load domain feature sequence, and constraint domain feature sequence.

[0050] The specific steps for performing the correspondence calculation are as follows: For the same candidate base station positioning scheme, based on the order of candidate site locations, the associated objects corresponding to the candidate site locations are read sequentially in each search domain. The association processing is performed item by item according to the correspondence between the candidate site locations and the associated objects, and the processing results are written into the feature sequence of the corresponding search domain according to the order of the candidate site locations. Among them, the associated objects in the coverage search domain are the user distribution locations, the associated objects in the interference search domain are the existing base station locations, the associated objects in the load search domain are the service demand distributions, and the associated objects in the constraint search domain are the constraint condition distributions. The four feature sequences are aligned and arranged according to the order of candidate sites under the same candidate base station positioning scheme, keeping the boundaries of each search domain independent, generating multi-domain search vectors, and writing them into the fish swarm individual location vector association table of the corresponding candidate base station positioning scheme to complete the correspondence binding between the multi-domain search vectors and the fish swarm individual location vectors.

[0051] The alignment arrangement is specifically as follows: the order of candidate site locations in the same candidate base station positioning scheme is used as the unified index order, the feature sequences of each search domain are unified, so that the feature results of the same candidate site location in different search domains are in the same index position, and then, under the condition of keeping the boundaries of each search domain independent, they are combined in a preset order to generate a multi-domain search vector.

[0052] In this embodiment, the process of constructing the initial fish population includes:

[0053] According to the candidate site order in the candidate base station positioning scheme, the positions of each candidate site are read sequentially; the spatial coordinates of the candidate site positions are used as position encoding units and concatenated according to the candidate site order to generate the individual fish position vectors corresponding to the candidate base station positioning scheme; based on all candidate base station positioning schemes in the communication planning dataset, the individual fish position vectors are arranged in a set to construct the initial fish population.

[0054] In this embodiment, the aggregate arrangement is specifically performed as follows: after the individual fish position vectors corresponding to each candidate base station positioning scheme are generated, the individual fish position vectors are read sequentially according to the preset scheme order and written sequentially into the initial population storage structure; a unique population number is assigned to each individual fish position vector, and a one-to-one correspondence is established between the candidate base station positioning scheme, the individual fish position vectors, and the population number; while keeping the internal candidate site order of each individual fish position vector unchanged, the external order arrangement of each individual fish position vector is completed to generate the initial fish population.

[0055] Associating multi-domain search vectors with individual fish populations involves sequentially reading the multi-domain search vectors corresponding to the position vectors of each individual fish population according to their population index, writing the multi-domain search vectors to the associated positions of fish populations with the same population index, establishing a one-to-one correspondence between the individual fish population position vectors and the multi-domain search vectors. After writing, a consistency check is performed on the length of the position vectors, the population index, and the multi-domain search vector index of each individual fish population to complete the association between the multi-domain search vectors and each individual fish population. According to the one-to-one correspondence between each candidate base station positioning scheme and the multi-domain search vectors, the multi-domain search vectors corresponding to each candidate base station positioning scheme are read; using the order of the individual fish population position vectors in the initial fish population as the index order, the corresponding multi-domain search vectors are written to the associated positions of fish populations with the same index; a correspondence check is performed on each individual fish population after the multi-domain search vectors are written to complete the association between the multi-domain search vectors and each individual fish population; specifically:

[0056] The system reads the position vectors and corresponding multi-domain search vectors of each fish group sequentially according to the population index. It checks whether the position vector index of each fish group matches the multi-domain search vector index, and whether the candidate base station positioning scheme identifier corresponding to the position vector of each fish group matches the candidate base station positioning scheme identifier corresponding to the multi-domain search vector. It also checks whether the order of candidate sites within the position vector of each fish group maintains a preset correspondence with the order of search domains within the multi-domain search vector. When the indexes, candidate base station positioning scheme identifiers, and internal order correspondences are consistent, the position vector of the current fish group is correctly associated with the multi-domain search vector. If inconsistencies exist, the incorrectly located fish group is identified, the corresponding association record is cleared, and the corresponding multi-domain search vector is rewritten according to the population index until the correspondence verification is completed for all fish group individuals.

[0057] In this embodiment, the process of generating a multi-domain independent memory pool includes:

[0058] The fish swarm individual position vectors and corresponding multi-domain search vectors from each historical iteration are read in the iteration order. Based on the search domain division results in the multi-domain search vectors, the candidate base station positioning schemes corresponding to each fish swarm individual are divided into coverage search domain segments, interference search domain segments, load search domain segments, and constraint search domain segments. Within the same historical iteration, the coverage search domain segments, interference search domain segments, load search domain segments, and constraint search domain segments are compared in single domains to select the single-domain advantageous segments in each search domain.

[0059] Specifically, within the same historical iteration, using the search domain partitioning results in the multi-domain search vector as boundaries, the candidate base station positioning schemes corresponding to each fish population are divided into coverage search domain segments, interference search domain segments, load search domain segments, and constraint search domain segments. Within each search domain, intra-domain sorting is performed separately. Search domain segments ranked within a preset priority interval are determined as single-domain dominant segments, while those ranked within a preset lower interval are determined as single-domain inferior segments. When intra-domain evaluation results are the same, search domain segments with closer historical iteration numbers are prioritized; when historical iteration numbers are the same, search domain segments with earlier population numbers are prioritized. The selected single-domain dominant segments are written into their respective independent memory areas according to the search domain, and this process is repeated along the historical iteration... The process of adding data continuously forms a sequence of historical experience accumulated within the same search domain. The selected single-domain advantageous fragments are bound and recorded with their corresponding historical iteration numbers, population numbers, candidate base station positioning scheme identifiers, and search domain identifiers. The single-domain advantageous fragments that have been bound and recorded are appended to their respective independent memory areas according to the search domain identifiers, ensuring that previously written content from previous historical iterations within the same search domain is not deleted. After each append write, the single-domain advantageous fragments within the corresponding independent memory areas are sequentially arranged according to their historical iteration numbers and population numbers, forming a sequence of historical experience accumulated within the same search domain. The coverage independent memory area, interference independent memory area, load independent memory area, and constraint independent memory area are combined to generate a multi-domain independent memory pool.

[0060] In this embodiment, the process of constructing the inter-domain conflict matrix includes:

[0061] Read the multi-domain search vectors corresponding to each individual fish in the population group sequentially according to their population numbers. Using the arrangement of the covering search domain, interference search domain, load search domain, and constraint search domain in the multi-domain search vectors of the same individual fish as a unified correspondence benchmark, perform position alignment on the domain vector segments of any two search domains, calculate the difference results item by item according to the corresponding positions, and generate inter-domain difference sequences between each search domain. Perform segment aggregation on each inter-domain difference sequence to obtain the inter-domain difference values ​​between the covering search domain and the interference search domain, between the covering search domain and the load search domain, between the covering search domain and the constraint search domain, between the interference search domain and the load search domain, between the interference search domain and the constraint search domain, and between the load search domain and the constraint search domain. Using the pairwise correspondence of the search domains as the matrix row and column indices, write the inter-domain difference values ​​into the corresponding matrix positions, making the matrix symmetrical. The inter-domain difference values ​​of the positions are kept consistent to generate an inter-domain conflict matrix. The inter-domain difference values ​​of all matrix positions in the inter-domain conflict matrix are sorted according to their size. The search domain combination corresponding to the first matrix position in the sorting result is taken as the primary conflict domain pair, and the search domain combination corresponding to the second matrix position in the sorting result is taken as the secondary conflict domain pair. The search domain combination corresponding to the primary conflict domain pair is determined as the conflict type of the current fish population. When the primary conflict domain pair and the secondary conflict domain pair contain the same search domain, the search domain corresponding to the primary conflict domain pair and the non-repeating search domain combination in the secondary conflict domain pair are determined as the composite conflict type. The inter-domain difference value corresponding to the primary conflict domain pair is read. Combined with the inter-domain difference value corresponding to the secondary conflict domain pair and the distribution of inter-domain difference values ​​of the remaining matrix positions in the inter-domain conflict matrix, the conflict concentration is calculated to generate the conflict intensity under the current conflict type.

[0062] Performing conflict concentration calculation involves reading the inter-domain difference values ​​corresponding to the primary conflict domain pair, secondary conflict domain pair, and other matrix positions, comparing the differences between the primary and secondary conflict domain pairs, as well as the differences between the primary conflict domain pair and the other matrix positions. When the inter-domain difference value corresponding to the primary conflict domain pair is significantly higher than the inter-domain difference values ​​of the other pairs, the conflict concentration is determined to be high. When the inter-domain difference values ​​corresponding to the primary and secondary conflict domain pairs are both higher than the inter-domain difference values ​​of the other pairs, the conflict concentration is determined to be composite concentration. When the inter-domain difference values ​​are close, the conflict concentration is determined to be low, and the determination result is used as the conflict intensity.

[0063] In this embodiment, the process of generating the composite guiding vector includes:

[0064] Based on the conflict type and intensity corresponding to the current fish population, the target search domain to be invoked is determined. Single-domain dominant fragments are read from the corresponding independent memory areas of the multi-domain independent memory pool according to the target search domain, and the retrieval is performed sequentially according to the historical iteration number and population number. Single-domain dominant features consistent with the candidate site order in the current fish population's position vector are extracted from the sequential retrieval results. The extracted single-domain dominant features corresponding to different target search domains are aligned according to the candidate site order, maintaining the independence of each target search domain boundary. Based on the conflict type corresponding to the current fish population, the aligned single-domain dominant features are segmented and recombined according to the order of primary conflict domain priority and secondary conflict domain priority. The aligned single-domain dominant features are divided into multiple corresponding position segments according to the candidate site order. The corresponding single-domain dominant features in each position segment are selected sequentially according to the order of primary conflict domain priority and secondary conflict domain priority. The selected position segments are then reassembled according to the original candidate site order to generate a composite feature sequence. The composite feature sequence is written into the guide vector construction position according to the arrangement order of the current fish population's position vector to generate a composite guide vector.

[0065] In this embodiment, the process of establishing a heterogeneous search mechanism oriented towards conflict states includes:

[0066] Based on the conflict type, the position vector of the current fish population is divided into a main conflict domain update segment, a secondary conflict domain update segment, and a non-conflict domain preservation segment. The guide segment corresponding to the main conflict domain update segment in the composite guide vector is written into the main conflict domain update segment, and the guide segment corresponding to the secondary conflict domain update segment in the composite guide vector is written into the secondary conflict domain update segment, while keeping the position content of the non-conflict domain preservation segment unchanged. The correction magnitude of the main conflict domain update segment and the secondary conflict domain update segment are determined based on the conflict intensity, ensuring that the correction magnitude of the main conflict domain update segment is greater than that of the secondary conflict domain update segment. A primary correction is performed on the main conflict domain update segment after the guide segment is written, and an auxiliary correction is performed on the secondary conflict domain update segment after the guide segment is written. The corrected main conflict domain update segment, secondary conflict domain update segment, and non-conflict domain preservation segment are then backfilled into the position vector of the current fish population in their original position order, generating a split update position vector.

[0067] In this embodiment, the main correction is specifically performed as follows: using the guiding fragment written in the main conflict domain update segment as the main correction reference, according to the candidate station order of the main conflict domain update segment in the current position vector, the current position content and the corresponding guiding fragment are read segment by segment, the position difference between the current position content and the corresponding guiding fragment is compared, and the main correction direction of each candidate station position is determined; according to the correction magnitude corresponding to the main conflict domain, the current position content is adjusted along the main correction direction, and the adjustment result is written back to the corresponding position in the main conflict domain update segment. After the adjustment of all candidate station positions is completed, the main conflict domain update segment after main correction is obtained.

[0068] The auxiliary correction process is as follows: using the guidance fragment written in the secondary conflict domain update segment as the auxiliary correction benchmark, according to the candidate site order of the secondary conflict domain update segment in the current position vector, the current position content and the corresponding guidance fragment are read segment by segment, the positional difference between the current position content and the corresponding guidance fragment is compared, the auxiliary correction direction of each candidate site position is determined, and the positional adjustment of the current position content is performed along the auxiliary correction direction according to the correction magnitude corresponding to the secondary conflict domain. The adjustment result is written back to the corresponding position in the secondary conflict domain update segment. After the adjustment of all candidate site positions is completed, the secondary conflict domain update segment after auxiliary correction is obtained. The candidate site order consistency check and the search domain boundary consistency check are performed on the split update position vector. When the check passes, the split update position vector is determined as the corrected fish group individual position vector. When the check fails, the composite guidance vector is reread according to the original population number and the segmented writing and position backfilling are repeated. The corrected fish group individual position vector is written into the hierarchical sub-fish group cooperative search process, and fish group individuals under different conflict types undergo differentiated position correction according to the corresponding update segment.

[0069] In this embodiment, the process of generating the optimal base station deployment result includes:

[0070] Based on the updated candidate base station positioning scheme, the target communication area is divided into hotspot search areas, bearer search areas, and edge search areas according to user distribution density, service demand distribution, and regional boundary locations. Based on the regional division results of the hotspot search areas, bearer search areas, and edge search areas, position segments corresponding to the areas are extracted from the corrected individual position vectors of the fish swarms to construct hotspot sub-fish swarms, bearer sub-fish swarms, and edge sub-fish swarms respectively. Within the hotspot sub-fish swarms, position correction is performed by prioritizing the composite guiding vector segment corresponding to the load search domain, according to the position segment order within the hotspot search area. Within the bearer sub-fish swarms, position correction is performed by prioritizing the composite guiding vector segments corresponding to the coverage search domain and interference search domain, according to the position segment order within the bearer search area. Within the edge sub-fish swarms, position correction is performed by prioritizing the composite guiding vector segments corresponding to the coverage search domain and constraint search domain, according to the position segment order within the edge search area. This generates the updated sub-fish swarm results for each search area. The updated sub-fish swarm results for each search area are then backfilled into the corrected area according to the spatial position order within the original target communication area. The corresponding region location in the individual fish location vector is used to generate a cooperative localization result. Using the cooperative localization result as input, the domain features corresponding to the coverage search domain, interference search domain, load search domain, and constraint search domain are re-extracted according to the candidate site order. These features are then recombined according to the original search domain boundaries to generate an updated multi-domain search vector. Based on the updated multi-domain search vector, the inter-domain difference values ​​between each search domain are recalculated and written back according to the original matrix positions to generate an updated inter-domain conflict matrix. Based on the updated multi-domain search vector and the updated inter-domain conflict matrix, single-domain advantageous segments in each search domain are re-selected and appended to the corresponding independent memory area according to the search domain to generate an updated multi-domain independent memory pool. It is then determined whether the updated multi-domain search vector, the updated inter-domain conflict matrix, and the updated multi-domain independent memory pool meet the preset termination conditions. If the preset termination conditions are met, the current cooperative localization result is determined as the optimal base station deployment result. If the preset termination conditions are not met, the updated multi-domain search vector, the updated inter-domain conflict matrix, and the updated multi-domain independent memory pool are written into the next round of individual fish update process to continue iterating.

[0071] In this embodiment, the preset termination condition refers to the set of judgment conditions used to determine whether the current iteration process has ended. After each round of hierarchical sub-fish swarm cooperative search is completed, the updated multi-domain search vector, the updated inter-domain conflict matrix, and the updated multi-domain independent memory pool are read sequentially, and the current cooperative localization result is compared with the previous round's cooperative localization result. When the cooperative localization result remains unchanged within consecutive preset rounds, it is determined that the result stability termination condition has been met. When the search domain arrangement result in the multi-domain search vector remains unchanged within consecutive preset rounds, and the main conflict domain pairs in the inter-domain conflict matrix remain unchanged, it is determined that the conflict stability termination condition has been met. When the number of newly added single-domain dominant segments written in the multi-domain independent memory pool within consecutive preset rounds is lower than a preset number threshold, it is determined that the memory convergence termination condition has been met. When any one of the result stability termination condition, conflict stability termination condition, and memory convergence termination condition is met, it is determined that the preset termination condition is met, and the current iteration ends. When none of the conditions are met, the updated multi-domain search vector, the updated inter-domain conflict matrix, and the updated multi-domain independent memory pool are written into the next round of fish swarm individual update process to continue iteration.

[0072] Example 1: To verify the feasibility of this invention in practice, it was applied to a base station location optimization scenario in a continuous coverage communication area. The communication area includes high-density service clusters, conventional bearer areas, and edge coverage areas. The original base stations within the area are unevenly distributed, with some candidate sites located near high-service-demand areas and others in boundary extension areas. Network operation suffers from coverage holes, excessive overlap between adjacent base stations, concentrated load in hotspot areas, and limited deployment in edge areas. When using conventional single-fitness site selection methods, the planning process often compresses coverage, interference, load, and constraints into the same evaluation channel for unified sorting. While the results superficially meet the overall score requirements, in actual deployment, situations arise where local coverage is good but interference is increased, or load distribution is improved but edge area coverage is insufficient. Multiple rounds of manual correction are required, resulting in poor stability of the location results. The search process easily remains near locally optimal solutions, making it difficult to continue moving towards better site combinations.

[0073] In this embodiment, basic data for the target communication area planning is first collected. This basic data covers the distribution of candidate sites, existing base stations, user space distribution, service demand distribution, and constraint conditions. After field parsing, coordinate unification, data cleaning, spatial grid division, location mapping, unit unification, and normalization, a communication planning dataset is generated. The communication planning dataset contains 146 candidate sites, 58 existing base stations, 1200 user distribution grids, 260 high-value service demand grids, and 190 constraint-sensitive grids. Based on the above data, multi-domain decoupling processing is performed on each candidate base station positioning scheme. The scheme is split into a coverage search domain, an interference search domain, a load search domain, and a constraint search domain. Then, features of the four domains are extracted and combined into a multi-domain search vector. This processing method separates and preserves the originally mixed evaluation relationships, ensuring that the advantages of the coverage direction are not offset by the shortcomings of the interference direction, and that the improvement results of the load direction are not directly covered by the limitations of the constraint direction. Subsequent searches can adopt different correction paths for different conflict states.

[0074] During the individual fish swarm construction phase, each candidate base station positioning scheme is encoded as a fish swarm individual location vector, and the multi-domain search vectors are associated one by one with the corresponding fish swarm individuals. The initial population size is set to eighty candidate schemes, with each scheme containing twelve newly deployed site location codes. After the start of historical iterations, the globally optimal single retention method is no longer used. Instead, single-domain advantageous fragments are selected for each of the four search domains: coverage, interference, load, and constraints, and written into the multi-domain independent memory pool according to the search domain. After the first ten iterations, the multi-domain independent memory pool has accumulated 128 coverage advantageous fragments, 114 interference advantageous fragments, 121 load advantageous fragments, and 109 constraint advantageous fragments. The direct effect of this accumulation process is that subsequent updates are no longer limited to following the overall optimal scheme of a single round, but can extract single-domain advantageous features more suitable for the current conflict state from different historical iterations, making the search path more flexible.

[0075] In the conflict identification phase, the inter-domain correlation differences are calculated for the multi-domain search vector corresponding to each individual fish in the swarm, constructing an inter-domain conflict matrix. The conflict matrix identifies the most prominent conflict domain pairs in the current scheme and provides the conflict intensity. During implementation, it was found that the most frequent conflict type in the initial stage was coverage and interference conflict, accounting for over 40% of all conflict determinations, while load and constraint conflict accounted for nearly 30%, and composite conflict accounted for less than 20%. After entering the mid-term iteration, the proportion of coverage and interference conflict decreased significantly, and load and constraint conflict became the main source of conflict. This indicates that after the initial search, coverage holes and overlapping coverage have been significantly compressed, and the search focus has shifted to the coordination between hotspot area carrying capacity and edge site constraints. Based on this change in conflict state, the method no longer uses a unified update but instead adopts a conflict-driven split update. The main conflict domain uses a larger correction magnitude, the secondary conflict domain uses a smaller correction magnitude, and non-conflict domains remain stable. The composite guiding vector is recombined from single-domain advantageous features extracted from multi-domain independent memory pools, and its correction of the current position vector is highly targeted. After this treatment, the search process continued to iterate effectively even after twenty rounds, without exhibiting the early convergence stagnation phenomenon commonly seen in traditional fish swarm algorithms.

[0076] In the regional coordination phase, the target communication area is divided into hotspot search area, bearer search area, and edge search area. Hotspot search area prioritizes load-direction advantage features, bearer search area prioritizes coverage and interference-direction advantage features, and edge search area prioritizes coverage and constraint-direction advantage features. Each search area forms a sub-group and performs local updates. The update results are then backfilled according to the original spatial order to form the coordinated positioning result. In practical applications, this approach is very intuitive. In service-intensive areas, the focus shifts from simply achieving signal coverage to prioritizing user bearer balance; in conventional bearer areas, the focus shifts from even site distribution to coverage continuity and adjacent interference compression; and in edge areas, the focus shifts from simply expanding coverage radius to finding a more suitable location between gap-filling needs and deployment limitations. After thirty-five iterations, the coordinated positioning result stabilizes. The main conflict domain pairs remain unchanged for three consecutive rounds, and the number of newly added single-domain advantage segments in the multi-domain independent memory pool decreases below a preset threshold, meeting the termination condition. The optimal base station deployment result is then output, as shown in Table 1 below.

[0077] Table 1: Comparison of the Implementation Effects of Communication Base Station Positioning Optimization

[0078] Number of candidate sites 146 146 146 Number of existing base stations 58 58 58 Number of newly deployed stations 12 12 12 Comprehensive coverage 91.8% 94.3% 97.1% Coverage of void ratio 8.2% 5.7% 2.9% Overlapping coverage ratio 27.4% 22.6% 15.8% High load in hotspot areas, number of grid cells 68 51 29 Site utilization can be implemented in peripheral areas 67.9% 72.4% 84.6% Number of adjacent high-interference site pairs 36 25 17 Average service load balance 78.5 84.7 90.2 Marginal blindness filling efficiency 69.3% 76.8% 87.4% Stable convergence iteration rounds 40 32 35 Effective improvement in retention rate after 20 rounds 31.6% 42.8% 68.5% Number of manual review and corrections 9 5 3 Location result first pass rate 56.8% 71.2% 85.9%

[0079] As shown in Table 1, under the condition that the number of candidate sites, the number of existing base stations, and the number of newly deployed sites remain the same, the method of this invention shows superior optimization effects in multiple core indicators. This indicates that the performance improvement is not achieved by increasing the number of sites or expanding the deployment scale, but by improving the search and update mechanism under the same planning conditions. Compared with the original planning scheme, this invention increases the overall coverage rate from 91.8% to 97.1%, a further improvement of 2.8 percentage points compared with the conventional single-fitness fish swarm algorithm; the coverage hole ratio decreases from 8.2% to 2.9%, a significant decrease, indicating that this invention has stronger blind spot filling capabilities in edge areas and weak coverage areas. The overlapping coverage ratio decreases from 27.4% to 15.8%, a reduction of 6.8 percentage points compared with the conventional method, indicating that while improving coverage, this invention does not simply expand the coverage area to obtain results, but rather takes into account both coverage continuity and overlap suppression effects.

[0080] From the perspective of load and interference-related indicators, this invention has a more significant improvement effect on hotspot bearer conflicts and inter-site interference problems in complex communication areas. The number of grids with high load in hotspot areas has decreased from 68 to 29, a reduction of 39 compared to the original planning scheme and 22 compared to the conventional algorithm. This reflects that this invention has a stronger load distribution capability in high service density areas. The average service bearer balance has increased from 78.5 to 90.2, indicating that the bearer distribution among sites is more balanced and the concentrated congestion in hotspot areas has been effectively alleviated. The number of adjacent high-interference site pairs has decreased from 36 to 17, a further reduction of 8 pairs compared to the conventional algorithm. This indicates that this invention, through inter-domain conflict identification and conflict-driven updates, can more accurately compress high-interference site combinations and enhance inter-site coordination. The site utilization rate in edge areas reaches 84.6%, and the edge blind spot filling efficiency has increased to 87.4%. This shows that this invention not only improves the coverage quality in edge areas but also enhances the effective utilization capability under constrained site conditions.

[0081] From the perspectives of the search process and engineering application effects, this invention maintains high optimization quality while also demonstrating better stability and practicality. Although the stable convergence iteration rounds are 35, slightly higher than the 32 rounds of the conventional single-fitness fish swarm algorithm, the effective improvement retention rate reaches 68.5% after 20 rounds, significantly higher than the 42.8% of the conventional algorithm and the 31.6% of the original scheme. This indicates that this invention can still maintain effective search in the mid-to-late stages, alleviating the problem of premature convergence of conventional methods. The number of manual verification and corrections has been reduced from 9 to 3, and the first-pass rate of the positioning results has increased from 56.8% to 85.9%, indicating that the site location scheme output by this invention is closer to the actual deployable results, and the pressure of subsequent manual corrections has been significantly reduced. Combining all the data, it can be seen that this invention has achieved simultaneous improvements in coverage improvement, interference suppression, load balancing, edge blind spot filling, and result stability, verifying the effectiveness of multi-domain decoupled search, inter-domain conflict matrix, single-domain advantage memory accumulation, and hierarchical sub-fish swarm collaborative update mechanism in communication base station positioning optimization.

[0082] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. An adaptive optimization method for communication base station positioning based on an improved fish swarm algorithm, characterized in that, Includes the following steps: Collect basic data for the target communication area planning and preprocess it to generate a communication planning dataset. Each candidate base station positioning scheme is divided into a coverage search domain, an interference search domain, a load search domain, and a constraint search domain. Domain features are extracted and combined to generate a multi-domain search vector. The candidate base station positioning scheme is encoded into the individual fish population position vector, the initial fish population is constructed, and the multi-domain search vector is associated with each individual fish population. Based on the multi-domain search vectors corresponding to each individual fish group, the single-domain dominant solutions in the historical iterations are stored separately according to the search domain, generating a multi-domain independent memory pool. Based on the multi-domain search vector calculation, the correlation difference between each search domain is calculated, an inter-domain conflict matrix is ​​constructed, and the conflict type and conflict intensity of the candidate base station positioning scheme for each individual fish group are determined. Based on the conflict type and intensity, the fish group individuals are controlled to execute the corresponding split update path. Different single-domain advantageous features are extracted from the multi-domain independent memory pool and recombined to generate a composite guiding vector. The composite guiding vector is used to correct the position vector of the fish group individuals, changing the unified update to conflict-driven split update. The single-domain advantageous features are recombined to guide the correction of the current position, thus establishing a heterogeneous search mechanism oriented towards conflict state. The target communication area is divided, and a hierarchical sub-fish swarm cooperative search is performed on the updated candidate base station positioning scheme. The positioning results are fed back to the update process, and the process is iterated until the termination condition is met, and the optimal base station deployment result is output.

2. The adaptive optimization method for communication base station positioning based on the improved fish swarm algorithm according to claim 1, characterized in that, The preprocessing specifically includes: parsing the fields of the target communication area planning basic data, and sequentially performing coordinate unification, data cleaning, spatial grid division, location mapping, unit unification and normalization on the parsed target communication area planning basic data, and then aggregating and encapsulating the data to generate a communication planning dataset.

3. The adaptive optimization method for communication base station positioning based on the improved fish swarm algorithm according to claim 2, characterized in that, The process of generating the multi-domain search vector includes: For each candidate base station positioning scheme in the communication planning dataset, the candidate site location, existing base station location, user distribution location, service demand distribution, and constraint condition distribution are read. Centered on the candidate site location, the correspondence between the candidate site location and the user distribution location is divided into the coverage search domain, the correspondence between the candidate site location and the existing base station location is divided into the interference search domain, the correspondence between the candidate site location and the service demand distribution is divided into the load search domain, and the correspondence between the candidate site location and the constraint condition distribution is divided into the constraint search domain. Within the coverage search domain, interference search domain, load search domain, and constraint search domain, the correspondence calculation is performed according to the candidate site order to generate the coverage domain feature sequence, interference domain feature sequence, load domain feature sequence, and constraint domain feature sequence. The four feature sequences are aligned and arranged according to the candidate site order under the same candidate base station positioning scheme, keeping the boundaries of each search domain independent, generating multi-domain search vectors, and writing them into the fish swarm individual position vector association table of the corresponding candidate base station positioning scheme to complete the correspondence binding between multi-domain search vectors and fish swarm individual position vectors.

4. The adaptive optimization method for communication base station positioning based on the improved fish swarm algorithm according to claim 3, characterized in that, The process of constructing the initial fish population includes: According to the candidate site order in the candidate base station positioning scheme, the positions of each candidate site are read sequentially; the spatial coordinates of the candidate site positions are used as position encoding units and concatenated according to the candidate site order to generate the individual fish position vectors corresponding to the candidate base station positioning scheme; based on all candidate base station positioning schemes in the communication planning dataset, the individual fish position vectors are arranged in a set to construct the initial fish population. According to the one-to-one correspondence between each candidate base station positioning scheme and the multi-domain search vector, the multi-domain search vector corresponding to each candidate base station positioning scheme is read; the multi-domain search vector is written to the associated position of the fish individual with the same sequence number, using the arrangement order of the fish individual position vector in the initial fish population as the index order; the correspondence verification is performed on each fish individual after writing the multi-domain search vector to complete the association between the multi-domain search vector and each fish individual.

5. The adaptive optimization method for communication base station positioning based on the improved fish swarm algorithm according to claim 4, characterized in that, The process of generating a multi-domain independent memory pool includes: The fish swarm individual position vectors and corresponding multi-domain search vectors from each historical iteration are read in iterative order. Based on the search domain division results in the multi-domain search vectors, the candidate base station positioning schemes corresponding to each fish swarm individual are divided into coverage search domain segments, interference search domain segments, load search domain segments, and constraint search domain segments. Within the same historical iteration, single-domain comparisons are performed on the coverage search domain segments, interference search domain segments, load search domain segments, and constraint search domain segments to select the single-domain advantageous segments in each search domain. The selected single-domain advantageous segments are bound and recorded with the corresponding historical iteration sequence number, population sequence number, candidate base station positioning scheme identifier, and search domain identifier. According to the search domain identifier, the single-domain advantageous segments that have been bound and recorded are appended to the corresponding independent memory area, keeping the content written in previous historical iterations within the same search domain from being deleted. After each append, the single-domain advantageous segments in the corresponding independent memory area are arranged sequentially according to the historical iteration sequence number and population sequence number to form a continuously accumulated historical experience sequence under the same search domain. The coverage independent memory area, interference independent memory area, load independent memory area, and constraint independent memory area are combined to generate a multi-domain independent memory pool.

6. The adaptive optimization method for communication base station positioning based on the improved fish swarm algorithm according to claim 5, characterized in that, The process of constructing the inter-domain conflict matrix includes: Read the multi-domain search vectors corresponding to each individual fish in the population group sequentially according to their population numbers. Using the arrangement of the covering search domain, interference search domain, load search domain, and constraint search domain in the multi-domain search vectors of the same individual fish as a unified correspondence benchmark, perform position alignment on the domain vector segments of any two search domains, calculate the difference results item by item according to the corresponding positions, and generate inter-domain difference sequences between each search domain. Perform segment aggregation on each inter-domain difference sequence to obtain the inter-domain difference values ​​between the covering search domain and the interference search domain, between the covering search domain and the load search domain, between the covering search domain and the constraint search domain, between the interference search domain and the load search domain, between the interference search domain and the constraint search domain, and between the load search domain and the constraint search domain. Using the pairwise correspondence of the search domains as the matrix row and column indices, write the inter-domain difference values ​​into the corresponding matrix positions, making the matrix symmetrical. To generate an inter-domain conflict matrix, the inter-domain difference values ​​of all matrix positions in the inter-domain conflict matrix are kept consistent. The inter-domain difference values ​​of all matrix positions in the inter-domain conflict matrix are sorted according to their size. The search domain combination corresponding to the first matrix position in the sorting result is taken as the primary conflict domain pair, and the search domain combination corresponding to the second matrix position in the sorting result is taken as the secondary conflict domain pair. The search domain combination corresponding to the primary conflict domain pair is determined as the conflict type of the current fish population. When the primary and secondary conflict domain pairs contain the same search domain, the search domain combination corresponding to the primary conflict domain pair and the non-repeating search domain combination in the secondary conflict domain pair are determined as the composite conflict type. The inter-domain difference values ​​corresponding to the primary conflict domain pair are read, and combined with the inter-domain difference values ​​corresponding to the secondary conflict domain pairs and the distribution of inter-domain difference values ​​of the remaining matrix positions in the inter-domain conflict matrix, the conflict concentration is calculated to generate the conflict intensity under the current conflict type.

7. The adaptive optimization method for communication base station positioning based on the improved fish swarm algorithm according to claim 6, characterized in that, The process of generating the composite guiding vector includes: Based on the conflict type and intensity corresponding to the current fish population, the target search domain to be invoked is determined. Single-domain dominant fragments are read from the corresponding independent memory regions of the multi-domain independent memory pool according to the target search domain, and the retrieval is performed sequentially according to the historical iteration number and population number. Single-domain dominant features consistent with the candidate site order in the current fish population's position vector are extracted from the sequential retrieval results. The extracted single-domain dominant features corresponding to different target search domains are aligned according to the candidate site order, maintaining the independence of each target search domain boundary. Based on the conflict type corresponding to the current fish population, the aligned single-domain dominant features are segmented and recombined according to the order of primary conflict domain pairing and secondary conflict domain pairing, generating a composite feature sequence. The composite feature sequence is written into the guide vector construction position according to the arrangement order of the current fish population's position vector, generating a composite guide vector.

8. The adaptive optimization method for communication base station positioning based on the improved fish swarm algorithm according to claim 7, characterized in that, The process of establishing a heterogeneous search mechanism oriented towards conflict states includes: Based on the conflict type, the position vectors of the current fish swarm individuals are divided into a main conflict domain update segment, a secondary conflict domain update segment, and a non-conflict domain preservation segment. The guide segment corresponding to the main conflict domain update segment in the composite guide vector is written into the main conflict domain update segment, and the guide segment corresponding to the secondary conflict domain update segment in the composite guide vector is written into the secondary conflict domain update segment, while keeping the position content of the non-conflict domain preservation segment unchanged. The correction magnitudes of the main conflict domain update segment and the secondary conflict domain update segment are determined based on the conflict intensity, ensuring that the correction magnitude of the main conflict domain update segment is greater than that of the secondary conflict domain update segment. A principal correction is performed on the main conflict domain update segment after the guide segment is written, and a principal correction is performed on the secondary conflict domain update segment after the guide segment is written. The new segment performs auxiliary correction, backfilling the corrected main conflict domain update segment, secondary conflict domain update segment, and non-conflict domain preservation segment to the current fish population's position vector in the original position order, generating a split update position vector; the split update position vector performs candidate site order consistency check and search domain boundary consistency check. When the check passes, the split update position vector is determined as the corrected fish population's position vector. When the check fails, the composite guide vector is reread according to the original population number, and the segment writing and position backfilling are repeated; the corrected fish population's position vector is written into the hierarchical sub-fish population cooperative search process, and fish populations under different conflict types perform differentiated position correction according to the corresponding update segment.

9. The adaptive optimization method for communication base station positioning based on the improved fish swarm algorithm according to claim 8, characterized in that, The process of generating the optimal base station deployment result includes: Based on the updated candidate base station positioning scheme, the target communication area is divided into hotspot search areas, bearer search areas, and edge search areas according to user distribution density, service demand distribution, and regional boundary locations. Based on the regional division results of the hotspot search areas, bearer search areas, and edge search areas, position segments corresponding to the areas are extracted from the corrected individual position vectors of the fish swarms to construct hotspot sub-fish swarms, bearer sub-fish swarms, and edge sub-fish swarms respectively. Within the hotspot sub-fish swarms, position correction is performed by prioritizing the composite guiding vector segment corresponding to the load search domain, according to the position segment order within the hotspot search area. Within the bearer sub-fish swarms, position correction is performed by prioritizing the composite guiding vector segments corresponding to the coverage search domain and interference search domain, according to the position segment order within the bearer search area. Within the edge sub-fish swarms, position correction is performed by prioritizing the composite guiding vector segments corresponding to the coverage search domain and constraint search domain, according to the position segment order within the edge search area. This generates the updated sub-fish swarm results for each search area. The updated sub-fish swarm results for each search area are then backfilled into the corrected area according to the spatial position order within the original target communication area. The corresponding region location in the individual fish location vector is used to generate a cooperative localization result. Using the cooperative localization result as input, the domain features corresponding to the coverage search domain, interference search domain, load search domain, and constraint search domain are re-extracted according to the candidate site order. These features are then recombined according to the original search domain boundaries to generate an updated multi-domain search vector. Based on the updated multi-domain search vector, the inter-domain difference values ​​between each search domain are recalculated and written back according to the original matrix positions to generate an updated inter-domain conflict matrix. Based on the updated multi-domain search vector and the updated inter-domain conflict matrix, single-domain advantageous segments in each search domain are re-selected and appended to the corresponding independent memory area according to the search domain to generate an updated multi-domain independent memory pool. It is then determined whether the updated multi-domain search vector, the updated inter-domain conflict matrix, and the updated multi-domain independent memory pool meet the preset termination conditions. If the preset termination conditions are met, the current cooperative localization result is determined as the optimal base station deployment result. If the preset termination conditions are not met, the updated multi-domain search vector, the updated inter-domain conflict matrix, and the updated multi-domain independent memory pool are written into the next round of individual fish update process to continue iterating.

Images