A resource allocation method, apparatus and device

Abstract

This invention provides a resource allocation method, apparatus, and device. The resource allocation method includes: acquiring resource allocation parameter information corresponding to at least one terminal within a target area; obtaining first information on the traffic demand of each terminal within at least one time period based on the resource allocation parameter information; and obtaining resource allocation result information corresponding to each terminal based on the resource allocation parameter information and the first information. The resource allocation parameter information includes at least one of: base station information, terminal information, sub-frequency band information, and the upper limit of the base station's transmit power. The first information includes: information on the traffic demand of the terminal within the current time period. The resource allocation result information includes at least one of: base station allocation information, sub-frequency band allocation information, and transmit power allocation information. This solution solves the problems of existing resource allocation schemes failing to effectively meet traffic demand and being overly complex.

Description

Technical Field

[0001] This invention relates to the field of wireless network technology, and in particular to a resource allocation method, apparatus, and device. Background Technology

[0002] In recent years, user demand for higher-speed mobile networks has been continuously growing. Given this massive traffic demand, more efficient allocation of existing wireless network resources has become crucial. However, in current technologies, heuristic allocation schemes struggle to adapt to changing environments and often fall far short of the optimal solution. Traditional optimization methods face multiple constraints, including limitations in hardware resources and long-term user traffic demands, as well as the scale problem caused by a large number of users. The former (multiple constraints) makes constructing specific optimization algorithms extremely complex, while the latter (scale problem) results in enormous computational resources required for solving the problem.

[0003] Furthermore, existing wireless network resource allocation technologies do not take into account users' traffic demand constraints, which makes it difficult to effectively meet the traffic demands of some users.

[0004] As shown above, existing resource allocation schemes suffer from problems such as ineffectiveness in meeting traffic demands and complexity. Summary of the Invention

[0005] The purpose of this invention is to provide a resource allocation method, apparatus, and device to solve the problems of existing resource allocation schemes being unable to effectively meet traffic demands and being too complex.

[0006] To address the aforementioned technical problems, embodiments of the present invention provide a resource allocation method, comprising:

[0007] Obtain resource allocation parameter information corresponding to at least one terminal within the target area;

[0008] Based on the resource allocation parameter information, the first information on the traffic demand of each terminal within at least one time period is obtained;

[0009] Based on the resource allocation parameter information and the first information, the resource allocation result information corresponding to each terminal is obtained;

[0010] The resource allocation parameter information includes at least one of the following: base station information corresponding to the target area, terminal information corresponding to the target area, sub-band information of spectrum resources shared by base stations in the target area, and upper limit value of transmit power for each base station.

[0011] The first information includes: information on the terminal's traffic demand during the current time period;

[0012] The resource allocation result information includes at least one of the following: base station allocation information, sub-frequency band allocation information, and transmit power allocation information.

[0013] Optionally, obtaining the first information on the traffic demand of each terminal within at least one time period based on the resource allocation parameter information includes:

[0014] For at least one time period prior to the current time period of the terminal, a first operation is performed to obtain the corresponding first terminal codes;

[0015] Based on the first terminal encoding, the encoding history sequence is obtained;

[0016] Based on the encoded history sequence, the first information of the terminal's traffic demand in the current time period is obtained;

[0017] The first operation includes: constructing a corresponding first relationship graph based on the resource allocation parameter information corresponding to the time period; and obtaining a first terminal code corresponding to the time period based on the first relationship graph.

[0018] Optionally, constructing the corresponding first relationship graph based on the resource allocation parameter information corresponding to the time period includes:

[0019] Based on the resource allocation parameter information corresponding to the time period, a first relationship graph is constructed with the first relevant information of the base station and the second relevant information of the terminal as nodes.

[0020] The first relevant information includes: the base station information and the upper limit of the transmission power, wherein the base station information includes: the identity information of the base station;

[0021] The second relevant information includes: the terminal information, which includes: second information on the unmet traffic demand of the terminal and the number of remaining time periods after the current time period of the terminal;

[0022] In the first relationship diagram, the weights between the first relevant information and the second relevant information are determined based on the signal attenuation factor between the base station and the terminal.

[0023] Optionally, obtaining the first information on the terminal's traffic demand in the current time period based on the encoded historical sequence includes:

[0024] Based on the encoded history sequence, the terminal state change trend code is obtained;

[0025] Based on the terminal status change trend encoding, the first information of the terminal's traffic demand in the current time period is obtained.

[0026] Optionally, obtaining the first terminal code corresponding to the time period based on the first relationship graph includes:

[0027] Using a graph neural network, the first terminal code corresponding to the time period is obtained based on the first relationship graph;

[0028] And / or, obtaining the terminal state change trend code based on the encoded history sequence includes:

[0029] Using a recurrent neural network, the terminal state change trend code is obtained based on the encoded history sequence.

[0030] Optionally, obtaining the first information on the terminal's traffic demand within the current time period based on the terminal state change trend encoding includes:

[0031] Using a two-layer fully connected network, the first probability parameter information is obtained by encoding the terminal state change trend.

[0032] Based on the first probability parameter information, the first Gaussian probability distribution is obtained;

[0033] From the first Gaussian probability distribution, the traffic demand of the terminal in the current time period is determined by sampling, and the first information of the traffic demand of the terminal in the current time period is obtained.

[0034] Optionally, obtaining the resource allocation result information corresponding to each terminal based on the resource allocation parameter information and the first information includes:

[0035] Based on the resource allocation parameter information and the first information, construct the target relationship graph;

[0036] Based on the target relationship diagram, the reference code is obtained;

[0037] Based on the reference code, the resource allocation result information corresponding to the terminal is obtained.

[0038] Optionally, the target relationship graph includes a second relationship graph;

[0039] The step of constructing a target relationship graph based on the resource allocation parameter information and the first information includes:

[0040] Based on the resource allocation parameter information and the first information, a second relationship graph is constructed with the third relevant information of the base station and the fourth relevant information of the terminal as nodes;

[0041] And / or, the reference encoding includes: a first base station encoding and a second terminal encoding;

[0042] The step of obtaining the reference code based on the target relationship diagram includes:

[0043] Using a graph neural network, the first base station code and the second terminal code are obtained based on the second relationship graph;

[0044] And / or, obtaining the resource allocation result information corresponding to the terminal based on the reference encoding includes:

[0045] The first base station code and the second terminal code are concatenated to obtain the first concatenated code;

[0046] Using a two-layer fully connected network, the first traffic estimate of the terminal accessing the base station is obtained based on the first splicing code;

[0047] Based on the first traffic estimate, the target access base station of the terminal is determined, and the base station allocation information corresponding to the terminal is obtained;

[0048] The third relevant information includes: the base station information and the upper limit of the transmission power, wherein the base station information includes: the identity information of the base station;

[0049] The fourth relevant information includes: the first information;

[0050] In the second relationship diagram, the weights between the third relevant information and the fourth relevant information are determined based on the signal attenuation factor between the base station and the terminal.

[0051] Optionally, the target relationship graph includes a third relationship graph;

[0052] The step of constructing a target relationship graph based on the resource allocation parameter information and the first information includes:

[0053] Based on the resource allocation parameter information and the first information, a third relationship graph is constructed with the fifth relevant information of the base station and the sixth relevant information of the terminal as nodes;

[0054] And / or, the reference encoding includes: a third terminal encoding;

[0055] The step of obtaining the reference code based on the target relationship diagram includes:

[0056] Using a graph neural network, the first sub-encoder is obtained based on the second relationship graph;

[0057] Using a graph neural network, the second sub-encoder is obtained based on the third relationship graph;

[0058] The first sub-encode and the second sub-encode are concatenated to obtain the third terminal code;

[0059] And / or, obtaining the resource allocation result information corresponding to the terminal based on the reference encoding includes:

[0060] Using a two-layer fully connected network, the second traffic estimate for the terminal access sub-band is obtained based on the third terminal coding;

[0061] Based on the second traffic estimate, the target access sub-band of the terminal is determined, and the sub-band allocation information corresponding to the terminal is obtained;

[0062] The fifth relevant information includes: the base station information and the upper limit of the transmission power, wherein the base station information includes: the identity information of the base station;

[0063] The sixth relevant information includes: the first information;

[0064] In the third relationship diagram, the weights between the fifth and sixth related information are determined based on whether there is a connection between the base station and the terminal.

[0065] Optionally, the target relationship graph includes a fourth relationship graph;

[0066] The step of constructing a target relationship graph based on the resource allocation parameter information and the first information includes:

[0067] Based on the resource allocation parameter information and the first information, a fourth relationship graph is constructed with the seventh relevant information of the base station and the eighth relevant information of the terminal as nodes;

[0068] And / or, the reference encoding includes: a fourth terminal encoding;

[0069] The step of obtaining the reference code based on the target relationship diagram includes:

[0070] Using a graph neural network, the third sub-encoder is obtained based on the second relationship graph;

[0071] Using a graph neural network, the fourth sub-encoder is obtained based on the third relation graph;

[0072] Using a graph neural network, the fifth sub-encoder is obtained based on the fourth relation graph;

[0073] The third sub-encode, the fourth sub-encode, and the fifth sub-encode are concatenated to obtain the fourth terminal code;

[0074] And / or, obtaining the resource allocation result information corresponding to the terminal based on the reference encoding includes:

[0075] Using a two-layer fully connected network, the second probability parameter information is obtained based on the fourth terminal encoding;

[0076] Based on the second probability parameter information, the second Gaussian probability distribution is obtained;

[0077] Based on the second Gaussian probability distribution, the transmit power allocation information corresponding to the terminal is obtained;

[0078] The seventh relevant information includes: the base station information and the upper limit of the transmission power, wherein the base station information includes: the identity information of the base station;

[0079] The eighth relevant information includes: the first information;

[0080] In the fourth relationship diagram, the weights between the seventh and eighth related information are determined based on whether there is a connection between the sub-frequency band of the base station and the terminal.

[0081] Optionally, obtaining the transmit power allocation information corresponding to the terminal based on the second Gaussian probability distribution includes:

[0082] The power reference value corresponding to the terminal is determined by sampling from the second Gaussian probability distribution;

[0083] Based on the power reference value, the corresponding transmit power percentage of the terminal is obtained;

[0084] Based on the transmission power ratio and the upper limit of transmission power, the transmission power allocation information corresponding to the terminal is obtained.

[0085] Optional, also includes:

[0086] Based on the resource allocation result information, the traffic value allocated to the corresponding terminal is obtained;

[0087] Adjust the relevant network for resource allocation based on the traffic value;

[0088] The relevant network includes at least one of the following:

[0089] Graph neural networks;

[0090] Recurrent neural networks;

[0091] Two-layer fully connected network.

[0092] Optionally, the process of adjusting resource allocation based on the traffic value includes:

[0093] Based on the traffic value, obtain the traffic loss value corresponding to the resource allocation result information;

[0094] Based on the aforementioned traffic loss value, adjust the relevant network for resource allocation.

[0095] Optionally, obtaining the traffic loss value corresponding to the resource allocation result information based on the traffic value includes:

[0096] Using the first method, based on the traffic value, obtain the traffic loss value corresponding to the resource allocation result information;

[0097] The first method includes: converting constrained adjustment conditions into unconstrained adjustment conditions;

[0098] The constrained adjustment conditions include: ensuring that the traffic allocated to each terminal in the first time period is greater than the traffic demand in the first time period, and maximizing the total traffic obtained by all terminals in the first time period;

[0099] The unconstrained adjustment conditions include: maximizing the total traffic of all terminals in the first time period, and maximizing the weighted sum of the traffic requirements met by all terminals in the first time period.

[0100] Optionally, the base station information in the resource allocation parameter information corresponding to the terminal is: the base station information of the first base station in the target area whose first distance from the terminal is less than a first threshold;

[0101] And / or, the terminal information in the resource allocation parameter information corresponding to the terminal is: the terminal information of a second terminal in the target area whose second distance from the terminal is less than a second threshold;

[0102] And / or, the sub-frequency band information in the resource allocation parameter information corresponding to the terminal is: the sub-frequency band information of the spectrum resources shared by the first base station;

[0103] And / or, the upper limit of the transmit power in the resource allocation parameter information corresponding to the terminal is: the upper limit of the transmit power of the first base station.

[0104] This invention also provides a resource allocation device, comprising:

[0105] The first acquisition module is used to acquire resource allocation parameter information corresponding to at least one terminal within the target area;

[0106] The first processing module is used to obtain first information on the traffic demand of each terminal in at least one time period based on the resource allocation parameter information.

[0107] The second processing module is used to obtain resource allocation result information corresponding to each terminal based on the resource allocation parameter information and the first information;

[0108] The resource allocation parameter information includes at least one of the following: base station information corresponding to the target area, terminal information corresponding to the target area, sub-band information of spectrum resources shared by base stations in the target area, and upper limit value of transmit power for each base station.

[0109] The first information includes: information on the terminal's traffic demand during the current time period;

[0110] The resource allocation result information includes at least one of the following: base station allocation information, sub-frequency band allocation information, and transmit power allocation information.

[0111] Optionally, obtaining the first information on the traffic demand of each terminal within at least one time period based on the resource allocation parameter information includes:

[0112] For at least one time period prior to the current time period of the terminal, a first operation is performed to obtain the corresponding first terminal codes;

[0113] Based on the first terminal encoding, the encoding history sequence is obtained;

[0114] Based on the encoded history sequence, the first information of the terminal's traffic demand in the current time period is obtained;

[0115] The first operation includes: constructing a corresponding first relationship graph based on the resource allocation parameter information corresponding to the time period; and obtaining a first terminal code corresponding to the time period based on the first relationship graph.

[0116] Optionally, constructing the corresponding first relationship graph based on the resource allocation parameter information corresponding to the time period includes:

[0117] Based on the resource allocation parameter information corresponding to the time period, a first relationship graph is constructed with the first relevant information of the base station and the second relevant information of the terminal as nodes.

[0118] The first relevant information includes: the base station information and the upper limit of the transmission power, wherein the base station information includes: the identity information of the base station;

[0119] The second relevant information includes: the terminal information, which includes: second information on the unmet traffic demand of the terminal and the number of remaining time periods after the current time period of the terminal;

[0120] In the first relationship diagram, the weights between the first relevant information and the second relevant information are determined based on the signal attenuation factor between the base station and the terminal.

[0121] Optionally, obtaining the first information on the terminal's traffic demand in the current time period based on the encoded historical sequence includes:

[0122] Based on the encoded history sequence, the terminal state change trend code is obtained;

[0123] Based on the terminal status change trend encoding, the first information of the terminal's traffic demand in the current time period is obtained.

[0124] Optionally, obtaining the first terminal code corresponding to the time period based on the first relationship graph includes:

[0125] Using a graph neural network, the first terminal code corresponding to the time period is obtained based on the first relationship graph;

[0126] And / or, obtaining the terminal state change trend code based on the encoded history sequence includes:

[0127] Using a recurrent neural network, the terminal state change trend code is obtained based on the encoded history sequence.

[0128] Optionally, obtaining the first information on the terminal's traffic demand within the current time period based on the terminal state change trend encoding includes:

[0129] Using a two-layer fully connected network, the first probability parameter information is obtained by encoding the terminal state change trend.

[0130] Based on the first probability parameter information, the first Gaussian probability distribution is obtained;

[0131] From the first Gaussian probability distribution, the traffic demand of the terminal in the current time period is determined by sampling, and the first information of the traffic demand of the terminal in the current time period is obtained.

[0132] Optionally, obtaining the resource allocation result information corresponding to each terminal based on the resource allocation parameter information and the first information includes:

[0133] Based on the resource allocation parameter information and the first information, construct the target relationship graph;

[0134] Based on the target relationship diagram, the reference code is obtained;

[0135] Based on the reference code, the resource allocation result information corresponding to the terminal is obtained.

[0136] Optionally, the target relationship graph includes a second relationship graph;

[0137] The step of constructing a target relationship graph based on the resource allocation parameter information and the first information includes:

[0138] Based on the resource allocation parameter information and the first information, a second relationship graph is constructed with the third relevant information of the base station and the fourth relevant information of the terminal as nodes;

[0139] And / or, the reference encoding includes: a first base station encoding and a second terminal encoding;

[0140] The step of obtaining the reference code based on the target relationship diagram includes:

[0141] Using a graph neural network, the first base station code and the second terminal code are obtained based on the second relationship graph;

[0142] And / or, obtaining the resource allocation result information corresponding to the terminal based on the reference encoding includes:

[0143] The first base station code and the second terminal code are concatenated to obtain the first concatenated code;

[0144] Using a two-layer fully connected network, the first traffic estimate of the terminal accessing the base station is obtained based on the first splicing code;

[0145] Based on the first traffic estimate, the target access base station of the terminal is determined, and the base station allocation information corresponding to the terminal is obtained;

[0146] The third relevant information includes: the base station information and the upper limit of the transmission power, wherein the base station information includes: the identity information of the base station;

[0147] The fourth relevant information includes: the first information;

[0148] In the second relationship diagram, the weights between the third relevant information and the fourth relevant information are determined based on the signal attenuation factor between the base station and the terminal.

[0149] Optionally, the target relationship graph includes a third relationship graph;

[0150] The step of constructing a target relationship graph based on the resource allocation parameter information and the first information includes:

[0151] Based on the resource allocation parameter information and the first information, a third relationship graph is constructed with the fifth relevant information of the base station and the sixth relevant information of the terminal as nodes;

[0152] And / or, the reference encoding includes: a third terminal encoding;

[0153] The step of obtaining the reference code based on the target relationship diagram includes:

[0154] Using a graph neural network, the first sub-encoder is obtained based on the second relationship graph;

[0155] Using a graph neural network, the second sub-encoder is obtained based on the third relationship graph;

[0156] The first sub-encode and the second sub-encode are concatenated to obtain the third terminal code;

[0157] And / or, obtaining the resource allocation result information corresponding to the terminal based on the reference encoding includes:

[0158] Using a two-layer fully connected network, the second traffic estimate for the terminal access sub-band is obtained based on the third terminal coding;

[0159] Based on the second traffic estimate, the target access sub-band of the terminal is determined, and the sub-band allocation information corresponding to the terminal is obtained;

[0160] The fifth relevant information includes: the base station information and the upper limit of the transmission power, wherein the base station information includes: the identity information of the base station;

[0161] The sixth relevant information includes: the first information;

[0162] In the third relationship diagram, the weights between the fifth and sixth related information are determined based on whether there is a connection between the base station and the terminal.

[0163] Optionally, the target relationship graph includes a fourth relationship graph;

[0164] The step of constructing a target relationship graph based on the resource allocation parameter information and the first information includes:

[0165] Based on the resource allocation parameter information and the first information, a fourth relationship graph is constructed with the seventh relevant information of the base station and the eighth relevant information of the terminal as nodes;

[0166] And / or, the reference encoding includes: a fourth terminal encoding;

[0167] The step of obtaining the reference code based on the target relationship diagram includes:

[0168] Using a graph neural network, the third sub-encoder is obtained based on the second relationship graph;

[0169] Using a graph neural network, the fourth sub-encoder is obtained based on the third relation graph;

[0170] Using a graph neural network, the fifth sub-encoder is obtained based on the fourth relation graph;

[0171] The third sub-encode, the fourth sub-encode, and the fifth sub-encode are concatenated to obtain the fourth terminal code;

[0172] And / or, obtaining the resource allocation result information corresponding to the terminal based on the reference encoding includes:

[0173] Using a two-layer fully connected network, the second probability parameter information is obtained based on the fourth terminal encoding;

[0174] Based on the second probability parameter information, the second Gaussian probability distribution is obtained;

[0175] Based on the second Gaussian probability distribution, the transmit power allocation information corresponding to the terminal is obtained;

[0176] The seventh relevant information includes: the base station information and the upper limit of the transmission power, wherein the base station information includes: the identity information of the base station;

[0177] The eighth relevant information includes: the first information;

[0178] In the fourth relationship diagram, the weights between the seventh and eighth related information are determined based on whether there is a connection between the sub-frequency band of the base station and the terminal.

[0179] Optionally, obtaining the transmit power allocation information corresponding to the terminal based on the second Gaussian probability distribution includes:

[0180] The power reference value corresponding to the terminal is determined by sampling from the second Gaussian probability distribution;

[0181] Based on the power reference value, the corresponding transmit power percentage of the terminal is obtained;

[0182] Based on the transmission power ratio and the upper limit of transmission power, the transmission power allocation information corresponding to the terminal is obtained.

[0183] Optional, also includes:

[0184] The third processing module is used to obtain the traffic value allocated to the corresponding terminal based on the resource allocation result information.

[0185] The first adjustment module is used to adjust the relevant network for resource allocation based on the traffic value;

[0186] The relevant network includes at least one of the following:

[0187] Graph neural networks;

[0188] Recurrent neural networks;

[0189] Two-layer fully connected network.

[0190] Optionally, the process of adjusting resource allocation based on the traffic value includes:

[0191] Based on the traffic value, obtain the traffic loss value corresponding to the resource allocation result information;

[0192] Based on the aforementioned traffic loss value, adjust the relevant network for resource allocation.

[0193] Optionally, obtaining the traffic loss value corresponding to the resource allocation result information based on the traffic value includes:

[0194] Using the first method, based on the traffic value, obtain the traffic loss value corresponding to the resource allocation result information;

[0195] The first method includes: converting constrained adjustment conditions into unconstrained adjustment conditions;

[0196] The constrained adjustment conditions include: ensuring that the traffic allocated to each terminal in the first time period is greater than the traffic demand in the first time period, and maximizing the total traffic obtained by all terminals in the first time period;

[0197] The unconstrained adjustment conditions include: maximizing the total traffic of all terminals in the first time period, and maximizing the weighted sum of the traffic requirements met by all terminals in the first time period.

[0198] Optionally, the base station information in the resource allocation parameter information corresponding to the terminal is: the base station information of the first base station in the target area whose first distance from the terminal is less than a first threshold;

[0199] And / or, the terminal information in the resource allocation parameter information corresponding to the terminal is: the terminal information of a second terminal in the target area whose second distance from the terminal is less than a second threshold;

[0200] And / or, the sub-frequency band information in the resource allocation parameter information corresponding to the terminal is: the sub-frequency band information of the spectrum resources shared by the first base station;

[0201] And / or, the upper limit of the transmit power in the resource allocation parameter information corresponding to the terminal is: the upper limit of the transmit power of the first base station.

[0202] This invention also provides a resource allocation device, including: a processor;

[0203] The processor is used to acquire resource allocation parameter information corresponding to at least one terminal within the target area;

[0204] Based on the resource allocation parameter information, the first information on the traffic demand of each terminal within at least one time period is obtained;

[0205] Based on the resource allocation parameter information and the first information, the resource allocation result information corresponding to each terminal is obtained;

[0206] The resource allocation parameter information includes at least one of the following: base station information corresponding to the target area, terminal information corresponding to the target area, sub-band information of spectrum resources shared by base stations in the target area, and upper limit value of transmit power for each base station.

[0207] The first information includes: information on the terminal's traffic demand during the current time period;

[0208] The resource allocation result information includes at least one of the following: base station allocation information, sub-frequency band allocation information, and transmit power allocation information.

[0209] Optionally, obtaining the first information on the traffic demand of each terminal within at least one time period based on the resource allocation parameter information includes:

[0210] For at least one time period prior to the current time period of the terminal, a first operation is performed to obtain the corresponding first terminal codes;

[0211] Based on the first terminal encoding, the encoding history sequence is obtained;

[0212] Based on the encoded history sequence, the first information of the terminal's traffic demand in the current time period is obtained;

[0213] The first operation includes: constructing a corresponding first relationship graph based on the resource allocation parameter information corresponding to the time period; and obtaining a first terminal code corresponding to the time period based on the first relationship graph.

[0214] Optionally, constructing the corresponding first relationship graph based on the resource allocation parameter information corresponding to the time period includes:

[0215] Based on the resource allocation parameter information corresponding to the time period, a first relationship graph is constructed with the first relevant information of the base station and the second relevant information of the terminal as nodes.

[0216] The first relevant information includes: the base station information and the upper limit of the transmission power, wherein the base station information includes: the identity information of the base station;

[0217] The second relevant information includes: the terminal information, which includes: second information on the unmet traffic demand of the terminal and the number of remaining time periods after the current time period of the terminal;

[0218] In the first relationship diagram, the weights between the first relevant information and the second relevant information are determined based on the signal attenuation factor between the base station and the terminal.

[0219] Optionally, obtaining the first information on the terminal's traffic demand in the current time period based on the encoded historical sequence includes:

[0220] Based on the encoded history sequence, the terminal state change trend code is obtained;

[0221] Based on the terminal status change trend encoding, the first information of the terminal's traffic demand in the current time period is obtained.

[0222] Optionally, obtaining the first terminal code corresponding to the time period based on the first relationship graph includes:

[0223] Using a graph neural network, the first terminal code corresponding to the time period is obtained based on the first relationship graph;

[0224] And / or, obtaining the terminal state change trend code based on the encoded history sequence includes:

[0225] Using a recurrent neural network, the terminal state change trend code is obtained based on the encoded history sequence.

[0226] Optionally, obtaining the first information on the terminal's traffic demand within the current time period based on the terminal state change trend encoding includes:

[0227] Using a two-layer fully connected network, the first probability parameter information is obtained by encoding the terminal state change trend.

[0228] Based on the first probability parameter information, the first Gaussian probability distribution is obtained;

[0229] From the first Gaussian probability distribution, the traffic demand of the terminal in the current time period is determined by sampling, and the first information of the traffic demand of the terminal in the current time period is obtained.

[0230] Optionally, obtaining the resource allocation result information corresponding to each terminal based on the resource allocation parameter information and the first information includes:

[0231] Based on the resource allocation parameter information and the first information, construct the target relationship graph;

[0232] Based on the target relationship diagram, the reference code is obtained;

[0233] Based on the reference code, the resource allocation result information corresponding to the terminal is obtained.

[0234] Optionally, the target relationship graph includes a second relationship graph;

[0235] The step of constructing a target relationship graph based on the resource allocation parameter information and the first information includes:

[0236] Based on the resource allocation parameter information and the first information, a second relationship graph is constructed with the third relevant information of the base station and the fourth relevant information of the terminal as nodes;

[0237] And / or, the reference encoding includes: a first base station encoding and a second terminal encoding;

[0238] The step of obtaining the reference code based on the target relationship diagram includes:

[0239] Using a graph neural network, the first base station code and the second terminal code are obtained based on the second relationship graph;

[0240] And / or, obtaining the resource allocation result information corresponding to the terminal based on the reference encoding includes:

[0241] The first base station code and the second terminal code are concatenated to obtain the first concatenated code;

[0242] Using a two-layer fully connected network, the first traffic estimate of the terminal accessing the base station is obtained based on the first splicing code;

[0243] Based on the first traffic estimate, the target access base station of the terminal is determined, and the base station allocation information corresponding to the terminal is obtained;

[0244] The third relevant information includes: the base station information and the upper limit of the transmission power, wherein the base station information includes: the identity information of the base station;

[0245] The fourth relevant information includes: the first information;

[0246] In the second relationship diagram, the weights between the third relevant information and the fourth relevant information are determined based on the signal attenuation factor between the base station and the terminal.

[0247] Optionally, the target relationship graph includes a third relationship graph;

[0248] The step of constructing a target relationship graph based on the resource allocation parameter information and the first information includes:

[0249] Based on the resource allocation parameter information and the first information, a third relationship graph is constructed with the fifth relevant information of the base station and the sixth relevant information of the terminal as nodes;

[0250] And / or, the reference encoding includes: a third terminal encoding;

[0251] The step of obtaining the reference code based on the target relationship diagram includes:

[0252] Using a graph neural network, the first sub-encoder is obtained based on the second relationship graph;

[0253] Using a graph neural network, the second sub-encoder is obtained based on the third relationship graph;

[0254] The first sub-encode and the second sub-encode are concatenated to obtain the third terminal code;

[0255] And / or, obtaining the resource allocation result information corresponding to the terminal based on the reference encoding includes:

[0256] Using a two-layer fully connected network, the second traffic estimate for the terminal access sub-band is obtained based on the third terminal coding;

[0257] Based on the second traffic estimate, the target access sub-band of the terminal is determined, and the sub-band allocation information corresponding to the terminal is obtained;

[0258] The fifth relevant information includes: the base station information and the upper limit of the transmission power, wherein the base station information includes: the identity information of the base station;

[0259] The sixth relevant information includes: the first information;

[0260] In the third relationship diagram, the weights between the fifth relevant information and the sixth relevant information are determined based on whether there is a connection between the base station and the terminal.

[0261] Optionally, the target relationship graph includes a fourth relationship graph;

[0262] The step of constructing a target relationship graph based on the resource allocation parameter information and the first information includes:

[0263] Based on the resource allocation parameter information and the first information, a fourth relationship graph is constructed with the seventh relevant information of the base station and the eighth relevant information of the terminal as nodes;

[0264] And / or, the reference encoding includes: a fourth terminal encoding;

[0265] The step of obtaining the reference code based on the target relationship diagram includes:

[0266] Using a graph neural network, the third sub-encoder is obtained based on the second relationship graph;

[0267] Using a graph neural network, the fourth sub-encoder is obtained based on the third relation graph;

[0268] Using a graph neural network, the fifth sub-encoder is obtained based on the fourth relation graph;

[0269] The third sub-encode, the fourth sub-encode, and the fifth sub-encode are concatenated to obtain the fourth terminal code;

[0270] And / or, obtaining the resource allocation result information corresponding to the terminal based on the reference encoding includes:

[0271] Using a two-layer fully connected network, the second probability parameter information is obtained based on the fourth terminal encoding;

[0272] Based on the second probability parameter information, the second Gaussian probability distribution is obtained;

[0273] Based on the second Gaussian probability distribution, the transmit power allocation information corresponding to the terminal is obtained;

[0274] The seventh relevant information includes: the base station information and the upper limit of the transmission power, wherein the base station information includes: the identity information of the base station;

[0275] The eighth relevant information includes: the first information;

[0276] In the fourth relationship diagram, the weights between the seventh and eighth related information are determined based on whether there is a connection between the sub-frequency band of the base station and the terminal.

[0277] Optionally, obtaining the transmit power allocation information corresponding to the terminal based on the second Gaussian probability distribution includes:

[0278] The power reference value corresponding to the terminal is determined by sampling from the second Gaussian probability distribution;

[0279] Based on the power reference value, the corresponding transmit power percentage of the terminal is obtained;

[0280] Based on the transmission power ratio and the upper limit of transmission power, the transmission power allocation information corresponding to the terminal is obtained.

[0281] Optionally, the processor is further configured to:

[0282] Based on the resource allocation result information, the traffic value allocated to the corresponding terminal is obtained;

[0283] Adjust the relevant network for resource allocation based on the traffic value;

[0284] The relevant network includes at least one of the following:

[0285] Graph neural networks;

[0286] Recurrent neural networks;

[0287] Two-layer fully connected network.

[0288] Optionally, the process of adjusting resource allocation based on the traffic value includes:

[0289] Based on the traffic value, obtain the traffic loss value corresponding to the resource allocation result information;

[0290] Based on the aforementioned traffic loss value, adjust the relevant network for resource allocation.

[0291] Optionally, obtaining the traffic loss value corresponding to the resource allocation result information based on the traffic value includes:

[0292] Using the first method, based on the traffic value, obtain the traffic loss value corresponding to the resource allocation result information;

[0293] The first method includes: converting constrained adjustment conditions into unconstrained adjustment conditions;

[0294] The constrained adjustment conditions include: ensuring that the traffic allocated to each terminal in the first time period is greater than the traffic demand in the first time period, and maximizing the total traffic obtained by all terminals in the first time period;

[0295] The unconstrained adjustment conditions include: maximizing the total traffic of all terminals in the first time period, and maximizing the weighted sum of the traffic requirements met by all terminals in the first time period.

[0296] Optionally, the base station information in the resource allocation parameter information corresponding to the terminal is: the base station information of the first base station in the target area whose first distance from the terminal is less than a first threshold;

[0297] And / or, the terminal information in the resource allocation parameter information corresponding to the terminal is: the terminal information of a second terminal in the target area whose second distance from the terminal is less than a second threshold;

[0298] And / or, the sub-frequency band information in the resource allocation parameter information corresponding to the terminal is: the sub-frequency band information of the spectrum resources shared by the first base station;

[0299] And / or, the upper limit of the transmit power in the resource allocation parameter information corresponding to the terminal is: the upper limit of the transmit power of the first base station.

[0300] This invention also provides a resource allocation device, including a memory, a processor, and a program stored in the memory and executable on the processor; when the processor executes the program, it implements the resource allocation method described above.

[0301] This invention also provides a readable storage medium storing a program that, when executed by a processor, implements the steps in the resource allocation method described above.

[0302] The beneficial effects of the above-described technical solution of the present invention are as follows:

[0303] In the above scheme, the resource allocation method obtains resource allocation parameter information corresponding to at least one terminal in the target area; based on the resource allocation parameter information, it obtains first information on the traffic demand of each terminal in at least one time period; based on the resource allocation parameter information and the first information, it obtains resource allocation result information corresponding to each terminal; wherein, the resource allocation parameter information includes at least one of the following: base station information corresponding to the target area, terminal information corresponding to the target area, sub-frequency band information of spectrum resources shared by base stations in the target area, and upper limit value of transmit power of each base station; the first information includes: information on the traffic demand of the terminal in the current time period; the resource allocation result information includes at least one of the following: base station allocation information, sub-frequency band allocation information, and transmit power allocation information; it can realize resource allocation based on traffic demand to effectively meet traffic demand, and decomposes resource allocation into base station allocation, sub-frequency band allocation, and transmit power allocation, thereby simplifying the resource allocation strategy, reducing the difficulty of the scheme, and effectively solving the problems of existing resource allocation schemes being unable to effectively meet traffic demand and being complex. Attached Figure Description

[0304] Figure 1 This is a schematic diagram of the resource allocation method according to an embodiment of the present invention;

[0305] Figure 2 This is a schematic diagram illustrating the implementation framework of the resource allocation method according to an embodiment of the present invention;

[0306] Figure 3 This is a schematic diagram of the resource allocation device according to an embodiment of the present invention;

[0307] Figure 4 This is a schematic diagram of the resource allocation device structure according to an embodiment of the present invention. Detailed Implementation

[0308] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0309] This invention addresses the problems of existing resource allocation schemes failing to effectively meet traffic demands and being overly complex, by providing a resource allocation method, such as... Figure 1 As shown, it includes:

[0310] Step 11: Obtain resource allocation parameter information corresponding to at least one terminal in the target area;

[0311] Step 12: Based on the resource allocation parameter information, obtain the first information on the traffic demand of each terminal within at least one time period;

[0312] Step 13: Based on the resource allocation parameter information and the first information, obtain the resource allocation result information corresponding to each terminal; wherein, the resource allocation parameter information includes at least one of the following: base station information corresponding to the target area, terminal information corresponding to the target area, sub-frequency band information of the spectrum resources shared by the base stations in the target area, and the upper limit value of the transmit power of each base station; the first information includes: information on the traffic demand of the terminal in the current time period; the resource allocation result information includes at least one of the following: base station allocation information, sub-frequency band allocation information, and transmit power allocation information.

[0313] The resource allocation method provided in this embodiment of the invention obtains resource allocation parameter information corresponding to at least one terminal in a target area; obtains first information on the traffic demand of each terminal in at least one time period based on the resource allocation parameter information; and obtains resource allocation result information corresponding to each terminal based on the resource allocation parameter information and the first information. The resource allocation parameter information includes at least one of the following: base station information corresponding to the target area, terminal information corresponding to the target area, sub-band information of spectrum resources shared by base stations in the target area, and upper limit values ​​of the transmit power of each base station. The first information includes: information on the traffic demand of the terminal in the current time period. The resource allocation result information includes at least one of: base station allocation information, sub-band allocation information, and transmit power allocation information. This method enables resource allocation based on traffic demand to effectively meet traffic demand, and decomposes resource allocation into base station allocation, sub-band allocation, and transmit power allocation, thereby simplifying the resource allocation strategy, reducing the complexity of the scheme, and effectively solving the problems of existing resource allocation schemes failing to effectively meet traffic demand and being overly complex.

[0314] The step of obtaining the first information on the traffic demand of each terminal within at least one time period based on the resource allocation parameter information includes: performing a first operation on at least one time period prior to the current time period of the terminal to obtain a corresponding first terminal code; obtaining an encoding history sequence based on the first terminal code; and obtaining the first information on the traffic demand of the terminal within the current time period based on the encoding history sequence. The first operation includes: constructing a corresponding first relationship graph based on the resource allocation parameter information corresponding to the time period; and obtaining the first terminal code corresponding to the time period based on the first relationship graph.

[0315] This allows us to accurately obtain the first piece of information about traffic demand.

[0316] In this embodiment of the invention, constructing a corresponding first relationship graph based on the resource allocation parameter information corresponding to the time period includes: constructing a first relationship graph with first relevant information of the base station and second relevant information of the terminal as nodes based on the resource allocation parameter information corresponding to the time period; wherein, the first relevant information includes: the base station information and the upper limit of the transmission power, the base station information including: the identity information of the base station; the second relevant information includes: the terminal information, the terminal information including: second information on the unmet traffic demand of the terminal and the number of remaining time periods after the current time period of the terminal; in the first relationship graph, the weight between the first relevant information and the second relevant information is determined based on the signal attenuation factor between the base station and the terminal.

[0317] This allows for the accurate construction of the first relationship graph.

[0318] The step of obtaining the first information on the terminal's traffic demand in the current time period based on the encoded historical sequence includes: obtaining the terminal state change trend code based on the encoded historical sequence; and obtaining the first information on the terminal's traffic demand in the current time period based on the terminal state change trend code.

[0319] This allows for a more accurate acquisition of primary information regarding traffic demand.

[0320] In this embodiment of the invention, obtaining the first terminal code corresponding to the time period based on the first relationship graph includes: using a graph neural network to obtain the first terminal code corresponding to the time period based on the first relationship graph; and / or, obtaining the terminal state change trend code based on the encoding history sequence includes: using a recurrent neural network to obtain the terminal state change trend code based on the encoding history sequence.

[0321] This allows for the accurate acquisition of the first terminal code and / or the terminal status change trend code.

[0322] The step of obtaining the first information of the terminal's traffic demand in the current time period based on the terminal state change trend encoding includes: using a two-layer fully connected network to obtain first probability parameter information based on the terminal state change trend encoding; obtaining a first Gaussian probability distribution based on the first probability parameter information; and sampling from the first Gaussian probability distribution to determine the terminal's traffic demand in the current time period, thereby obtaining the first information of the terminal's traffic demand in the current time period.

[0323] This allows us to obtain specific information about the terminal's traffic demand within the current time period. This first probability parameter information may include, but is not limited to, the mean and standard deviation of a Gaussian probability distribution.

[0324] In this embodiment of the invention, obtaining the resource allocation result information corresponding to each terminal based on the resource allocation parameter information and the first information includes: constructing a target relationship graph based on the resource allocation parameter information and the first information; obtaining a reference code based on the target relationship graph; and obtaining the resource allocation result information corresponding to the terminal based on the reference code.

[0325] This allows us to accurately obtain the resource allocation results for the corresponding terminal.

[0326] Resource allocation can include base station allocation, sub-band allocation, and power allocation, which are illustrated with examples below.

[0327] Regarding base station allocation, the target relationship graph includes a second relationship graph; constructing the target relationship graph based on the resource allocation parameter information and the first information includes: constructing a second relationship graph with the third relevant information of the base station and the fourth relevant information of the terminal as nodes based on the resource allocation parameter information and the first information; and / or, the reference encoding includes: a first base station encoding and a second terminal encoding; obtaining the reference encoding based on the target relationship graph includes: using a graph neural network to obtain the first base station encoding and the second terminal encoding based on the second relationship graph; and / or, obtaining the resource allocation result information corresponding to the terminal based on the reference encoding includes: […]. The first base station code and the second terminal code are concatenated to obtain a first concatenated code; using a two-layer fully connected network, a first traffic estimate of the terminal accessing the base station is obtained based on the first concatenated code; based on the first traffic estimate, the target access base station of the terminal is determined, and the base station allocation information corresponding to the terminal is obtained; wherein, the third relevant information includes: the base station information and the upper limit of the transmit power, the base station information includes: the identity information of the base station; the fourth relevant information includes: the first information; in the second relationship diagram, the weight between the third relevant information and the fourth relevant information is determined based on the signal attenuation factor between the base station and the terminal.

[0328] This allows for accurate base station allocation.

[0329] Regarding sub-frequency band allocation, the target relationship graph includes a third relationship graph; constructing the target relationship graph based on the resource allocation parameter information and the first information includes: constructing a third relationship graph with the fifth relevant information of the base station and the sixth relevant information of the terminal as nodes based on the resource allocation parameter information and the first information; and / or, the reference coding includes: a third terminal coding; obtaining the reference coding based on the target relationship graph includes: using a graph neural network to obtain a first sub-code based on the second relationship graph; using a graph neural network to obtain a second sub-code based on the third relationship graph; concatenating the first sub-code and the second sub-code to obtain the third terminal coding; and / or, The step of obtaining the resource allocation result information corresponding to the terminal based on the reference coding includes: using a two-layer fully connected network, obtaining a second traffic estimate of the terminal access sub-frequency band based on the third terminal coding; determining the target access sub-frequency band of the terminal based on the second traffic estimate, and obtaining the sub-frequency band allocation information corresponding to the terminal; wherein, the fifth related information includes: the base station information and the upper limit of the transmit power, the base station information includes: the identity information of the base station; the sixth related information includes: the first information; in the third relationship diagram, the weight between the fifth related information and the sixth related information is determined based on whether there is a connection between the base station and the terminal.

[0330] This allows for accurate sub-band allocation.

[0331] Regarding power allocation, the target relationship graph includes a fourth relationship graph; constructing the target relationship graph based on the resource allocation parameter information and the first information includes: constructing a fourth relationship graph with the seventh relevant information of the base station and the eighth relevant information of the terminal as nodes based on the resource allocation parameter information and the first information; and / or, the reference coding includes: a fourth terminal coding; obtaining the reference coding based on the target relationship graph includes: using a graph neural network to obtain a third sub-code based on the second relationship graph; using a graph neural network to obtain a fourth sub-code based on the third relationship graph; using a graph neural network to obtain a fifth sub-code based on the fourth relationship graph; concatenating the third, fourth, and fifth sub-codes to obtain the... The fourth terminal encoding; and / or, obtaining the resource allocation result information corresponding to the terminal based on the reference encoding, includes: using a two-layer fully connected network, obtaining second probability parameter information based on the fourth terminal encoding; obtaining a second Gaussian probability distribution based on the second probability parameter information; obtaining the transmit power allocation information corresponding to the terminal based on the second Gaussian probability distribution; wherein, the seventh related information includes: the base station information and the upper limit of transmit power, the base station information includes: the identity information of the base station; the eighth related information includes: the first information; in the fourth relationship diagram, the weight between the seventh related information and the eighth related information is determined based on whether the sub-frequency band of the base station is connected to the terminal.

[0332] This allows for accurate allocation of transmission power. The second probability parameter information may include, but is not limited to, the mean and standard deviation of a Gaussian probability distribution.

[0333] The step of obtaining the transmission power allocation information corresponding to the terminal based on the second Gaussian probability distribution includes: sampling and determining the power reference value corresponding to the terminal from the second Gaussian probability distribution; obtaining the transmission power ratio corresponding to the terminal based on the power reference value; and obtaining the transmission power allocation information corresponding to the terminal based on the transmission power ratio and the upper limit value of the transmission power.

[0334] This allows for more accurate allocation of transmission power. Specifically, "obtaining the transmission power percentage corresponding to the terminal based on the power reference value" can include, but is not limited to, using a softmax layer to obtain the transmission power percentage corresponding to the terminal based on the power reference value and a constant (which can be set empirically or arbitrarily).

[0335] Furthermore, the resource allocation method further includes: obtaining the traffic value allocated to the corresponding terminal based on the resource allocation result information; adjusting the relevant network for resource allocation based on the traffic value; wherein the relevant network includes at least one of the following: graph neural network; recurrent neural network; two-layer fully connected network.

[0336] This can improve the accuracy of subsequent resource allocation.

[0337] The step of adjusting the relevant network for resource allocation based on the traffic value includes: obtaining the traffic loss value corresponding to the resource allocation result information based on the traffic value; and adjusting the relevant network for resource allocation based on the traffic loss value.

[0338] This allows for accurate adjustment of the relevant networks.

[0339] In this embodiment of the invention, obtaining the traffic loss value corresponding to the resource allocation result information based on the traffic value includes: using a first method to obtain the traffic loss value corresponding to the resource allocation result information based on the traffic value; wherein, the first method includes: converting a constrained adjustment condition into an unconstrained adjustment condition; the constrained adjustment condition includes: ensuring that the traffic allocated to each terminal in the first time period is greater than the traffic demand in the first time period, and maximizing the total traffic obtained by all terminals in the first time period; the unconstrained adjustment condition includes: maximizing the total traffic of all terminals in the first time period, and maximizing the weighted sum of the traffic demand satisfied by all terminals in the first time period.

[0340] This allows for accurate determination of flow loss values, and this method of obtaining flow loss values ​​is applicable to situations involving power allocation.

[0341] Wherein, the base station information in the resource allocation parameter information corresponding to the terminal is: the base station information of the first base station in the target area whose first distance from the terminal is less than a first threshold; and / or, the terminal information in the resource allocation parameter information corresponding to the terminal is: the terminal information of the second terminal in the target area whose second distance from the terminal is less than a second threshold; and / or, the sub-frequency band information in the resource allocation parameter information corresponding to the terminal is: the sub-frequency band information of the spectrum resources shared by the first base station; and / or, the upper limit value of the transmit power in the resource allocation parameter information corresponding to the terminal is: the upper limit value of the transmit power of the first base station.

[0342] This avoids excessive consumption of computing resources such as memory and video memory as the number of base stations and terminals increases, and increases the possibility of solving large-scale problems.

[0343] The resource allocation method provided in the embodiments of the present invention will be illustrated below with examples. The recurrent neural network is LSTM (Long Short-Term Memory) artificial neural network, and the graph neural network is graph convolutional network.

[0344] To address the aforementioned technical problems, and considering that existing wireless network resource allocation technologies have a relatively small decision space (e.g., only involving sub-band allocation and addressing small-scale optimization issues, thus failing to comprehensively optimize large-scale problems), this invention provides a resource allocation method, specifically a wireless network resource allocation scheme based on security reinforcement learning. This scheme simultaneously optimizes base station selection, sub-band selection, and power allocation for large-scale problems (hundreds of base stations and thousands of users).

[0345] Specifically, let's assume the wireless network resource allocation scenario is as follows: For users (corresponding to the aforementioned terminals) within a certain area and time period, wireless network resources, including base stations, sub-frequency bands, and transmit power, are allocated to users based on their location and traffic requirements. Specifically, assume that within a certain area (corresponding to the aforementioned target area), there are M base stations (included in the aforementioned base station information), N active users (included in the aforementioned terminal information), and K sub-frequency bands (included in the aforementioned sub-frequency band information), with each base station having its own transmit power limit. (Corresponding to the aforementioned transmit power upper limit, the transmit power upper limit for base station 1 is P1, and other base stations are similar). Each user can only access one sub-frequency band of one base station, and the total transmit power allocated to all users accessing the same base station cannot exceed the transmit power upper limit of that base station. Considering T time slices (i.e., time periods), at the beginning of each time slice, wireless network resources are allocated—connecting each user to a suitable base station, allocating a suitable sub-frequency band, and allocating appropriate power to meet the total traffic demand of users in the T time slices. (The total traffic requirement for User 1 is D1, and the same applies to other users) while maximizing cumulative throughput. The formula for calculating throughput is as follows:

[0346]

[0347] Where B represents network bandwidth, Indicates user u within time slice t n Connected to base station b m The signal interference ratio in sub-band k is calculated using the following formula:

[0348]

[0349] in, User u within time slice t n Is it connected to base station b?m , User u within time slice t n Is it connected to base station b? M sub-band k, User u within time slice t n At base station b M The power allocated to sub-band k, Indicates base station b within time slice t M To user u n Signal attenuation factor, p noise Indicates background noise power. Whether all users have accessed base station b within time slice t m′ , Whether all users have accessed base station b within time slice t m′ sub-band k, All users within time slice t at base station b m′ The power allocated to sub-band k, Indicates base station b within time slice t m′ To user u n The signal attenuation factor. User u n Within time slice t, the satisfied flow rate is:

[0350]

[0351] In this scheme, the wireless network resource allocation problem can be formally described as a constrained Markov decision process:

[0352] state s t Includes: all base stations Location information All users Historical location information Signal attenuation factor from each base station to each user Unmet traffic needs of each user

[0353] Actions include: base station allocation. That is, within time slice t, give user u n Allocated base station b m Information (corresponding to the above base station allocation information), sub-frequency band allocation status That is, within time slice t, give user u n Information on the allocated sub-band k (corresponding to the sub-band allocation information mentioned above) and transmit power allocation. That is, within time slice t, give user un Information on the allocated transmission power (corresponding to the transmission power allocation information mentioned above).

[0354] The state transition function is as follows: update the location of all users, update the corresponding signal attenuation factor, and update the unmet traffic demand of users.

[0355] The reward is: state s t Execute a certain combination of actions The total throughput H of the later system t .

[0356] The constraint is: the total traffic obtained by the user within the preset total time slice T satisfies:

[0357]

[0358] Based on the above, this solution considers that the wireless network resource allocation problem is a complex constrained optimization problem. Therefore, the original problem is broken down into two sequential decision problems: demand decomposition and immediate demand satisfaction. The overall process is as follows: Figure 2 The diagram illustrates a decision-making framework for wireless network resource allocation. The former (demand decomposition) determines how much traffic demand needs to be met in the current time slice based on user historical trajectories (a sequence of user historical locations), remaining traffic demand (i.e., the unmet traffic demand mentioned above), and base station locations. The latter (immediate demand fulfillment) allocates resources based on base station and user locations (e.g., real-time coordinates) to satisfy the immediate traffic demand set by the former (i.e., the unmet demand). Figure 2 (The immediate needs in the process). The latter is further decomposed into three sub-tasks: base station allocation, sub-band allocation, and (transmit) power allocation. These three sub-tasks can be configured with an inherent sequence relationship—the decision of the subsequent task depends on the decision result of the preceding task. The above task decomposition method allows the base station allocation strategy, sub-band allocation strategy, and power allocation strategy to only consider single-step allocation optimization, simplifying the training difficulty of these strategies and making the construction of these modules more flexible. Among them, resource allocation can obtain at least one of the following: base station allocation information (such as the access base station number, i.e., the target access base station number), sub-band allocation information (such as the sub-band number), and transmit power allocation information; the location of the base station and the terminal can be used to obtain path loss and calculate subsequent nearest neighbor relationships, but is not limited to this; the above base station location includes Figure 2 The base station information.

[0359] Furthermore, considering that the scale of the optimization problem increases with the number of base stations and users, leading to excessive consumption of computing resources such as memory and GPU memory; and that the scope of influence of each decision in the wireless network resource allocation problem is very limited—for example, increasing or decreasing the transmit power allocated to a user only affects the signal interference of nearby users—each module in this scheme (the modules corresponding to the demand decomposition strategy, base station allocation strategy, sub-band allocation strategy, and power allocation strategy) makes decisions for each user sequentially during execution. Each decision uses only neighboring base stations and users to construct a (relationship) graph and performs corresponding encoding. Correspondingly, when updating the demand decomposition strategy using rewards, the rewards are also reduced to rewards related to these neighboring users, i.e., only the sum of the throughput of neighboring users is considered as the reward; the constraints are also reduced to the satisfaction of the demands of neighboring users.

[0360] Specifically, the solution provided in this embodiment of the invention may include the following steps:

[0361] Input: ① Number of base stations M, and the maximum transmit power of each base station. Geographic coordinates of each base station ② Number of active users N, maximum user traffic demand D max ③ Number of sub-bands K. ④ Total number of decision-making iterations T. ⑤ Total number of training iterations I. ⑥ Constraint violation penalty weight λ. m .

[0362] ⑦ Reinforcement learning discount factor γ.

[0363] Initialization: Initialize the demand decomposition strategy network, base station allocation strategy network, sub-band allocation strategy network, and power allocation strategy network using random parameters. Training iteration count i = 0.

[0364] Step 1: Reset the environment.

[0365] Step 1.1: Randomize the starting coordinates for each user and termination coordinates Based on the start and end coordinates and the total time T, assuming the user moves at a constant speed in a straight line, construct the movement speed and direction of each user.

[0366] Step 1.2: Randomize the traffic requirements of each user Traffic demand D n Follows a uniform distribution U(0,D) max ).

[0367] Step 1.3: Reset time t to 0.

[0368] Step 1.4: Set the trajectory set to

[0369] Step 2: Obtain the current coordinates of each user Path loss from each base station to each user is obtained based on user coordinates. (i.e., the signal attenuation factor from the base station to each user).

[0370] Step 3: Use the demand decomposition strategy network to assign an immediate demand to each user (corresponding to the traffic demand within the current time period mentioned above).

[0371] Step 3.1: For user u n A graph G1 was constructed by selecting neighboring base stations and users (including itself) to form a graph with base station and user information as nodes and their relationship as weights (corresponding to the first relationship graph constructed above based on the resource allocation parameters corresponding to the time period, with the first relevant information of the base station and the second relevant information of the terminal as nodes). Base station b m The corresponding node information includes: the one-hot code of the base station number (corresponding to the identity information of the aforementioned base station), and the base station transmit power limit P. m User u n The corresponding node information includes: unmet traffic demands. The remaining time (Tt) corresponds to the second information of the unmet traffic demand of the aforementioned terminal and the number of remaining time periods after the current time period of the terminal, respectively; the weight between the base station node and the user node is the signal attenuation factor. (Corresponding to the above in the first relationship diagram, the weights between the first relevant information and the second relevant information are determined based on the signal attenuation factor between the base station and the terminal), there are no edge connections between base station nodes and between user nodes.

[0372] Step 3.2: Embed graph G1 using a graph convolutional network to obtain the encodings of the base station and the user in the t-th time slice, denoted as follows: Here, the user's encoding at time slice t is obtained. The operation corresponds to the above-mentioned use of a graph neural network to obtain the first terminal code corresponding to the time period based on the first relationship graph.

[0373] In one loop, step 3.2 is executed at different time slices t, and the code for each time slice is saved. These saved codes constitute the historical sequence of user codes (corresponding to the above-mentioned coding history sequence). Specifically, each user code in the historical sequence is obtained by encoding information such as the user's location in the corresponding time slice.

[0374] Step 3.3: Historical sequence of user encoding The user code is obtained by further encoding using LSTM (Long Short-Term Memory). (Corresponding to the above terminal state change trend encoding), corresponding to the above use of recurrent neural networks to obtain the terminal state change trend encoding based on the encoding history sequence.

[0375] Step 3.4: User Code After passing through two fully connected layers, the output action probability parameters (the mean and standard deviation of a Gaussian probability distribution, corresponding to the first probability parameter information mentioned above) are obtained: Construct a Gaussian probability distribution N(μ) using this parameter. n ,σ n 2 And (can be randomly) sample actions As the result of the network's allocation strategy for this base station, the traffic that each user needs to satisfy in this time slice (the demand for time slice t) is denoted as... (The first information corresponding to the traffic demand of the terminal in the current time period). Corresponding to the above-mentioned use of a two-layer fully connected network, the first probability parameter information is obtained by encoding the terminal state change trend; the first Gaussian probability distribution is obtained based on the first probability parameter information; the traffic demand of the terminal in the current time period is determined by sampling from the first Gaussian probability distribution, thus obtaining the first information of the traffic demand of the terminal in the current time period.

[0376] Step 4: Based on the traffic demand calculated in Step 3 The network uses a base station allocation strategy to assign a base station to each user.

[0377] Step 4.1: For user u n A graph G2 is constructed by selecting neighboring base stations and users (including itself) with base station and user information as nodes and their relationship as weights (corresponding to the second relationship graph constructed above based on the resource allocation parameter information and the first information, with the third relevant information of the base station and the fourth relevant information of the terminal as nodes), where base station b m The corresponding node information includes: the one-hot code of the base station number (corresponding to the identity information of the aforementioned base station), and the base station transmit power limit P. m User u n The corresponding node information includes: the traffic requirements that the current time slice needs to meet. (i.e., the traffic demand calculated in step 3); the weight between base station nodes and user nodes is the signal attenuation factor. (Corresponding to the second relationship diagram above, the weights between the third relevant information and the fourth relevant information are determined based on the signal attenuation factor between the base station and the terminal), there are no edge connections between base station nodes and between user nodes.

[0378] Step 4.2: Embed graph G2 using a graph convolutional network to obtain the base station and user information encodings: Base station encoding User Code Corresponding to the above-described use of graph neural networks, the first base station code and the second terminal code are obtained based on the second relationship graph.

[0379] Step 4.3: After concatenating the base station code and user code pairwise (corresponding to concatenating the first base station code and the second terminal code to obtain the first concatenated code) and passing it through two layers of fully connected networks, user u is obtained. n Access base station b m The value estimate (which can be understood as a flow estimate, corresponding to the first flow estimate mentioned above): This value estimate corresponds to the immediate reward a user will receive after accessing the network, which is the traffic a user will receive within time slice t. This step corresponds to the above-mentioned use of a two-layer fully connected network to obtain the first traffic estimate of the terminal accessing the base station based on the first splicing encoding.

[0380] Step 4.4: Construct a bipartite graph consisting of two node sets. One node set corresponds to all users, and the other node set contains the number of nodes M times the number of base stations K times the number of sub-bands. Each K node set corresponds to one base station. The weights of the edges between the two node sets correspond to the value estimates calculated in Step 4.3.

[0381] Step 4.5: Use bipartite graph matching to construct a one-to-one matching relationship between nodes with the largest total weight, i.e. Indicates: User u n Base station b should be accessed in the current time slice t. m Otherwise, the connection will not be established.

[0382] Steps 4.4 and 4.5 correspond to the above steps of determining the target access base station of the terminal based on the first traffic estimate and obtaining the base station allocation information corresponding to the terminal.

[0383] Step 5: Based on the traffic demand calculated in Step 3 And the base station selection (i.e., allocation result) calculated in step 4. The network uses a sub-band allocation strategy to allocate sub-bands to each user.

[0384] Step 5.1: For user u nSelect its neighboring base stations and users (users include itself), and based on retaining graph G2, additionally construct graph G3 with base stations and users as nodes (corresponding to the third relationship graph constructed above based on the resource allocation parameter information and the first information, with the fifth relevant information of the base station and the sixth relevant information of the terminal as nodes). The node information is the same as in graph G2. If user u n Access base station b m If the weight between the corresponding nodes is 1, then the weight between the corresponding nodes is 0 (corresponding to the weight between the fifth and sixth related information in the third relationship diagram mentioned above, which is determined based on whether the base station and the terminal are connected). There are no edge connections between base station nodes and user nodes.

[0385] Step 5.2: Embed graphs G2 and G3 using a graph convolutional network to obtain the corresponding base station and user information codes, respectively. Then merge them (merge the two base station information codes together, and merge the two user information codes together) to obtain the final base station and user information codes. Among them, the final user information encoding is obtained. The operation corresponds to the above-mentioned use of a graph neural network to obtain a first sub-encoder based on the second relationship graph; use a graph neural network to obtain a second sub-encoder based on the third relationship graph; and concatenate the first sub-encoder and the second sub-encoder to obtain the third terminal code.

[0386] Step 5.3: The user code is processed through a two-layer fully connected network to obtain the value estimate of the user's access to different sub-frequency bands (which can be understood as a traffic estimate, corresponding to the second traffic estimate value mentioned above): This value estimate corresponds to the immediate reward a user will receive after accessing the network, which is the traffic a user will receive within time slice t. This step corresponds to the above-mentioned use of a two-layer fully connected network to obtain the second traffic estimate of the terminal access sub-band based on the third terminal coding.

[0387] Step 5.4: Construct M bipartite graphs, each consisting of two node sets. One node set corresponds to users accessing the base station, and the other node set represents the K selectable sub-frequency bands. Each K node set corresponds to one base station. The weights of the edges between the two node sets correspond to the value estimates calculated in Step 5.3.

[0388] Step 5.5: Use the bipartite graph matching method to construct a one-to-one matching relationship between nodes with the largest total weight for each bipartite graph, i.e. Indicates: User u n Base station b should be accessed in the current time slice t. m If the kth sub-band is selected, then the connection will be made; otherwise, the connection will not be made.

[0389] Steps 5.4 and 5.5 correspond to the above-mentioned determination of the target access sub-frequency band of the terminal based on the second traffic estimation value, and obtaining the sub-frequency band allocation information corresponding to the terminal.

[0390] Step 6: Based on the traffic demand calculated in Step 3 Step 4 calculates the base station selection (i.e., the allocation result). And the sub-band selection (i.e., allocation result) calculated in step 5. The network uses a power allocation strategy to allocate (transmit) power to each user.

[0391] Step 6.1: For user u n Select neighboring base stations and users (including itself) and, while retaining graphs G2 and G3, construct K additional graphs with base station and user information as nodes. (Corresponding to the above-mentioned fourth relationship graph constructed based on the resource allocation parameter information and the first information, with the seventh relevant information of the base station and the eighth relevant information of the terminal as nodes), one sub-frequency band corresponds to one graph, and the node information is the same as that of graph G2. If user u n Access base station b m The k-th sub-band, then Figure G 3+k The weight between corresponding nodes is 1, otherwise the weight between corresponding nodes is 0 (corresponding to the weight between the seventh and eighth related information in the fourth relationship diagram, which is determined based on whether the sub-frequency band of the base station is connected to the terminal). There are no edge connections between base station nodes and user nodes in each diagram.

[0392] Using graph convolutional networks for graphs G2 and G3 The information codes for the base stations and users are embedded separately to obtain their respective base station and user information codes. These are then merged (the three base station information codes are combined together, and the three user information codes are combined together) to obtain the final base station and user information codes. Among them, the final user information encoding is obtained. The operation (corresponding to the fourth terminal encoding above) corresponds to the above-mentioned use of graph neural network to obtain the third sub-encoding according to the second relationship graph; use graph neural network to obtain the fourth sub-encoding according to the third relationship graph; use graph neural network to obtain the fifth sub-encoding according to the fourth relationship graph; and concatenate the third sub-encoding, the fourth sub-encoding and the fifth sub-encoding to obtain the fourth terminal encoding.

[0393] Step 6.2: User Coding After passing through two fully connected layers, the output action probability parameters (the mean and standard deviation of the Gaussian probability distribution, corresponding to the second Gaussian probability distribution mentioned above) are obtained: Construct a Gaussian probability distribution using this parameter. And (can be randomly) sampled This is used as a sampled value for allocating transmit power (corresponding to the power reference value mentioned above). This step corresponds to obtaining the second Gaussian probability distribution based on the second probability parameter information; and sampling from the second Gaussian probability distribution to determine the power reference value corresponding to the terminal.

[0394] Step 6.3: For each base station b m Perform subsequent operations respectively, and record the number of users accessing the base station as I. m One, respectively The sampled values ​​corresponding to these users (i.e., the sampled values ​​obtained in step 6.2) are: These sampled values ​​and a constant C (e.g., 0) are passed through a softmax layer to obtain the proportion of transmit power allocated to these users (corresponding to the transmit power proportion of the terminal obtained based on the power reference value mentioned above); multiplying this proportion by the maximum transmit power of the corresponding base station yields the power allocated to these users. Corresponding to the above, based on the transmission power ratio and transmission power upper limit, the transmission power allocation information corresponding to the terminal is obtained. The power allocated to users under all base stations is calculated using the above method, and all results are denoted as follows.

[0395] Step 7: Calculate u for each user n The traffic obtained in the t-th time slice

[0396] Specifically, this may include: using the connection relationship between the user and the base station calculated in steps 3-6. User connection to sub-band and the transmit power allocated to the user According to the traffic calculation formula mentioned above:

[0397]

[0398] Calculate each user u n The traffic obtained in the t-th time slice Based on the resource allocation results information described above, the traffic value allocated to the corresponding terminal is obtained.

[0399] Further steps may include:

[0400] Step 8: Based on the flow rate obtained in Step 7 Training base station allocation strategy network, sub-band allocation strategy network, and power allocation strategy network; corresponding to the relevant networks that adjust resource allocation based on the traffic value.

[0401] Step 8.1: Calculate user u according to the methods in steps 4.1-4.4. n Access base station b m Value estimation If user u n Access base station b m Then calculate the value estimate and user u n Corresponding rewards (user u) n Single-time intra-chip flow Predicted loss between ) The total prediction loss is obtained by summing the prediction losses for all users. The derivative of this total prediction loss is then used to obtain the parameters of the network for the base station allocation strategy. The derivative is used to update the parameters. Corresponding to the above, the traffic loss value is obtained based on the traffic value, and the relevant network for resource allocation is adjusted based on the traffic loss value.

[0402] Among them, parameters This refers to the parameters of the base station allocation policy network; step 8.1 is to explain the parameter update process of the policy network. Here and below, the policy network refers to a deep neural network, and the parameters (e.g., Specifically, this can refer to the weights between nodes in a neural network. Since the loss is a function of the parameters, the direction of parameter updates can be obtained by differentiating the loss. Generally, the performance of the neural network improves as the parameters are updated.

[0403] Step 8.2: Calculate user u according to the methods in steps 5.1-5.4. n Value estimation for accessing different sub-bands If user u n Access base station b m For the k-th sub-band, calculate the value estimate and user u. n Corresponding rewards (user u) n Single-time intra-chip flow Predicted loss between ) The total prediction loss is obtained by summing the prediction losses for all users. The derivative of the total prediction loss is then used to obtain the parameters of the sub-band allocation strategy network. The derivative is used to update the parameters. Corresponding to the above, based on the traffic value, the traffic loss value corresponding to the resource allocation result information is obtained; based on the traffic loss value, the relevant network for resource allocation is adjusted. Wherein, parameters... This refers to the parameters of the sub-band allocation strategy network. The value estimate of a user accessing different sub-bands can be used to predict the traffic a user will receive when accessing different bands.

[0404] Step 8.3: This can be calculated using the methods described in steps 6.1-6.3. Then calculate the corresponding reward for the power allocation (traffic of all users). The constrained optimization problem corresponding to the power allocation strategy is:

[0405]

[0406] in, Indicates based on parameters The calculated transmit power for each user;

[0407] Transform it into an unconstrained optimization problem:

[0408]

[0409] Regarding the aforementioned losses Taking the derivative, we obtain the parameters of the power allocation strategy network. The derivative is used to update the parameters. Corresponding to the above, based on the traffic value, the traffic loss value corresponding to the resource allocation result information is obtained; based on the traffic loss value, the relevant network for resource allocation is adjusted. Wherein, parameters... The parameters of a power allocation strategy network.

[0410] The formula in this step represents the transformation of the constrained optimization problem represented by the first and second formulas into the unconstrained optimization problem represented by the third formula. Specifically:

[0411] The constrained optimization problem is expressed as follows (corresponding to the constrained adjustment conditions above): under the premise that the traffic of each user is greater than the user's immediate demand (corresponding to the second formula, that the traffic allocated to each terminal in the first time period is greater than the traffic demand in the first time period), maximize the total traffic obtained by all users in the time slice (corresponding to the first formula, that is, maximize the total traffic obtained by all terminals in the first time period).

[0412] The unconstrained optimization problem (corresponding to the unconstrained adjustment conditions above) is: maximizing the total traffic of all users in a single time slice. (corresponding to maximizing the total traffic of all terminals in the first time period as described above) and meeting the immediate needs of all users in a single time slice. The weighted sum (corresponding to the weighted sum that maximizes the traffic demand of all terminals in the first time period), with a weight of λ. m Because of λ m Since it is significantly greater than 1, this formula (i.e., the third formula) emphasizes increasing the traffic for unmet users in order to ensure that the constraints are met.

[0413] The content of the above "representation of constrained optimization problems" and the content of "representation of unconstrained optimization problems" correspond to obtaining the traffic loss value corresponding to the resource allocation result information according to the above first method based on the traffic value.

[0414] Step 9: Add a sample to the trajectory set Traj, where the sample includes: time slice t and the coordinates of all users Corresponding path loss Calculated traffic demand Actually obtained traffic

[0415] Step 10: t = t + 1. If t < T, jump to step 2; otherwise, jump to step 11.

[0416] Step 11: Train the demand decomposition policy network.

[0417] Step 11.1: Calculate the user encoding according to the method in steps 3.1 - 3.4 and the action probability parameters (μ n , σ n ).

[0418] Step 11.2: Pass the user encoding through a two - layer fully - connected network to obtain the corresponding state value estimate That is, the evaluation of the decision result. This value estimate corresponds to the cumulative reward that the user will obtain from time slice t to time slice T - 1, that is, the traffic that the user will obtain from time slice t to time slice T - 1. In this solution, use the reward of time slice t and the discounted value estimate of time slice t + 1 to calculate the target value Calculate the error loss between and Take the derivative of the loss to obtain the derivative of the parameters of the value network corresponding to the demand decomposition policy, and use this derivative to update the parameters The parameter

[0419] refers to the parameters of the value network that predicts "the total traffic obtained by the user within T time slices" (this value network is for assisting the training of the demand decomposition policy network). Specifically, if the obtained total traffic is greater than or equal to the traffic demand, it is marked as 1; otherwise, it is marked as 0.

[0420] Step 11.4: The constrained optimization problem corresponding to the demand decomposition policy is:

[0421]

[0421]

[0422] Where ρ represents the demand decomposition strategy The generated trajectory distribution includes information about the base station and users at each time step, as well as the output of the demand decomposition strategy at each time step. τ represents the trajectory sampled according to this distribution. E represents the expected total user traffic corresponding to trajectory τ under distribution ρ. Indicates user u n The set of IDs of neighboring users, n ′ Indicates user u n The IDs of neighboring users; Indicates user u n′ The flow rate D obtained in the t-th time slice n′ Indicates user u n′ Total traffic demand;

[0423] Transform it into an unconstrained optimization problem:

[0424]

[0425] in, For action gain:

[0426]

[0427] Taking the derivative with respect to the above objective yields the parameter derivatives of the value network corresponding to the demand decomposition strategy. These derivatives are then used to update the parameters. Corresponding to the above, based on the traffic value, the traffic loss value corresponding to the resource allocation result information is obtained; based on the traffic loss value, the relevant network for resource allocation is adjusted. Wherein, parameters... The parameters refer to the network of demand decomposition strategy.

[0428] The formulas mentioned above in this step represent the transformation of the constrained optimization problem represented by the first and second formulas into the unconstrained optimization problem represented by the third and fourth formulas.

[0429] The constrained optimization problem is: to maximize the total traffic obtained by a user and its neighboring users within T time slices (corresponding to the first formula) while ensuring that "for a user and its neighboring users, the sum of traffic obtained by each user within T time slices is greater than their respective overall demand (corresponding to the second formula)".

[0430] The unconstrained optimization problem is represented by the loss function corresponding to the policy network (corresponding to the third formula). The fourth formula represents the user u's immediate needs based on the output of this iteration. n Increase in traffic Weight value. The weighting method is as follows: when the total user demand is met, the corresponding mark is 1 and the actual weight is 1; otherwise, the mark is 0 and the actual weight is 1 + λ m . Since λ m is significantly greater than 1, this formula (i.e., the fourth formula) emphasizes the improvement of the traffic for unmet users to ensure meeting the constraint conditions. Among them, the actual weight refers to the coefficient [1 + (1 - mark n )λ m taking the actual value according to the different values of mark n .

[0431] Step 12: i = i + 1. If i < I, jump to Step 1; otherwise, jump to Step 13.

[0432] Step 13: Output the trained demand decomposition policy network, base station allocation policy network, sub - frequency band allocation policy network, and power allocation policy network.

[0433] Here it is noted that the above steps involve "for user u n , select its neighboring base stations and users" to construct (relationship) graphs (such as G1, G2, etc.), corresponding to the following: the base station information in the resource allocation parameter information corresponding to the said terminal is: the base station information of the first base station whose first distance from the terminal in the target area is less than the first threshold; and / or, the terminal information in the resource allocation parameter information corresponding to the said terminal is: the terminal information of the second terminal whose second distance from the terminal in the target area is less than the second threshold; and / or, the sub - frequency band information in the resource allocation parameter information corresponding to the said terminal is: the sub - frequency band information of the spectrum resources shared by the first base station; and / or, the upper limit value of the transmission power in the resource allocation parameter information corresponding to the said terminal is: the upper limit value of the transmission power of the first base station.

[0434] Furthermore, in order to verify the effectiveness of this solution, a certain wireless network was used as the problem scenario for testing; this scenario includes M = 184 base stations, N = 8000 users, K = 45 sub - frequency bands, and the signal attenuation factor from each base station to each location in this area can be obtained through the ray - tracing method.

[0435] First, test the performance of this solution in a single time slice. Randomly determine the user positions and user demands, and the user demands are the traffic demands corresponding to random numbers in [0, 2B], where B represents the bandwidth; the following are the experimental results:

[0436] Total throughput Demand fulfillment ratio Heuristics <![CDATA[1.28×10 4 ]]> 60.55％ This plan <![CDATA[1.01×10 4 ]]> 84.12％

[0437] Obviously, compared with the heuristic method, this solution can greatly improve the proportion of user demand satisfaction.

[0438] Then, the performance of this scheme was tested when the total time slice was T=20. The user's start and end positions were randomly determined, and it was assumed that the user was performing uniform linear motion. The user's demand was the bandwidth requirement corresponding to a random number in the range [0, 40B]. The relaxation factor was 0.95, meaning that 95% of the user's demand was to be met. The experimental results are as follows:

[0439] Total throughput Demand fulfillment ratio Unconstrained methods <![CDATA[5.18×10 5 ]]> 91.45％ This plan <![CDATA[4.49×10 5 ]]> 96.28％

[0440] Clearly, compared to unconstrained methods, this approach can find a resource distribution scheme that fits the relaxation factor, maximizing total throughput while meeting the needs of the vast majority of users, thereby optimizing the user experience.

[0441] Based on the above, this solution addresses the wireless resource allocation problem, which simultaneously optimizes base station selection, sub-frequency band selection, and power allocation under user traffic demand constraints. It presents a method for decomposing the problem into sub-problems and their individual solution methods. Furthermore, this solution provides a local optimization scheme for large-scale wireless network resource allocation problems. Specifically:

[0442] This solution addresses the wireless network resource allocation scenario where, for users within a specific area and time period, wireless network resources, including base stations, sub-frequency bands, and transmit power, are allocated based on their location and traffic demands. This scenario encompasses the allocation of various resources across different time periods, making it highly valuable for practical reference. More specifically, this solution provides a wireless resource allocation scheme, including a demand segmentation module (corresponding to the aforementioned demand decomposition strategy network), a base station selection module (corresponding to the aforementioned base station allocation strategy network), a sub-frequency band selection module (corresponding to the aforementioned sub-frequency band allocation strategy network), and a transmit power allocation module (corresponding to the aforementioned power allocation strategy network). This scheme solves the optimization problem by decomposing it and constructing a constraint optimization method, and introduces local observation and local rewards to optimize resource allocation for each user. While meeting user needs, it maximizes the system's total throughput, thereby optimizing the user experience.

[0443] In summary, this solution has the following advantages:

[0444] (1) This scheme can simultaneously optimize base station selection, sub-band selection, and power allocation, and construct a wireless network resource allocation scheme based on user traffic demand, thereby better meeting user traffic demand.

[0445] To meet users' traffic demands, this solution transforms the wireless resource allocation problem into a multi-step constraint-based timing optimization problem. This problem is further decomposed into several constrained sub-optimization problems: demand decomposition, base station allocation, sub-frequency band allocation, and power allocation. Compared to the original optimization problem, each sub-problem has relatively simple constraints, making it easier to solve. For these constrained sub-problems, this solution transforms them into unconstrained optimization problems for solution.

[0446] (2) This solution can solve the problem of large-scale wireless network resource allocation.

[0447] This scheme takes into account that the decision results related to each user will only affect a subset of neighboring users. Therefore, when allocating resources to each user, the scheme makes decisions based only on local information (including using only neighboring base stations and users to construct the graph and perform corresponding encoding for each decision mentioned above). This makes it possible to solve large-scale problems.

[0448] (3) This solution has high commercial value. For the problem of wireless network resource allocation, the usual approach is to transform it into a numerical optimization problem and then optimize it. However, the computational cost increases exponentially with the problem size, which means that such solutions are difficult to solve large-scale resource allocation problems in real time. Moreover, such solutions require the optimization objective to be concise and clear, which greatly limits the application scenarios. Compared with traditional solutions, the deep reinforcement learning applied in this solution can be flexible and diverse in the form of the optimization objective. In addition, deep networks trained with rich and diverse samples often have a certain generalization ability, are more adaptable to new scenarios, have lower costs, and the inference process of deep networks is less time-consuming, which can provide resource allocation solutions for large-scale problems in real time. Furthermore, in the process of commercial promotion, it has a stronger adaptability to new environments, greater flexibility, and stronger real-time performance.

[0449] This invention also provides a resource allocation device, such as... Figure 3 As shown, it includes:

[0450] The first acquisition module 31 is used to acquire resource allocation parameter information corresponding to at least one terminal in the target area;

[0451] The first processing module 32 is used to obtain first information on the traffic demand of each terminal in at least one time period based on the resource allocation parameter information.

[0452] The second processing module 33 is used to obtain resource allocation result information corresponding to each terminal based on the resource allocation parameter information and the first information;

[0453] The resource allocation parameter information includes at least one of the following: base station information corresponding to the target area, terminal information corresponding to the target area, sub-band information of spectrum resources shared by base stations in the target area, and upper limit value of transmit power for each base station.

[0454] The first information includes: information on the terminal's traffic demand during the current time period;

[0455] The resource allocation result information includes at least one of the following: base station allocation information, sub-frequency band allocation information, and transmit power allocation information.

[0456] The resource allocation device provided in this embodiment of the invention obtains resource allocation parameter information corresponding to at least one terminal in a target area; obtains first information on the traffic demand of each terminal in at least one time period based on the resource allocation parameter information; and obtains resource allocation result information corresponding to each terminal based on the resource allocation parameter information and the first information. The resource allocation parameter information includes at least one of the following: base station information corresponding to the target area, terminal information corresponding to the target area, sub-band information of spectrum resources shared by base stations in the target area, and upper limit values ​​of the transmit power of each base station. The first information includes: information on the traffic demand of the terminal in the current time period. The resource allocation result information includes at least one of: base station allocation information, sub-band allocation information, and transmit power allocation information. This device can effectively meet traffic demand by allocating resources according to traffic demand, and decomposes resource allocation into base station allocation, sub-band allocation, and transmit power allocation, thereby simplifying the resource allocation strategy, reducing the difficulty of the scheme, and effectively solving the problems of existing resource allocation schemes failing to effectively meet traffic demand and being complex.

[0457] The step of obtaining the first information on the traffic demand of each terminal within at least one time period based on the resource allocation parameter information includes: performing a first operation on at least one time period prior to the current time period of the terminal to obtain a corresponding first terminal code; obtaining an encoding history sequence based on the first terminal code; and obtaining the first information on the traffic demand of the terminal within the current time period based on the encoding history sequence. The first operation includes: constructing a corresponding first relationship graph based on the resource allocation parameter information corresponding to the time period; and obtaining the first terminal code corresponding to the time period based on the first relationship graph.

[0458] In this embodiment of the invention, constructing a corresponding first relationship graph based on the resource allocation parameter information corresponding to the time period includes: constructing a first relationship graph with first relevant information of the base station and second relevant information of the terminal as nodes based on the resource allocation parameter information corresponding to the time period; wherein, the first relevant information includes: the base station information and the upper limit of the transmission power, the base station information including: the identity information of the base station; the second relevant information includes: the terminal information, the terminal information including: second information on the unmet traffic demand of the terminal and the number of remaining time periods after the current time period of the terminal; in the first relationship graph, the weight between the first relevant information and the second relevant information is determined based on the signal attenuation factor between the base station and the terminal.

[0459] The step of obtaining the first information on the terminal's traffic demand in the current time period based on the encoded historical sequence includes: obtaining the terminal state change trend code based on the encoded historical sequence; and obtaining the first information on the terminal's traffic demand in the current time period based on the terminal state change trend code.

[0460] In this embodiment of the invention, obtaining the first terminal code corresponding to the time period based on the first relationship graph includes: using a graph neural network to obtain the first terminal code corresponding to the time period based on the first relationship graph; and / or, obtaining the terminal state change trend code based on the encoding history sequence includes: using a recurrent neural network to obtain the terminal state change trend code based on the encoding history sequence.

[0461] The step of obtaining the first information of the terminal's traffic demand in the current time period based on the terminal state change trend encoding includes: using a two-layer fully connected network to obtain first probability parameter information based on the terminal state change trend encoding; obtaining a first Gaussian probability distribution based on the first probability parameter information; and sampling from the first Gaussian probability distribution to determine the terminal's traffic demand in the current time period, thereby obtaining the first information of the terminal's traffic demand in the current time period.

[0462] In this embodiment of the invention, obtaining the resource allocation result information corresponding to each terminal based on the resource allocation parameter information and the first information includes: constructing a target relationship graph based on the resource allocation parameter information and the first information; obtaining a reference code based on the target relationship graph; and obtaining the resource allocation result information corresponding to the terminal based on the reference code.

[0463] The target relationship graph includes a second relationship graph; constructing the target relationship graph based on the resource allocation parameter information and the first information includes: constructing a second relationship graph with the third relevant information of the base station and the fourth relevant information of the terminal as nodes based on the resource allocation parameter information and the first information; and / or, the reference coding includes: a first base station coding and a second terminal coding; obtaining the reference coding based on the target relationship graph includes: using a graph neural network to obtain the first base station coding and the second terminal coding based on the second relationship graph; and / or, obtaining the resource allocation result information corresponding to the terminal based on the reference coding includes: concatenating the first base station coding and the second terminal coding to obtain a first concatenated coding; using a two-layer fully connected network to obtain a first traffic estimate of the terminal accessing the base station based on the first concatenated coding; determining the target access base station of the terminal based on the first traffic estimate to obtain the base station allocation information corresponding to the terminal; wherein, the third relevant information includes: the base station information and the upper limit of the transmit power, the base station information includes: the identity information of the base station; the fourth relevant information includes: the first information; in the second relationship graph, the weight between the third relevant information and the fourth relevant information is determined based on the signal attenuation factor between the base station and the terminal.

[0464] In this embodiment of the invention, the target relationship graph includes a third relationship graph; constructing the target relationship graph based on the resource allocation parameter information and the first information includes: constructing a third relationship graph with the fifth relevant information of the base station and the sixth relevant information of the terminal as nodes based on the resource allocation parameter information and the first information; and / or, the reference coding includes: a third terminal coding; obtaining the reference coding based on the target relationship graph includes: using a graph neural network to obtain a first sub-code based on the second relationship graph; using a graph neural network to obtain a second sub-code based on the third relationship graph; concatenating the first sub-code and the second sub-code to obtain the third terminal coding; and / or, the According to the reference coding, the resource allocation result information corresponding to the terminal is obtained, including: using a two-layer fully connected network, according to the third terminal coding, to obtain a second traffic estimate of the terminal access sub-frequency band; according to the second traffic estimate, to determine the target access sub-frequency band of the terminal, and to obtain the sub-frequency band allocation information corresponding to the terminal; wherein, the fifth related information includes: the base station information and the upper limit of the transmit power, the base station information includes: the identity information of the base station; the sixth related information includes: the first information; in the third relationship diagram, the weight between the fifth related information and the sixth related information is determined according to whether there is a connection between the base station and the terminal.

[0465] The target relationship graph includes a fourth relationship graph; constructing the target relationship graph based on the resource allocation parameter information and the first information includes: constructing a fourth relationship graph with the seventh relevant information of the base station and the eighth relevant information of the terminal as nodes; and / or, the reference coding includes: a fourth terminal coding; obtaining the reference coding based on the target relationship graph includes: using a graph neural network to obtain a third sub-code based on the second relationship graph; using a graph neural network to obtain a fourth sub-code based on the third relationship graph; using a graph neural network to obtain a fifth sub-code based on the fourth relationship graph; concatenating the third sub-code, the fourth sub-code, and the fifth sub-code to obtain the fourth terminal coding. Terminal coding; and / or, obtaining the resource allocation result information corresponding to the terminal based on the reference coding includes: using a two-layer fully connected network, obtaining second probability parameter information based on the fourth terminal coding; obtaining a second Gaussian probability distribution based on the second probability parameter information; obtaining transmit power allocation information corresponding to the terminal based on the second Gaussian probability distribution; wherein, the seventh related information includes: the base station information and the transmit power upper limit value, the base station information includes: the base station identification information; the eighth related information includes: the first information; in the fourth relationship diagram, the weight between the seventh related information and the eighth related information is determined based on whether the sub-frequency band of the base station is connected to the terminal.

[0466] In this embodiment of the invention, obtaining the transmission power allocation information corresponding to the terminal based on the second Gaussian probability distribution includes: sampling and determining the power reference value corresponding to the terminal from the second Gaussian probability distribution; obtaining the transmission power ratio corresponding to the terminal based on the power reference value; and obtaining the transmission power allocation information corresponding to the terminal based on the transmission power ratio and the upper limit value of the transmission power.

[0467] Furthermore, the resource allocation device further includes: a third processing module, used to obtain the traffic value allocated to the corresponding terminal based on the resource allocation result information; and a first adjustment module, used to adjust the relevant network for resource allocation based on the traffic value; wherein the relevant network includes at least one of the following: a graph neural network; a recurrent neural network; or a two-layer fully connected network.

[0468] The step of adjusting the relevant network for resource allocation based on the traffic value includes: obtaining the traffic loss value corresponding to the resource allocation result information based on the traffic value; and adjusting the relevant network for resource allocation based on the traffic loss value.

[0469] In this embodiment of the invention, obtaining the traffic loss value corresponding to the resource allocation result information based on the traffic value includes: using a first method to obtain the traffic loss value corresponding to the resource allocation result information based on the traffic value; wherein, the first method includes: converting a constrained adjustment condition into an unconstrained adjustment condition; the constrained adjustment condition includes: ensuring that the traffic allocated to each terminal in the first time period is greater than the traffic demand in the first time period, and maximizing the total traffic obtained by all terminals in the first time period; the unconstrained adjustment condition includes: maximizing the total traffic of all terminals in the first time period, and maximizing the weighted sum of the traffic demand satisfied by all terminals in the first time period.

[0470] In this embodiment of the invention, the base station information in the resource allocation parameter information corresponding to the terminal is: the base station information of a first base station in the target area whose first distance from the terminal is less than a first threshold; and / or, the terminal information in the resource allocation parameter information corresponding to the terminal is: the terminal information of a second terminal in the target area whose second distance from the terminal is less than a second threshold; and / or, the sub-frequency band information in the resource allocation parameter information corresponding to the terminal is: the sub-frequency band information of the spectrum resources shared by the first base station; and / or, the upper limit value of the transmit power in the resource allocation parameter information corresponding to the terminal is: the upper limit value of the transmit power of the first base station.

[0471] The implementation embodiments of the above-described resource allocation method are all applicable to the embodiments of the resource allocation device, and can achieve the same technical effect.

[0472] This invention also provides a resource allocation device, such as... Figure 4 As shown, it includes: processor 41;

[0473] The processor 41 is used to obtain resource allocation parameter information corresponding to at least one terminal in the target area;

[0474] Based on the resource allocation parameter information, the first information on the traffic demand of each terminal within at least one time period is obtained;

[0475] Based on the resource allocation parameter information and the first information, the resource allocation result information corresponding to each terminal is obtained;

[0476] The resource allocation parameter information includes at least one of the following: base station information corresponding to the target area, terminal information corresponding to the target area, sub-band information of spectrum resources shared by base stations in the target area, and upper limit value of transmit power for each base station.

[0477] The first information includes: information on the terminal's traffic demand during the current time period;

[0478] The resource allocation result information includes at least one of the following: base station allocation information, sub-frequency band allocation information, and transmit power allocation information.

[0479] In this embodiment of the invention, the resource allocation device may further include a transceiver 42 capable of communicating with the processor 41, but is not limited thereto.

[0480] The resource allocation device provided in this embodiment of the invention obtains resource allocation parameter information corresponding to at least one terminal in a target area; based on the resource allocation parameter information, it obtains first information on the traffic demand of each terminal in at least one time period; based on the resource allocation parameter information and the first information, it obtains resource allocation result information corresponding to each terminal; wherein, the resource allocation parameter information includes at least one of: base station information corresponding to the target area, terminal information corresponding to the target area, sub-frequency band information of spectrum resources shared by base stations in the target area, and upper limit value of transmit power of each base station; the first information includes: information on the traffic demand of the terminal in the current time period; the resource allocation result information includes at least one of: base station allocation information, sub-frequency band allocation information, and transmit power allocation information; it can realize resource allocation according to traffic demand to effectively meet traffic demand, and decomposes resource allocation into base station allocation, sub-frequency band allocation, and transmit power allocation, thereby simplifying the resource allocation strategy, reducing the difficulty of the scheme, and effectively solving the problems of existing resource allocation schemes being unable to effectively meet traffic demand and the schemes being complex.

[0481] The step of obtaining the first information on the traffic demand of each terminal within at least one time period based on the resource allocation parameter information includes: performing a first operation on at least one time period prior to the current time period of the terminal to obtain a corresponding first terminal code; obtaining an encoding history sequence based on the first terminal code; and obtaining the first information on the traffic demand of the terminal within the current time period based on the encoding history sequence. The first operation includes: constructing a corresponding first relationship graph based on the resource allocation parameter information corresponding to the time period; and obtaining the first terminal code corresponding to the time period based on the first relationship graph.

[0482] In this embodiment of the invention, constructing a corresponding first relationship graph based on the resource allocation parameter information corresponding to the time period includes: constructing a first relationship graph with first relevant information of the base station and second relevant information of the terminal as nodes based on the resource allocation parameter information corresponding to the time period; wherein, the first relevant information includes: the base station information and the upper limit of the transmission power, the base station information including: the identity information of the base station; the second relevant information includes: the terminal information, the terminal information including: second information on the unmet traffic demand of the terminal and the number of remaining time periods after the current time period of the terminal; in the first relationship graph, the weight between the first relevant information and the second relevant information is determined based on the signal attenuation factor between the base station and the terminal.

[0483] The step of obtaining the first information on the terminal's traffic demand in the current time period based on the encoded historical sequence includes: obtaining the terminal state change trend code based on the encoded historical sequence; and obtaining the first information on the terminal's traffic demand in the current time period based on the terminal state change trend code.

[0484] In this embodiment of the invention, obtaining the first terminal code corresponding to the time period based on the first relationship graph includes: using a graph neural network to obtain the first terminal code corresponding to the time period based on the first relationship graph; and / or, obtaining the terminal state change trend code based on the encoding history sequence includes: using a recurrent neural network to obtain the terminal state change trend code based on the encoding history sequence.

[0485] The step of obtaining the first information of the terminal's traffic demand in the current time period based on the terminal state change trend encoding includes: using a two-layer fully connected network to obtain first probability parameter information based on the terminal state change trend encoding; obtaining a first Gaussian probability distribution based on the first probability parameter information; and sampling from the first Gaussian probability distribution to determine the terminal's traffic demand in the current time period, thereby obtaining the first information of the terminal's traffic demand in the current time period.

[0486] In this embodiment of the invention, obtaining the resource allocation result information corresponding to each terminal based on the resource allocation parameter information and the first information includes: constructing a target relationship graph based on the resource allocation parameter information and the first information; obtaining a reference code based on the target relationship graph; and obtaining the resource allocation result information corresponding to the terminal based on the reference code.

[0487] The target relationship graph includes a second relationship graph; constructing the target relationship graph based on the resource allocation parameter information and the first information includes: constructing a second relationship graph with the third relevant information of the base station and the fourth relevant information of the terminal as nodes based on the resource allocation parameter information and the first information; and / or, the reference coding includes: a first base station coding and a second terminal coding; obtaining the reference coding based on the target relationship graph includes: using a graph neural network to obtain the first base station coding and the second terminal coding based on the second relationship graph; and / or, obtaining the resource allocation result information corresponding to the terminal based on the reference coding includes: concatenating the first base station coding and the second terminal coding to obtain a first concatenated coding; using a two-layer fully connected network to obtain a first traffic estimate of the terminal accessing the base station based on the first concatenated coding; determining the target access base station of the terminal based on the first traffic estimate to obtain the base station allocation information corresponding to the terminal; wherein, the third relevant information includes: the base station information and the upper limit of the transmit power, the base station information includes: the identity information of the base station; the fourth relevant information includes: the first information; in the second relationship graph, the weight between the third relevant information and the fourth relevant information is determined based on the signal attenuation factor between the base station and the terminal.

[0488] In this embodiment of the invention, the target relationship graph includes a third relationship graph; constructing the target relationship graph based on the resource allocation parameter information and the first information includes: constructing a third relationship graph with the fifth relevant information of the base station and the sixth relevant information of the terminal as nodes based on the resource allocation parameter information and the first information; and / or, the reference coding includes: a third terminal coding; obtaining the reference coding based on the target relationship graph includes: using a graph neural network to obtain a first sub-code based on the second relationship graph; using a graph neural network to obtain a second sub-code based on the third relationship graph; concatenating the first sub-code and the second sub-code to obtain the third terminal coding; and / or, the According to the reference coding, the resource allocation result information corresponding to the terminal is obtained, including: using a two-layer fully connected network, according to the third terminal coding, to obtain a second traffic estimate of the terminal access sub-frequency band; according to the second traffic estimate, to determine the target access sub-frequency band of the terminal, and to obtain the sub-frequency band allocation information corresponding to the terminal; wherein, the fifth related information includes: the base station information and the upper limit of the transmit power, the base station information includes: the identity information of the base station; the sixth related information includes: the first information; in the third relationship diagram, the weight between the fifth related information and the sixth related information is determined according to whether there is a connection between the base station and the terminal.

[0489] The target relationship graph includes a fourth relationship graph; constructing the target relationship graph based on the resource allocation parameter information and the first information includes: constructing a fourth relationship graph with the seventh relevant information of the base station and the eighth relevant information of the terminal as nodes; and / or, the reference coding includes: a fourth terminal coding; obtaining the reference coding based on the target relationship graph includes: using a graph neural network to obtain a third sub-code based on the second relationship graph; using a graph neural network to obtain a fourth sub-code based on the third relationship graph; using a graph neural network to obtain a fifth sub-code based on the fourth relationship graph; concatenating the third sub-code, the fourth sub-code, and the fifth sub-code to obtain the fourth terminal coding. Terminal coding; and / or, obtaining the resource allocation result information corresponding to the terminal based on the reference coding includes: using a two-layer fully connected network, obtaining second probability parameter information based on the fourth terminal coding; obtaining a second Gaussian probability distribution based on the second probability parameter information; obtaining transmit power allocation information corresponding to the terminal based on the second Gaussian probability distribution; wherein, the seventh related information includes: the base station information and the transmit power upper limit value, the base station information includes: the base station identification information; the eighth related information includes: the first information; in the fourth relationship diagram, the weight between the seventh related information and the eighth related information is determined based on whether the sub-frequency band of the base station is connected to the terminal.

[0490] In this embodiment of the invention, obtaining the transmission power allocation information corresponding to the terminal based on the second Gaussian probability distribution includes: sampling and determining the power reference value corresponding to the terminal from the second Gaussian probability distribution; obtaining the transmission power ratio corresponding to the terminal based on the power reference value; and obtaining the transmission power allocation information corresponding to the terminal based on the transmission power ratio and the upper limit value of the transmission power.

[0491] Furthermore, the processor is also configured to: obtain the traffic value allocated to the corresponding terminal based on the resource allocation result information; adjust the relevant network for resource allocation based on the traffic value; wherein the relevant network includes at least one of the following: graph neural network; recurrent neural network; two-layer fully connected network.

[0492] The step of adjusting the relevant network for resource allocation based on the traffic value includes: obtaining the traffic loss value corresponding to the resource allocation result information based on the traffic value; and adjusting the relevant network for resource allocation based on the traffic loss value.

[0493] In this embodiment of the invention, obtaining the traffic loss value corresponding to the resource allocation result information based on the traffic value includes: using a first method to obtain the traffic loss value corresponding to the resource allocation result information based on the traffic value; wherein, the first method includes: converting a constrained adjustment condition into an unconstrained adjustment condition; the constrained adjustment condition includes: ensuring that the traffic allocated to each terminal in the first time period is greater than the traffic demand in the first time period, and maximizing the total traffic obtained by all terminals in the first time period; the unconstrained adjustment condition includes: maximizing the total traffic of all terminals in the first time period, and maximizing the weighted sum of the traffic demand satisfied by all terminals in the first time period.

[0494] In this embodiment of the invention, the base station information in the resource allocation parameter information corresponding to the terminal is: the base station information of a first base station in the target area whose first distance from the terminal is less than a first threshold; and / or, the terminal information in the resource allocation parameter information corresponding to the terminal is: the terminal information of a second terminal in the target area whose second distance from the terminal is less than a second threshold; and / or, the sub-frequency band information in the resource allocation parameter information corresponding to the terminal is: the sub-frequency band information of the spectrum resources shared by the first base station; and / or, the upper limit value of the transmit power in the resource allocation parameter information corresponding to the terminal is: the upper limit value of the transmit power of the first base station.

[0495] The implementation embodiments of the above resource allocation method are all applicable to the embodiments of the resource allocation device and can achieve the same technical effect.

[0496] This invention also provides a resource allocation device, including a memory, a processor, and a program stored in the memory and executable on the processor; when the processor executes the program, it implements the resource allocation method described above.

[0497] The implementation embodiments of the above-described resource allocation method are all applicable to the embodiments of the resource allocation device and can achieve the same technical effect.

[0498] This invention also provides a readable storage medium storing a program that, when executed by a processor, implements the steps in the resource allocation method described above.

[0499] The implementation embodiments of the above resource allocation method are all applicable to the embodiments of the readable storage medium and can achieve the same technical effect.

[0500] It should be noted that many of the functional components described in this specification are referred to as modules in order to more specifically emphasize the independence of their implementation.

[0501] In this embodiment of the invention, the module can be implemented in software so that it can be executed by various types of processors. For example, an identified executable code module may include one or more physical or logical blocks of computer instructions, which may be constructed as objects, procedures, or functions. Nevertheless, the executable code of the identified module does not need to be physically located together, but may include different instructions stored in different bits, which, when logically combined, constitute the module and achieve the module's intended purpose.

[0502] In practice, an executable code module can be a single instruction or many instructions, and can even be distributed across multiple different code segments, different programs, and across multiple memory devices. Similarly, operational data can be identified within the module and can be implemented in any suitable form and organized within any suitable type of data structure. This operational data can be collected as a single dataset or distributed across different locations (including different storage devices), and can exist, at least in part, solely as electronic signals within the system or network.

[0503] When a module can be implemented using software, considering the current level of hardware technology, modules that can be implemented in software can be implemented using hardware circuits by those skilled in the art to achieve the corresponding functions, without considering cost. These hardware circuits include conventional very-large-scale integrated circuits (VLSI) or gate arrays, as well as existing semiconductors such as logic chips and transistors, or other discrete components. Modules can also be implemented using programmable hardware devices, such as field-programmable gate arrays, programmable array logic, and programmable logic devices.

[0504] The above describes the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A resource allocation method, characterized in that, include: Obtain resource allocation parameter information corresponding to at least one terminal within the target area; Based on the resource allocation parameter information, the first information on the traffic demand of each terminal within at least one time period is obtained; Based on the resource allocation parameter information and the first information, the resource allocation result information corresponding to each terminal is obtained; The resource allocation parameter information includes at least one of the following: base station information corresponding to the target area, terminal information corresponding to the target area, sub-band information of spectrum resources shared by base stations in the target area, and upper limit value of transmit power for each base station. The first information includes: information on the terminal's traffic demand during the current time period; The resource allocation result information includes at least one of the following: base station allocation information, sub-frequency band allocation information, and transmit power allocation information; The step of obtaining first information on the traffic demand of each terminal within at least one time period based on the resource allocation parameter information includes: For at least one time period prior to the current time period of the terminal, a first operation is performed to obtain the corresponding first terminal codes; Based on the first terminal encoding, the encoding history sequence is obtained; Based on the encoded history sequence, the first information of the terminal's traffic demand in the current time period is obtained; The first operation includes: constructing a corresponding first relationship graph based on the resource allocation parameter information corresponding to the time period; and obtaining a first terminal code corresponding to the time period based on the first relationship graph. The step of obtaining the resource allocation result information corresponding to each terminal based on the resource allocation parameter information and the first information includes: Based on the resource allocation parameter information and the first information, construct the target relationship graph; Based on the target relationship diagram, the reference code is obtained; Based on the reference code, the resource allocation result information corresponding to the terminal is obtained.

2. The resource allocation method according to claim 1, characterized in that, The step of constructing a corresponding first relationship graph based on the resource allocation parameter information corresponding to the time period includes: Based on the resource allocation parameter information corresponding to the time period, a first relationship graph is constructed with the first relevant information of the base station and the second relevant information of the terminal as nodes. The first relevant information includes: the base station information and the upper limit of the transmission power, wherein the base station information includes: the identity information of the base station; The second relevant information includes: the terminal information, which includes: second information on the unmet traffic demand of the terminal and the number of remaining time periods after the current time period of the terminal; In the first relationship diagram, the weights between the first relevant information and the second relevant information are determined based on the signal attenuation factor between the base station and the terminal.

3. The resource allocation method according to claim 1, characterized in that, The step of obtaining the first information on the terminal's traffic demand in the current time period based on the encoded historical sequence includes: Based on the encoded history sequence, the terminal state change trend code is obtained; Based on the terminal status change trend encoding, the first information of the terminal's traffic demand in the current time period is obtained.

4. The resource allocation method according to claim 3, characterized in that, The step of obtaining the first terminal code corresponding to the time period based on the first relationship graph includes: Using a graph neural network, the first terminal code corresponding to the time period is obtained based on the first relationship graph; And / or, obtaining the terminal state change trend code based on the encoded history sequence includes: Using a recurrent neural network, the terminal state change trend code is obtained based on the encoded history sequence.

5. The resource allocation method according to claim 3, characterized in that, The step of encoding the terminal's state change trend to obtain the first information on the terminal's traffic demand in the current time period includes: Using a two-layer fully connected network, the first probability parameter information is obtained by encoding the terminal state change trend. Based on the first probability parameter information, the first Gaussian probability distribution is obtained; From the first Gaussian probability distribution, the traffic demand of the terminal in the current time period is determined by sampling, and the first information of the traffic demand of the terminal in the current time period is obtained.

6. The resource allocation method according to claim 1, characterized in that, The target relationship graph includes a second relationship graph; The step of constructing a target relationship graph based on the resource allocation parameter information and the first information includes: Based on the resource allocation parameter information and the first information, a second relationship graph is constructed with the third relevant information of the base station and the fourth relevant information of the terminal as nodes; And / or, the reference encoding includes: a first base station encoding and a second terminal encoding; The step of obtaining the reference code based on the target relationship diagram includes: Using a graph neural network, the first base station code and the second terminal code are obtained based on the second relationship graph; And / or, obtaining the resource allocation result information corresponding to the terminal based on the reference encoding includes: The first base station code and the second terminal code are concatenated to obtain the first concatenated code; Using a two-layer fully connected network, the first traffic estimate of the terminal accessing the base station is obtained based on the first splicing code; Based on the first traffic estimate, the target access base station of the terminal is determined, and the base station allocation information corresponding to the terminal is obtained; The third relevant information includes: the base station information and the upper limit of the transmission power, wherein the base station information includes: the identity information of the base station; The fourth relevant information includes: the first information; In the second relationship diagram, the weights between the third relevant information and the fourth relevant information are determined based on the signal attenuation factor between the base station and the terminal.

7. The resource allocation method according to claim 6, characterized in that, The target relationship graph includes a third relationship graph; The step of constructing a target relationship graph based on the resource allocation parameter information and the first information includes: Based on the resource allocation parameter information and the first information, a third relationship graph is constructed with the fifth relevant information of the base station and the sixth relevant information of the terminal as nodes; And / or, the reference encoding includes: a third terminal encoding; The step of obtaining the reference code based on the target relationship diagram includes: Using a graph neural network, the first sub-encoder is obtained based on the second relationship graph; Using a graph neural network, the second sub-encoder is obtained based on the third relationship graph; The first sub-encode and the second sub-encode are concatenated to obtain the third terminal code; And / or, obtaining the resource allocation result information corresponding to the terminal based on the reference encoding includes: Using a two-layer fully connected network, the second traffic estimate for the terminal access sub-band is obtained based on the third terminal coding; Based on the second traffic estimate, the target access sub-band of the terminal is determined, and the sub-band allocation information corresponding to the terminal is obtained; The fifth relevant information includes: the base station information and the upper limit of the transmission power, wherein the base station information includes: the identity information of the base station; The sixth relevant information includes: the first information; In the third relationship diagram, the weights between the fifth and sixth related information are determined based on whether there is a connection between the base station and the terminal.

8. The resource allocation method according to claim 7, characterized in that, The target relationship diagram includes a fourth relationship diagram; The step of constructing a target relationship graph based on the resource allocation parameter information and the first information includes: Based on the resource allocation parameter information and the first information, a fourth relationship graph is constructed with the seventh relevant information of the base station and the eighth relevant information of the terminal as nodes; And / or, the reference encoding includes: a fourth terminal encoding; The step of obtaining the reference code based on the target relationship diagram includes: Using a graph neural network, the third sub-encoder is obtained based on the second relationship graph; Using a graph neural network, the fourth sub-encoder is obtained based on the third relation graph; Using a graph neural network, the fifth sub-encoder is obtained based on the fourth relation graph; The third sub-encode, the fourth sub-encode, and the fifth sub-encode are concatenated to obtain the fourth terminal code; And / or, obtaining the resource allocation result information corresponding to the terminal based on the reference encoding includes: Using a two-layer fully connected network, the second probability parameter information is obtained based on the fourth terminal encoding; Based on the second probability parameter information, the second Gaussian probability distribution is obtained; Based on the second Gaussian probability distribution, the transmit power allocation information corresponding to the terminal is obtained; The seventh relevant information includes: the base station information and the upper limit of the transmission power, wherein the base station information includes: the identity information of the base station; The eighth relevant information includes: the first information; In the fourth relationship diagram, the weights between the seventh and eighth related information are determined based on whether there is a connection between the sub-frequency band of the base station and the terminal.

9. The resource allocation method according to claim 8, characterized in that, The step of obtaining the transmit power allocation information corresponding to the terminal based on the second Gaussian probability distribution includes: The power reference value corresponding to the terminal is determined by sampling from the second Gaussian probability distribution; Based on the power reference value, the corresponding transmit power percentage of the terminal is obtained; Based on the transmission power ratio and the upper limit of transmission power, the transmission power allocation information corresponding to the terminal is obtained.

10. The resource allocation method according to any one of claims 1 to 9, characterized in that, Also includes: Based on the resource allocation result information, the traffic value allocated to the corresponding terminal is obtained; Adjust the relevant network for resource allocation based on the traffic value; The relevant network includes at least one of the following: Graph neural networks; Recurrent neural networks; Two-layer fully connected network.

11. The resource allocation method according to claim 10, characterized in that, The process of adjusting resource allocation based on the traffic value includes: Based on the traffic value, obtain the traffic loss value corresponding to the resource allocation result information; Based on the aforementioned traffic loss value, adjust the relevant network for resource allocation.

12. The resource allocation method according to claim 11, characterized in that, The step of obtaining the traffic loss value corresponding to the resource allocation result information based on the traffic value includes: Using the first method, based on the traffic value, obtain the traffic loss value corresponding to the resource allocation result information; The first method includes: converting constrained adjustment conditions into unconstrained adjustment conditions; The constrained adjustment conditions include: ensuring that the traffic allocated to each terminal in the first time period is greater than the traffic demand in the first time period, and maximizing the total traffic obtained by all terminals in the first time period; The unconstrained adjustment conditions include: maximizing the total traffic of all terminals in the first time period, and maximizing the weighted sum of the traffic requirements met by all terminals in the first time period.

13. The resource allocation method according to claim 1, characterized in that, The base station information in the resource allocation parameter information corresponding to the terminal is: the base station information of the first base station in the target area whose first distance from the terminal is less than a first threshold; And / or, the terminal information in the resource allocation parameter information corresponding to the terminal is: the terminal information of a second terminal in the target area whose second distance from the terminal is less than a second threshold; And / or, the sub-frequency band information in the resource allocation parameter information corresponding to the terminal is: the sub-frequency band information of the spectrum resources shared by the first base station; And / or, the upper limit of the transmit power in the resource allocation parameter information corresponding to the terminal is: the upper limit of the transmit power of the first base station.

14. A resource allocation device, characterized in that, include: The first acquisition module is used to acquire resource allocation parameter information corresponding to at least one terminal within the target area; The first processing module is used to obtain first information on the traffic demand of each terminal in at least one time period based on the resource allocation parameter information. The second processing module is used to obtain resource allocation result information corresponding to each terminal based on the resource allocation parameter information and the first information; The resource allocation parameter information includes at least one of the following: base station information corresponding to the target area, terminal information corresponding to the target area, sub-band information of spectrum resources shared by base stations in the target area, and upper limit value of transmit power for each base station. The first information includes: information on the terminal's traffic demand during the current time period; The resource allocation result information includes at least one of the following: base station allocation information, sub-frequency band allocation information, and transmit power allocation information; The step of obtaining first information on the traffic demand of each terminal within at least one time period based on the resource allocation parameter information includes: For at least one time period prior to the current time period of the terminal, a first operation is performed to obtain the corresponding first terminal codes; Based on the first terminal encoding, the encoding history sequence is obtained; Based on the encoded history sequence, the first information of the terminal's traffic demand in the current time period is obtained; The first operation includes: constructing a corresponding first relationship graph based on the resource allocation parameter information corresponding to the time period; and obtaining a first terminal code corresponding to the time period based on the first relationship graph. The step of obtaining the resource allocation result information corresponding to each terminal based on the resource allocation parameter information and the first information includes: Based on the resource allocation parameter information and the first information, construct the target relationship graph; Based on the target relationship diagram, the reference code is obtained; Based on the reference code, the resource allocation result information corresponding to the terminal is obtained.

15. A resource allocation device, characterized in that, include: processor; The processor is used to acquire resource allocation parameter information corresponding to at least one terminal within the target area; Based on the resource allocation parameter information, the first information on the traffic demand of each terminal within at least one time period is obtained; Based on the resource allocation parameter information and the first information, the resource allocation result information corresponding to each terminal is obtained; The resource allocation parameter information includes at least one of the following: base station information corresponding to the target area, terminal information corresponding to the target area, sub-band information of spectrum resources shared by base stations in the target area, and upper limit value of transmit power for each base station. The first information includes: information on the terminal's traffic demand during the current time period; The resource allocation result information includes at least one of the following: base station allocation information, sub-frequency band allocation information, and transmit power allocation information; The step of obtaining first information on the traffic demand of each terminal within at least one time period based on the resource allocation parameter information includes: For at least one time period prior to the current time period of the terminal, a first operation is performed to obtain the corresponding first terminal codes; Based on the first terminal encoding, the encoding history sequence is obtained; Based on the encoded history sequence, the first information of the terminal's traffic demand in the current time period is obtained; The first operation includes: constructing a corresponding first relationship graph based on the resource allocation parameter information corresponding to the time period; and obtaining a first terminal code corresponding to the time period based on the first relationship graph. The step of obtaining the resource allocation result information corresponding to each terminal based on the resource allocation parameter information and the first information includes: Based on the resource allocation parameter information and the first information, construct the target relationship graph; Based on the target relationship diagram, the reference code is obtained; Based on the reference code, the resource allocation result information corresponding to the terminal is obtained.

16. A resource allocation device, comprising a memory, a processor, and a program stored in the memory and executable on the processor; characterized in that, When the processor executes the program, it implements the resource allocation method as described in any one of claims 1 to 13.

17. A readable storage medium having a program stored thereon, characterized in that, When the program is executed by the processor, it implements the steps in the resource allocation method as described in any one of claims 1 to 13.

Images