Multi-base-station cooperative sensing resource optimization method and device based on generative AI

Abstract

The invention provides a multi-base-station cooperative sensing resource optimization method and device based on a generative AI. The multi-base-station cooperative sensing resource optimization method based on the generative AI comprises the steps that first communication sensing related information of a first communication sensing integrated ISAC system of multi-base-station cooperation is acquired; and inputting the first communication perception related information into a trained artificial intelligence (AI) resource optimization model to obtain first resource allocation information of a first multi-base-station cooperative ISAC system and / or a first perception mode of the first multi-base-station cooperative ISAC system output by the AI resource optimization model. The method breaks through the limitation of traditional fixed resource allocation, can be generalized to a new scene demand according to the existing scene information, dynamically adjusts a sensing mode and a resource allocation strategy, has certain robustness, and is high in robustness. Moreover, the method can adapt to dynamic environment changes, and solves the problems of low resource utilization efficiency and limited communication perception performance caused by insufficient flexibility of the system.

Description

Technical Field

[0001] The present invention relates to the field of communication perception integration technology, and in particular to a multi-base station collaborative perception resource optimization method based on generative AI. Background Art

[0002] In future mobile communication networks, Integrated Sensing and Communication (ISAC) realizes communication and environmental perception functions through the same platform, and is widely used in scenarios such as drone monitoring, autonomous driving, and intelligent transportation systems, significantly improving the overall efficiency of the network. For example, base stations can simultaneously detect and track targets while providing communication services. With the complexity of application scenarios, a single base station is difficult to support wide-area and multi-target perception needs due to limited coverage and resources. To this end, the multi-base station collaborative ISAC system optimizes resource allocation and coordination through the joint work of multiple base stations, and improves perception accuracy and communication quality. Base stations can achieve collaborative optimization through methods such as joint beamforming, power allocation, and resource scheduling, expand perception coverage, balance communication and perception performance, and provide support for efficient operation in complex dynamic scenarios.

[0003] However, the existing multi-base station collaborative ISAC system has limitations in resource allocation and system optimization, mainly reflected in the fact that the fixed sensing mode restricts the flexibility of the system and leads to low resource efficiency.

[0004] At present, the resource allocation optimization scheme for ISAC systems with multi-base station collaboration usually adopts active sensing or passive sensing mode (that is, the fixed sensing mode is active sensing mode or passive sensing mode), and interference modeling and resource optimization are performed based on this. However, in different sensing modes, there are obvious differences in the propagation path, channel state and interference characteristics of the signal. For example, in the active sensing mode, the base station continuously transmits the sensing signal, resulting in a large sensing echo power, and the weak echo signal of the adjacent base station is often regarded as interference; while in the passive sensing mode, the system relies on the low-power echo signal in the environment, and the channel state and interference characteristics are more susceptible to changes in the external environment. Therefore, the fixed sensing mode cannot adapt to the real-time changes of the dynamic environment and task requirements according to the channel state and interference level. In scenarios with high interference or limited resources, the fixed sensing mode may lead to unreasonable resource allocation, resource waste or insufficient sensing performance. In multi-target task scenarios, the fixed sensing mode is also difficult to combine the importance of different targets with the real-time changes of tasks, and dynamically adjust resource scheduling and sensing strategies, thereby limiting the collaborative optimization effect of communication and sensing tasks and affecting the overall performance of the system.

[0005] In summary, the fixed perception mode adopted in the existing multi-base station cooperative ISAC system has the problems of being difficult to adapt to dynamic environmental changes, low resource utilization efficiency and limited communication perception performance due to insufficient system flexibility. Summary of the invention

[0006] The present invention provides a method and device for optimizing multi-base station collaborative perception resources based on generative AI, so as to solve the problems that the fixed perception mode adopted in the existing multi-base station collaborative ISAC system is difficult to adapt to dynamic environmental changes, the lack of system flexibility leads to low resource utilization efficiency and limited communication perception performance.

[0007] In order to solve the above technical problems, the embodiments of the present invention provide the following technical solutions:

[0008] In a first aspect, an embodiment of the present invention provides a method for optimizing multi-base station collaborative sensing resources based on generative AI, including:

[0009] Acquire first communication perception related information of a first multi-base station coordinated communication perception integrated ISAC system;

[0010] Inputting the first communication perception related information into a trained artificial intelligence (AI) resource optimization model to obtain first resource allocation information of a first multi-base station collaborative ISAC system and / or a first perception mode of a first multi-base station collaborative ISAC system output by the AI ​​resource optimization model;

[0011] The AI ​​resource optimization model is used to indicate at least one of the following:

[0012] The correspondence between communication perception related information and resource allocation information;

[0013] The correspondence between communication perception related information and perception patterns.

[0014] Optionally, the method further comprises:

[0015] Acquire second communication perception related information, a second perception mode, and second resource allocation information of a second multi-base station coordinated ISAC system;

[0016] The AI ​​resource optimization model is obtained by training using the second communication perception related information and the second perception mode, and / or using the second communication perception related information and the second resource allocation information.

[0017] Optionally, the communication awareness related information includes at least one of the following:

[0018] Base station;

[0019] Communication users;

[0020] Perceived goals;

[0021] Resource constraint information;

[0022] Communication perception indicator information.

[0023] Optionally, the perception mode includes at least one of the following:

[0024] Active sensing mode;

[0025] Collaborative awareness model.

[0026] Optionally, obtaining second resource allocation information of a second multi-base station coordinated ISAC system includes:

[0027] Acquire, in the active sensing mode, a downlink communication signal received by a communication user from a base station in the second multi-base station collaborative ISAC system, an echo signal reflected by a sensing target received by the base station, and beam power allocation information of the base station;

[0028] Generating a first optimization target for resource allocation optimization in the active sensing mode according to an echo signal reflected by the sensing target received by the base station;

[0029] Generate a first optimization constraint condition for resource allocation optimization in the active sensing mode according to a downlink communication signal received by the communication user from the base station and beam power allocation information of the base station;

[0030] According to the first optimization target and the first optimization constraint, resource allocation optimization is performed on the second multi-base station cooperative ISAC system in the active sensing mode to obtain second resource allocation information in the active sensing mode.

[0031] Optionally, generating a first optimization target for performing resource allocation optimization in the active sensing mode according to an echo signal reflected by a sensing target and received by the base station includes:

[0032] Performing separation and filtering processing on the echo signal reflected by the sensing target received by the base station to obtain a first filtered echo signal;

[0033] generating a Cramer-Rao bound CRLB of the perception performance of the perception target according to the first filtered echo signal;

[0034] Obtaining a first global positioning performance of the perception target according to the CRLB of the perception target, the echo arrival azimuth AoA of the perception target relative to the base station, and the echo arrival elevation angle ZoA of the perception target relative to the base station;

[0035] The first optimization objective is generated according to the first global positioning performance.

[0036] Optionally, generating a first optimization constraint condition for performing resource allocation optimization in the active sensing mode according to a downlink communication signal received by the communication user from a base station and beam power allocation information of the base station includes:

[0037] Obtaining a first signal-to-noise ratio (SINR) received by the communication user according to a downlink communication signal received by the communication user from a base station;

[0038] Taking the first SINR as greater than or equal to the first preset SINR threshold as a first condition;

[0039] Obtaining the transmit power of the base station and the number of beams supported by the base station according to the power allocation information of the base station and the beam power allocation information of the base station;

[0040] Taking the transmit power of the base station as less than or equal to the first preset power threshold as the second condition;

[0041] The number of beams supported by the base station is less than or equal to a first preset number as a third condition;

[0042] Generate a fourth condition according to a preset beam selection matrix;

[0043] The first optimization constraint condition is obtained according to the first condition, the second condition, the third condition and the fourth condition.

[0044] Optionally, performing resource allocation optimization on the second multi-base station cooperative ISAC system in the active sensing mode according to the first optimization objective and the first optimization constraint condition to obtain second resource allocation information in the active sensing mode includes:

[0045] Performing an approximate convex relaxation process on the fourth condition in the first optimization constraint condition to obtain a fifth condition;

[0046] Decomposing the first optimization objective into a first sub-objective and a second sub-objective by using an initial beam selection matrix;

[0047] Obtaining, according to the initial beam selection matrix, numerator information and denominator information in a first SINR received by the communication user;

[0048] Converting the first condition in the first optimization constraint condition into a sixth condition according to the numerator information and the denominator information;

[0049] Generate a first optimization problem according to the first sub-goal, the sixth condition and the second condition in the first optimization constraint condition;

[0050] Generate a second optimization problem according to the second sub-objective, the sixth condition, and the third condition and the fifth condition in the first optimization constraint condition;

[0051] By alternately solving the first optimization problem and the second optimization problem, second resource allocation information in the active sensing mode is obtained.

[0052] Optionally, obtaining second resource allocation information of a second multi-base station coordinated ISAC system includes:

[0053] Acquire, in the cooperative sensing mode, a transmission signal of a first base station in the second multi-base station cooperative ISAC system, a reception signal of a first communication user communicating with the first base station, and a reflection signal received by the first base station;

[0054] Obtaining a second SINR of the first communication user according to a transmission signal of the first base station and a reception signal of the first communication user;

[0055] Obtaining, according to the reflected signal received by the first base station, a third SINR of an output signal of a radar for sensing a target at the first base station;

[0056] Generating a second optimization target according to the second SINR;

[0057] Generating a second optimization constraint condition for resource allocation optimization in the collaborative sensing mode according to the third SINR and the power budget information of the first base station;

[0058] According to the second optimization objective and the second optimization constraint, resource allocation optimization is performed on the second multi-base station cooperative ISAC system in the cooperative sensing mode to obtain second resource allocation information in the cooperative sensing mode.

[0059] Optionally, generating, according to the third SINR and the power budget information of the first base station, a second optimization constraint condition for performing resource allocation optimization in the collaborative sensing mode includes:

[0060] The third SINR is greater than or equal to the second preset SINR threshold as a seventh condition;

[0061] The transmit power of the first base station is less than or equal to the second preset power threshold indicated by the power budget information of the first base station as an eighth condition;

[0062] Generate the ninth condition according to the preset target scheduling matrix;

[0063] According to the seventh condition, the eighth condition and the ninth condition, the second optimization constraint condition is obtained.

[0064] Optionally, performing resource allocation optimization on the second multi-base station cooperative ISAC system in the cooperative sensing mode according to the second optimization objective and the second optimization constraint condition to obtain second resource allocation information in the cooperative sensing mode includes:

[0065] According to the transmit beamforming matrix and the target scheduling matrix, converting the second optimization objective into a third optimization problem;

[0066] Using generalized Rayleigh entropy, solving the third optimization problem to obtain an optimized receive beamforming matrix;

[0067] Using the Dinkelbach method, the second optimization objective is transformed into a third sub-objective;

[0068] According to the third sub-objective and the sixth and seventh conditions in the second optimization constraint condition, a fourth optimization problem is obtained;

[0069] Using the Lagrange dual method, the fourth optimization problem is solved to obtain an optimized transmit beamforming matrix;

[0070] According to the optimized receive beamforming matrix, the optimized transmit beamforming matrix and the second optimization target, the target scheduling matrix is ​​optimized to obtain the second resource allocation information in the cooperative sensing mode.

[0071] Optionally, using the second communication perception related information and the second perception mode, and / or using the second communication perception related information and the second resource allocation information for training to obtain the AI ​​resource optimization model includes:

[0072] Performing feature extraction on the second communication perception related information to obtain a first extracted feature;

[0073] Performing feature clustering on the first extracted features to obtain multiple feature domains;

[0074] Performing weighted processing on the extracted features within the same feature domain to obtain a first domain feature representation;

[0075] Performing feature extraction and fusion on the extracted features in different feature domains to obtain a second domain feature representation;

[0076] Obtaining a target feature representation according to the first domain feature representation and the second domain feature representation;

[0077] The AI ​​resource optimization model is obtained by training using the target feature representation and the second perception mode, and / or using the target feature representation and the second resource allocation information.

[0078] Optionally, training using the target feature representation and the second resource allocation information to obtain the AI ​​resource optimization model includes:

[0079] Obtaining first training resource allocation information according to the target feature representation using a preset optimization method;

[0080] Using a preset initial AI model, obtaining second training resource allocation information according to the target feature representation;

[0081] Obtaining a loss function according to the first training resource allocation information and the second training resource allocation information;

[0082] The AI ​​resource optimization model is obtained by adjusting the model parameters in the initial AI model to reduce the loss function.

[0083] In a second aspect, an embodiment of the present invention further provides a multi-base station collaborative sensing resource optimization device based on generative AI, comprising:

[0084] A first acquisition module, configured to acquire first communication perception related information of a first multi-base station coordinated communication perception integrated ISAC system;

[0085] A first processing module, configured to input the first communication perception related information into a trained artificial intelligence (AI) resource optimization model, and obtain first resource allocation information of a first multi-base station collaborative ISAC system and / or a first perception mode of a first multi-base station collaborative ISAC system output by the AI ​​resource optimization model;

[0086] The AI ​​resource optimization model is used to indicate at least one of the following:

[0087] The correspondence between communication perception related information and resource allocation information;

[0088] The correspondence between communication perception related information and perception patterns.

[0089] In the third aspect, an embodiment of the present invention also provides a multi-base station collaborative sensing resource optimization device based on generative AI, comprising: a processor, a memory, and a program stored on the memory and executable on the processor, wherein when the program is executed by the processor, the steps in the multi-base station collaborative sensing resource optimization method based on generative AI as described in any one of the first aspects are implemented.

[0090] In a fourth aspect, an embodiment of the present invention further provides a readable storage medium, on which a program is stored, and when the program is executed by a processor, the steps in the generative AI-based multi-base station collaborative sensing resource optimization method as described in any one of the first aspects are implemented.

[0091] In a fifth aspect, an embodiment of the present invention further provides a computer program product, comprising computer instructions, which, when executed by a processor, implement the steps in the generative AI-based multi-base station collaborative sensing resource optimization method as described in any one of the first aspects.

[0092] The beneficial effects of the present invention are:

[0093] The multi-base station collaborative perception resource optimization method based on generative AI provided by the solution of the present invention obtains an AI resource optimization model through training, and obtains the first communication perception related information of the first multi-base station collaborative ISAC system by using the correspondence between the communication perception related information indicated in the AI ​​resource optimization model and the resource allocation information, and / or obtains the first perception mode of the first multi-base station collaborative ISAC system output by the AI ​​resource optimization model by using the correspondence between the communication perception related information indicated in the AI ​​resource optimization model and the perception mode. It breaks through the limitations of traditional fixed resource allocation, can be generalized to new scenario requirements according to existing scenario information, dynamically adjust the perception mode and resource allocation strategy, has a certain robustness, and can adapt to dynamic environmental changes and solve the problems of low resource utilization efficiency and limited communication perception performance caused by insufficient system flexibility. BRIEF DESCRIPTION OF THE DRAWINGS

[0094] Figure 1 A flowchart showing a method for optimizing multi-base station collaborative sensing resources based on generative AI provided in an embodiment of the present invention;

[0095] Figure 2 A schematic diagram showing a multi-base station cooperative communication and perception method in an active perception mode provided in an embodiment of the present invention;

[0096] Figure 3 A flowchart showing acquisition of resource allocation information in an active sensing mode provided by an embodiment of the present invention;

[0097] Figure 4 A flowchart showing acquisition of resource allocation information in a collaborative sensing mode provided by an embodiment of the present invention;

[0098] Figure 5 A schematic diagram showing cooperative communication and perception of multiple base stations provided in an embodiment of the present invention;

[0099] Figure 6 A diagram showing the training process of the AI ​​resource optimization model provided by an embodiment of the present invention;

[0100] Figure 7A schematic diagram showing the structure of a multi-base station collaborative sensing resource optimization device based on generative AI provided in an embodiment of the present invention;

[0101] Figure 8 A schematic diagram showing the structure of a multi-base station collaborative sensing resource optimization device based on generative AI provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0102] In order to make the technical problems, technical solutions and advantages to be solved by the application clearer, the following will be described in detail in conjunction with the accompanying drawings and specific embodiments. In the following description, specific details such as specific configurations and components are provided only to help fully understand the embodiments of the application. Therefore, it should be clear to those skilled in the art that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of the application. In addition, for clarity and brevity, the description of known functions and structures has been omitted.

[0103] It should be understood that the references to "one embodiment" or "an embodiment" throughout the specification mean that the specific features, structures, or characteristics associated with the embodiment are included in at least one embodiment of the present application. Therefore, the references to "in one embodiment" or "in an embodiment" appearing throughout the specification do not necessarily refer to the same embodiment. In addition, these specific features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.

[0104] In the various embodiments of the present application, it should be understood that the size of the serial numbers of the following processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0105] In the embodiments provided in the present application, it should be understood that "B corresponding to A" means that B is associated with A, and B can be determined according to A. However, it should also be understood that determining B according to A does not mean determining B only according to A, and B can also be determined according to A and / or other information.

[0106] In the embodiments of the present invention, the term "and / or" describes the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B may represent three situations: A exists alone, A and B exist at the same time, and B exists alone. The character " / " generally indicates that the associated objects before and after are in an "or" relationship.

[0107] In order to solve the problems that the fixed perception mode adopted in the existing multi-base station cooperative ISAC system is difficult to adapt to dynamic environmental changes, the insufficient flexibility of the system leads to low resource utilization efficiency and limited communication perception performance, the embodiments of the present invention provide a multi-base station cooperative perception resource optimization method and device based on generative AI.

[0108] like Figure 1 As shown, an embodiment of the present invention provides a method for optimizing multi-base station collaborative sensing resources based on generative AI, including:

[0109] Step 101: Acquire first communication perception related information of a first multi-base station coordinated communication perception integrated ISAC system.

[0110] The first ISAC system for multi-base station cooperation may be any ISAC system for multi-base station cooperation. The first ISAC system for multi-base station cooperation is a system for which sensing resources are to be optimally allocated.

[0111] The communication awareness related information includes at least one of the following:

[0112] Base station, wherein the communication perception related information also includes location information, quantity information, etc. of the base station;

[0113] Communication users, wherein the communication perception related information further includes location information, quantity information, type information, etc. of the communication users, wherein the type information of the communication users includes vehicles and user equipment (UE);

[0114] Perceived target, wherein the communication perception related information also includes location information, quantity information, etc. of the perceived target. In this embodiment, a drone is taken as an example for description;

[0115] Resource constraint information, wherein the resource constraint information includes at least one of the following: a SINR threshold, a maximum number of beams that the base station can support, and a maximum transmit power that the base station can support;

[0116] Communication perception indicator information, wherein the communication perception indicator information includes at least one of the following: a downlink communication signal received by a communication user from a base station, an echo signal reflected by a perception target received by the base station, power allocation information of the base station, beam power allocation information of the base station, a transmission signal of the base station, a reception signal of a communication user communicating with the base station, and interference of a perception target received by the base station.

[0117] That is, in this step, first communication perception related information in the first multi-base station cooperative ISAC system to perform resource optimization allocation is obtained, and the first perception communication related information is used to perform subsequent resource optimization allocation.

[0118] Step 102: Input the first communication perception related information into a trained artificial intelligence (AI) resource optimization model to obtain first resource allocation information of a first multi-base station cooperative ISAC system and / or a first perception mode of a first multi-base station cooperative ISAC system output by the AI ​​resource optimization model;

[0119] The AI ​​resource optimization model is used to indicate at least one of the following:

[0120] The correspondence between communication perception related information and resource allocation information;

[0121] The correspondence between communication perception related information and perception patterns.

[0122] That is, in this step, the AI ​​resource optimization model is first trained using historical data, and the AI ​​resource optimization model stores the correspondence between communication perception related information and resource allocation information and / or the correspondence between communication perception related information and perception mode.

[0123] The first communication perception related information is input into the trained AI resource optimization model, and the correspondence between the communication perception related information and the resource allocation information is used to obtain the first resource allocation information of the first multi-base station collaborative ISAC system output by the AI ​​resource optimization model, and / or, the correspondence between the communication perception related information and the perception mode is used to obtain the first perception mode of the first multi-base station collaborative ISAC system output by the AI ​​resource optimization model, that is, the perception mode and resource allocation are determined by the model, which breaks through the limitations of traditional fixed resource allocation, can be generalized to new scenario requirements according to existing scenario information, dynamically adjust the perception mode and resource allocation strategy, has a certain degree of robustness, and can adapt to dynamic environmental changes and solve the problems of low resource utilization efficiency and limited communication perception performance caused by insufficient system flexibility.

[0124] The sensing mode includes at least one of the following:

[0125] Active sensing mode, wherein in the ISAC system in which multiple base stations cooperate, the same base station transmits a sensing signal and receives an echo signal reflected by a sensing target;

[0126] The cooperative sensing mode is that in the ISAC system with multiple base stations cooperating, one base station transmits a sensing signal, and another different base station receives an echo signal reflected by a sensing target.

[0127] Furthermore, the communication perception related information includes:

[0128] Communication sensing related information in active sensing mode;

[0129] Communication awareness related information in collaborative awareness mode.

[0130] It should be noted that the active perception mode can also be called the A-transmit and A-receive mode. In the A-transmit and A-receive mode, the base station is responsible for transmitting signals and receiving their echoes at the same time to achieve active perception. This mode is suitable for high-precision perception tasks within the range of a single base station. The key features of this mode are: Signal modeling: The base station transmits a perception signal and receives the echo signal reflected from the target. In this mode, the signal modeling of the base station is based on the Cramer-Rao lower bound (CRLB), and the perception accuracy is optimized by minimizing the CRLB; Beam and power allocation: While transmitting the signal, the base station performs beamforming and power allocation to ensure accurate perception of the target. Through the optimization of CRLB, the base station can use resources more efficiently in the perception task; Application scenario: Suitable for tasks that require high-precision perception, such as precise positioning and tracking of drone targets. In this mode, the base station can complete the perception task independently without relying on the cooperation of other base stations.

[0131] The collaborative sensing mode can also be called the A-transmit-B-receive mode. The A-transmit-B-receive mode completes the sensing task through the collaboration between multiple base stations. Base station A is responsible for transmitting the sensing signal, and base station B receives the echo signal and processes it, thereby realizing the perception of the target. The key features of this mode are: Signal modeling: In the A-transmit-B-receive mode, base station A transmits communication and sensing signals, and base station B is responsible for receiving the sensing signal reflected from the target. Through the collaboration of multiple base stations, the sensing coverage is expanded and the quality of the sensing signal is improved; Beam and power allocation: Due to the collaboration of multiple base stations, the system needs to reasonably allocate power and beam direction among multiple base stations to avoid mutual interference between base stations and ensure the overall performance of the system is optimal. Beamforming and power allocation are more complex in this mode, and the resources of multiple base stations need to be coordinated; Application scenarios: It is suitable for scenarios with wider coverage and higher target complexity, such as drone clusters or large-scale monitoring tasks. The collaboration of multiple base stations can improve the system's perception accuracy and anti-interference ability of the target.

[0132] The following specifically describes the training process of the AI ​​resource optimization model provided by the embodiment of the present invention:

[0133] In one embodiment, the method further comprises:

[0134] Acquire second communication perception related information, a second perception mode, and second resource allocation information of a second multi-base station coordinated ISAC system;

[0135] Among them, the number of the second multi-base station collaborative ISAC system can be one or more, and it is a system used for model training.

[0136] It should be noted that the parameters in the second communication perception related information and the first communication perception related information are the same, indicating that their specific values ​​are different.

[0137] Among them, the second resource allocation information is obtained according to the second notification perception information, and the second perception mode is set according to the application scenario of the ISAC system of multi-base station collaboration and other needs.

[0138] The AI ​​resource optimization model is obtained by training using the second communication perception related information and the second perception mode, and / or using the second communication perception related information and the second resource allocation information.

[0139] Specifically, the second communication perception-related information and the second perception mode are used to train the correspondence between the communication perception-related information and the perception mode in the AI ​​resource optimization model, and the second communication perception-related information and the second resource allocation information are used to train the correspondence between the communication perception-related information and the resource allocation information in the AI ​​resource optimization model.

[0140] In an optional embodiment, the process of obtaining communication sensing related information in the active sensing mode is specifically described:

[0141] In this active sensing mode, the base station is equipped with a massive multiple-input multiple-output (mMIMO) uniform planar array (UPA) and has the ability to sense through ISAC technology. The base station uses independent transmit and receive antennas to receive and sense echo signals while maintaining downlink communication.

[0142] In view of this mode, the embodiment of the present invention models the relationship between the downlink communication signal and the active sensing echo signal and the base station transmission beam in the multi-base station cooperation scenario, and derives the signal-to-interference-plus-noise ratio (SINR) expression of the ground communication user and the Cramer-Rao lower bound (CRLB) of the aerial drone sensing performance. By minimizing the CRLB, the sensing accuracy is improved when the position of the drone (i.e., the sensing target) is approximately known.

[0143] That is, in this optional embodiment, obtaining second resource allocation information of a second multi-base station coordinated ISAC system includes:

[0144] In the active sensing mode, the downlink communication signal received by the communication user from the base station in the second multi-base station collaborative ISAC system, the echo signal reflected by the sensing target received by the base station, the power allocation information of the base station, and the beam power allocation information of the base station are obtained.

[0145] Specifically, in this active sensing mode, all base stations (ISAC BSs) provide sensing and communication services simultaneously in a multi-beam mode. In terms of communication services, ground communication users with different quality of service (QoS) requirements for downlink communications, such as vehicles and user equipment (UE), are considered. In terms of sensing services, each ISAC BS receives the echo of its own transmitted signals, which are reflected by the unmanned aerial vehicle (UAV). By analyzing the reflected signals, each ISAC BS uploads its positioning results to the fusion center for data integration. Similar to ground communication users, UAVs present different threat levels in different areas, and thus have different perception QoS requirements.

[0146] In this active sensing mode, see Figure 2 , Figure 2 Figure 1 is a schematic diagram of the cooperative communication and sensing method of multiple base stations in the active sensing mode. The three adjacent sectors of three adjacent BSs are regarded as a cooperative communication and sensing integrated unit, namely, the cooperative ISAC unit (the area with the same color in the figure). Figure 2 The non-cooperative UAV shown in the figure is the sensing target. In the cooperative ISAC unit, the sensing targets include K UAVs and the communication users include M single-antenna UEs. Each base station is equipped with N t transmit antennas and N r The base station transmits multiple beams to the UAV or UE through beamforming (BF) technology. Each UAV can reflect the beam from one or more BSs. After receiving the echo from the UAV, each base station performs sensing signal level fusion in one base station through data interaction between base stations to obtain the sensing result; at the same time, each base station transmits multiple beams to transmit downlink data with the UE.

[0147] The baseband signal of the nth base station is defined as:

[0148]

[0149] Where J = K + M represents the total number of UAVs and UEs. The ISAC signal after beamforming is:

[0150]

[0151] in represents the transmit beamforming matrix, represents the beam selection vector, the binary variable g j,n When it is 1, it indicates that base station n allocates a beam pointing to sensing target j, and when it is 0, it indicates that base station n does not allocate a beam pointing to sensing target j.

[0152] For a communication UE, the downlink communication signal received by the mth UE from the nth base station is:

[0153]

[0154]

[0155] Where N represents the total number of cooperative base stations, represents antenna gain, M represents the set of UEs, α m,n represents the path loss between the mth UE and the nth base station, p m,n represents the beam power allocated by the nth base station to the mth UE, p k,n represents the beam power allocated by the nth base station to the kth UAV, p m,n and p k,n The beam power allocation information of the base station includes p m,n and p k,n ), and θ m,n They represent the azimuth and elevation angles of the mth UE relative to the nth base station, The variance is Additive Gaussian white noise, represents the steering vector of the transmitting antenna, which is calculated as:

[0156]

[0157] Where N y and N z Represent the number of UPA antennas in each row and column respectively.

[0158] The echo signal received by the nth base station from all UAVs is:

[0159]

[0160] in represents the antenna gain, β k,n , μ k,n and τ k,n They represent the reflection coefficient, Doppler frequency and delay of the kth UAV relative to the nth base station, The variance is Gaussian white noise, and θ k,n They represent the azimuth angle of arrival (AoA) and zenith angle of arrival (ZoA) of the echo of the k-th UAV relative to the n-th base station, represents the receiving antenna steering vector.

[0161] And, the beamformer is designed as follows:

[0162]

[0163] in and Represents the estimated elevation and azimuth angles.

[0164] According to the echo signal reflected by the sensing target received by the base station, the first optimization target for resource allocation optimization in the active sensing mode is generated, that is, according to the above formula (4), the first optimization target is generated.

[0165] According to the downlink communication signal received by the communication user from the base station and the beam power allocation information of the base station, a first optimization constraint condition for optimizing resource allocation in the active sensing mode is generated.

[0166] The first optimization objective and the first optimization constraint condition obtain the optimization problem after modeling.

[0167] According to the first optimization target and the first optimization constraint, resource allocation optimization is performed on the second multi-base station cooperative ISAC system in the active sensing mode to obtain second resource allocation information in the active sensing mode.

[0168] Optionally, generating a first optimization target for performing resource allocation optimization in the active sensing mode according to an echo signal reflected by a sensing target and received by the base station includes:

[0169] The echo signal reflected by the sensing target received by the base station is subjected to separation and filtering processing to obtain a first filtered echo signal, specifically:

[0170] According to formula (4), the echoes from different UAVs can be separated by a space-separation filter, and the echoes after filtering are:

[0171]

[0172] where w k,n represents the receive beamforming vector of the nth base station in the direction of the kth UAV, represents additive white Gaussian noise.

[0173] A Cramer-Rao bound CRLB of the perception performance of the perception target is generated according to the first filtered echo signal.

[0174] Specifically, the embodiment of the present invention uses CRLB to evaluate the performance of multi-base station collaborative perception of UAVs and derives its expression. On this basis, the beam and power allocation optimization problem of the multi-base station collaborative ISAC system is proposed, and then a multi-BS collaborative beam and power allocation (MCBPA) algorithm is proposed to solve this problem.

[0175] The derivation process of the perception CRLB of the drone under multi-base station cooperation is as follows:

[0176] When the drone position is approximately known a priori, the estimation performance can be improved by minimizing CRLB. CRLB is the lower bound of the variance of various unbiased estimators and is the inverse of the Fisher information matrix (FIM), which can be written as:

[0177]

[0178] in is the true value of the UAV distance, azimuth, and pitch angle, J k Given by:

[0179]

[0180] where r′ k,n is the echo without noise (i.e. the first filtered echo signal), J k,n is a 3×3 matrix, the (i,j)th element is denoted by J k,n (i, j), the perceived CRLB of the kth UAV is obtained as follows:

[0181]

[0182] Among them, α1, α2 and α3 depend on the base station frequency, bandwidth and number of antennas, etc. k,n Given by:

[0183]

[0184] Afterwards, the UAV differentiated QoS perception performance optimization problem is modeled:

[0185] When optimizing multi-target sensing performance, the prior art usually adopts the method of minimizing the overall CRLB or minimizing the maximum CRLB. However, this may lead to resource allocation being concentrated on a single target. Under such a criterion, the system tends to allocate resources to the target with the highest potential gain, even if the target does not require such high sensing accuracy. Taking this into account, this embodiment adopts an exponential utility function to represent the perception performance of drones with different QoS requirements.

[0186] Specifically, without loss of generality, using d k The FIM is used as the performance indicator of the k-th UAV perception and is expressed as:

[0187] ξ k =J k (1,1) (11)

[0188] If AoA and ZoA need to be considered, it can be expanded to a weighted FIM of three parameters. The global positioning performance of all UAVs (i.e., the first global positioning performance) can be expressed as:

[0189]

[0190] where η k represents the expected positioning CRLB of the kth UAV. Due to the characteristics of the exponential utility function, when the variable ξ k Greater than η k When φ is larger than when ξ is smaller than ξ k Less than η k This property suggests that allocating resources to drones that have not yet reached the expected sensing performance will yield more benefits.

[0191] Therefore, the first optimization objective in the optimization problem can be modeled as maximizing the first global positioning performance. It can be expressed as:

[0192]

[0193] Optionally, generating a first optimization constraint condition for performing resource allocation optimization in the active sensing mode according to a downlink communication signal received by the communication user from a base station and beam power allocation information of the base station includes:

[0194] Obtaining a first signal-to-noise ratio (SINR) received by the communication user according to a downlink communication signal received by the communication user from a base station;

[0195] Specifically, according to the above formula (1) to formula (3), the SINR of the mth UE is:

[0196]

[0197] The first SINR is greater than or equal to the first preset SINR threshold as the first condition, wherein the resource constraint information includes the first preset SINR threshold, and the first condition is expressed as follows:

[0198] SINR mm ≥γ m ,m=1,2,...,M,

[0199] Among them, γ m represents the receiving SINR threshold (ie, the first preset power threshold) of the mth UE.

[0200] Obtaining the transmit power of the base station and the number of beams supported by the base station according to the power allocation information of the base station and the beam power allocation information of the base station;

[0201] The transmit power of the base station is less than or equal to the first preset power threshold as the second condition, and the second condition is expressed as follows:

[0202]

[0203] Where P = [p1, p2, ..., p n ] T is the power allocation matrix (i.e., the beam power allocation information of the base station), p n =[p 1,n ,p 2,n ,...,p j,n ] T is the beam power allocation vector of the nth base station, G = [g1, g2, ..., g n ] T is the beam selection matrix, The maximum transmit power of the nth base station (ie, the first preset power threshold), wherein the resource constraint information includes the maximum transmit power of the base station.

[0204] The third condition is that the number of beams supported by the base station is less than or equal to the first preset number. The third condition is expressed as follows:

[0205]

[0206] Where G = [g1, g2, ..., g n ] T is the beam selection matrix, It indicates the maximum number of beams supported by the nth base station (ie, the first preset number), wherein the resource constraint information includes the maximum number of beams supported by the base station.

[0207] Generate a fourth condition according to a preset beam selection matrix;

[0208] Specifically, the fourth condition is expressed as follows:

[0209]

[0210] Where G = [g1, g2, ..., g n ] T is the beam selection matrix.

[0211] The first optimization constraint condition is obtained according to the first condition, the second condition, the third condition and the fourth condition.

[0212] In summary, the optimization problem obtained according to the first optimization objective and the first optimization constraint is expressed as follows:

[0213]

[0214] Where P = [p1, p2, ..., p n ] T is the power allocation matrix, p n =[p 1,n ,p 2,n ,...,p j,n ] T is the beam power allocation vector of the nth base station, G = [g1, g2, ..., g n ] T is the beam selection matrix, γ m represents the receiving SINR threshold of the mth UE, and They respectively represent the maximum transmit power and the maximum number of beams supported by the nth base station.

[0215] Optionally, performing resource allocation optimization on the second multi-base station cooperative ISAC system in the active sensing mode according to the first optimization objective and the first optimization constraint condition to obtain second resource allocation information in the active sensing mode includes:

[0216] The fourth condition in the first optimization constraint condition is subjected to approximate convex relaxation to obtain the fifth condition.

[0217] Specifically, due to the coupling and discontinuity of the optimization variables, it is challenging to solve the optimization problem directly. Therefore, we first constrain 0≤g j,n ≤1 instead of g j,n ∈{0,1} to make an approximate convex relaxation.

[0218] Using the initial beam selection matrix, the first optimization objective is decomposed into a first sub-objective and a second sub-objective. Specifically, the original optimization problem is decomposed into two sub-problems, and P and G are optimized alternately. The first sub-objective is to maximize P, and the second sub-objective is to maximize G.

[0219] Obtaining, according to the initial beam selection matrix, numerator information and denominator information in a first SINR received by the communication user;

[0220] Among them, through the initial beam selection matrix G (initialized as an all-one matrix), use a m (p) and bm (p) represents the numerator and denominator in the UE receive SINR expression.

[0221] According to the numerator information and the denominator information, the first condition in the first optimization constraint condition is converted into a sixth condition. Specifically, the sixth condition is expressed as follows:

[0222] a m (p)≥γ m b m (p),m=l,...,M,

[0223] A first optimization problem is generated according to the first sub-objective, the sixth condition and the second condition in the first optimization constraint condition.

[0224] That is, the original optimization problem can be written as subproblem 1 (i.e. the first optimization problem), as follows:

[0225]

[0226] Generate a second optimization problem according to the second sub-objective, the sixth condition, and the third condition and the fifth condition in the first optimization constraint condition;

[0227] Subproblem 2 (i.e., the second optimization problem) related to the beam selection matrix is ​​as follows:

[0228]

[0229] By alternately solving the first optimization problem and the second optimization problem, second resource allocation information in the active sensing mode is obtained.

[0230] Subproblem 1 is a convex problem and can be solved by a convex optimization solver. Then, the optimization result P * As a fixed variable in the original optimization problem, we get subproblem 2 about the beam selection matrix G. Similarly, subproblem 2 is convex and can be solved by a convex optimization solver. Then, the optimization result G * Use it as a fixed variable and continue to solve subproblem 1. By alternately solving the two subproblems, the objective function is finally converged. The resource allocation information in the converged system is the second resource allocation information in the active sensing mode.

[0231] Combine the following Figure 3 , specifically describes the process of acquiring resource allocation information in the active sensing mode provided by an embodiment of the present invention:

[0232] The prior location information of the unmanned aerial vehicle (UAV) and the UE is obtained, and the synaesthesia signal modeling and analysis of the collaborative perception unit are performed. The communication performance metric and the perception performance metric are calculated based on the signal modeling. The optimization problem is established according to the communication QoS and perception QoS requirements. After the optimization problem is solved, the beam selection vector and power allocation vector of each base station are obtained, and the base station performs beamforming and power regulation.

[0233] In view of this mode, the embodiment of the present invention models the relationship between the downlink communication signal and the active sensing echo signal and the base station transmission beam in the multi-base station cooperation scenario, and derives the signal-to-interference-plus-noise ratio (SINR) expression of the ground communication user and the Cramer-Rao lower bound (CRLB) of the aerial drone perception performance. By minimizing the CRLB, the perception accuracy is improved when the drone position is approximately known.

[0234] In terms of resource allocation optimization, in order to solve the problem that traditional methods lead to excessive concentration of resources on a single target, the embodiment of the present invention designs a global perception performance evaluation function based on an exponential utility function for multiple UAV targets with different perception QoS requirements. Taking this as the optimization goal, an optimization problem is constructed in combination with the base station beam capability, transmission power constraints, and ground user communication QoS requirements. It is decomposed into two sub-optimization problems through an approximate convex relaxation method, and solved using an alternating iterative algorithm, ultimately achieving efficient allocation of multi-base station collaborative beams and power.

[0235] By rationally allocating resources, the perception performance is significantly improved, multi-target requirements are taken into account, and the transmission power of the base station is reduced.

[0236] In an optional embodiment, the process of obtaining communication awareness related information in the collaborative awareness mode is specifically described:

[0237] like Figure 4 As shown in the figure, in the considered cooperative sensing mode, the ISAC scenario consists of J synchronized base stations, K communication users (CUs) and I UAV sensing targets, which can be expressed as and Each BS is equipped with a unified planar array (UPA) with M transmitting antenna arrays. t =M tx ×M ty , the receiving antenna array is configured as M r =M rx ×M ry To achieve ISAC capability, each BS can communicate with multiple CUs and sense multiple targets simultaneously. In addition, it is assumed that each CU is pre-assigned to a related BS. represents the CU set associated with BSj, k jDenotes the kth CU associated with BSj. For sensor target scheduling, define a binary target scheduling variable (i.e., target scheduling matrix, which is preset) a j,i ∈{0,1}. If a j,i =1, indicating that BS j receives and processes the reflected signal of target i, otherwise a j,i =0 means that BS j does not receive the reflected signal of sensing target i for processing.

[0238] In an optional embodiment, obtaining second resource allocation information of a second multi-base station coordinated ISAC system includes:

[0239] Acquire, in a cooperative sensing mode, a transmission signal of a first base station in the second multi-base station cooperative ISAC system, a reception signal of a first communication user communicating with the first base station, and a reflected signal received by the first base station; the first base station is any base station in the second multi-base station cooperative ISAC system; the first communication user is any communication user communicating with the first base station;

[0240] First, for the downlink signal model of the base station, the transmission signal of BS j (i.e., the first base station) can be expressed as:

[0241] x j′ =W j′ s j′ (17)

[0242] in represents the transmit beamforming matrix (preset) of the communication users served by BS j, Represents the communication symbol vector. Without loss of generality, it is assumed that the communication symbols from different CUs are unrelated, that is, for the communication user For other communication users

[0243] Therefore, based on the signal model shown in the above formula (17), CUk communicating with BS j′ can be constructed: j′ The received signal model of the (first communication user) is expressed as the following formula (18):

[0244]

[0245] in, They represent the distance from BS to CU k j′ The first part of formula (18) is the expected communication channel from BS j′ to CU k j′The second part is the same-cell interference from other CUs, and the third part is the inter-cell interference from other BSs. The last part is the communication signal of CU k j′ The additive white Gaussian noise (AWGN) at the receiver, where

[0246] Considering that the moving speed of general drone targets is low, the ability to sense the target's motion state change is relatively small compared to the ability to transmit and process the sensed signal. Therefore, the accurate direction of the target can be estimated in advance to design the transmission signal to detect the target. Then the reflected signal received by BSj can be expressed as:

[0247]

[0248] Among them, n r [l] is the receiving noise of receiving BS j, represents the direct channel from BS j′ to BS j. If j′≠j, then G′ j ,j=1, otherwise it is 0. β j′,j,i is the reflection coefficient from the transmitting BS j′ through the sensing target to the receiving BS j, including the round-trip path loss and RCSσ rcs , according to the radar equation, we get formula (20):

[0249]

[0250] Where λ represents the wavelength, d1 and d2 represent the distance from the transmitting BS to the sensing target and the distance from the sensing target to the receiving BS. Assuming that the antenna spacing of the BS is half a wavelength, the response vectors of the antenna array to the azimuth and elevation angles are expressed as:

[0251]

[0252]

[0253] Where m tx ∈[1,M tx ] and m ty ∈[1,M ty ] is the antenna index.

[0254] Obtaining a second SINR of the first communication user according to a transmission signal of the first base station and a reception signal of the first communication user;

[0255] According to formula (17) and formula (18), we get:

[0256] CU k communicates with BS j′ j′ The SINR (i.e., the second SINR) can be expressed as:

[0257]

[0258] in, and They are intra-cell interference and inter-cell interference, which are expressed as:

[0259]

[0260]

[0261] Obtaining, according to the reflected signal received by the first base station, a third SINR of an output signal of a radar for sensing a target at the first base station;

[0262] In the cooperative sensing ISAC scenario, BS j receives reflected sensing signals from multiple BSs through multiple sensing targets. In order to capture the desired signal of target i, a receive beamforming filter is applied on the received sensing signal y. Considering that the premise of perception is that the reflected signal can be detected, the perceived signal strength is usually positively correlated with the perception accuracy and detection probability. Therefore, the SINR of the received perception signal can be used as a measure of perception performance.

[0263] Based on the above formula (17), the radar output signal SINR (third SINR) of the target i sensed at BS j can be expressed as:

[0264]

[0265] in, In formula (26), Γ j,i represents the total interference perceived by target i at BS j, which can be expressed as:

[0266]

[0267] Among them, the first item represents the direct interference signal from other base stations, the second item is noise, and the third item is the interference signal from other sensing targets.

[0268] According to the second SINR, a second optimization target is generated. Under the maximum-minimum fairness principle, the SINR of the perceived signal of the receiving base station is maximized, that is, maximizing the second SINR is the second optimization target, which is expressed as follows:

[0269]

[0270] A second optimization constraint condition for resource allocation optimization in the collaborative sensing mode is generated according to the third SINR and the power budget information of the first base station, wherein the second optimization constraint condition satisfies the communication SINR of the communication user and the power budget constraint of the base station.

[0271] According to the second optimization objective and the second optimization constraint, resource allocation optimization is performed on the second multi-base station cooperative ISAC system in the cooperative sensing mode to obtain second resource allocation information in the cooperative sensing mode.

[0272] Optionally, generating, according to the third SINR and the power budget information of the first base station, a second optimization constraint condition for performing resource allocation optimization in the collaborative sensing mode includes:

[0273] The third SINR being greater than or equal to the second preset SINR threshold is taken as the seventh condition, which is expressed as follows:

[0274]

[0275] Among them, γ min It is the minimum SINR threshold required by the communication user, that is, the second preset SINR threshold, and the resource constraint information includes the minimum SINR threshold required by the communication user.

[0276] The transmit power of the first base station is less than or equal to the second preset power threshold indicated by the power budget information of the first base station as the eighth condition, which is expressed as follows:

[0277]

[0278] in, is the transmit beamforming matrix of the base station, that is, the transmit power of the first base station, P0 is the maximum transmit power of each base station, that is, the second preset power threshold, and the resource constraint information includes the maximum transmit power of each base station.

[0279] According to the preset target scheduling matrix, the ninth condition is generated, which is expressed as follows:

[0280]

[0281] According to the seventh condition, the eighth condition and the ninth condition, the second optimization constraint condition is obtained.

[0282] Under the maximum-minimum fairness principle, the SINR of the received base station's perceived signal is maximized while satisfying the communication SINR of the communication user and the power budget constraint of the base station. This problem is formulated as the target scheduling variable a j,i , Base station transmit beamforming matrix and the receive beamforming vector (receive beamforming matrix) The joint optimization problem is as follows:

[0283]

[0284]

[0285] Among them, γ min is the minimum SINR threshold required by the communication user, and P0 is the maximum transmit power of each base station.

[0286] Due to the multiplication of the beamforming matrix and the target scheduling matrix in the objective function, the optimization problem is a non-convex mixed integer program. In addition, the fractional structure of the objective function also leads to a non-convex optimization problem, which is difficult to solve. Therefore, the original problem needs to be transformed. The transformation of the original problem includes the following steps:

[0287] That is, optionally, according to the second optimization objective and the second optimization constraint, performing resource allocation optimization on the second multi-base station cooperative ISAC system in the cooperative sensing mode to obtain second resource allocation information in the cooperative sensing mode includes:

[0288] According to the transmit beamforming matrix and the target scheduling matrix, converting the second optimization objective into a third optimization problem;

[0289] Using generalized Rayleigh entropy, solving the third optimization problem to obtain an optimized receive beamforming matrix;

[0290] Specifically, the receive beamforming optimization problem: In this subproblem, the transmit beamforming matrix and the target scheduling matrix are given. Then, the original problem only depends on u j,i , the transformed problem (i.e., the third optimization problem) can be solved by generalized Rayleigh entropy.

[0291] Using the Dinkelbach method, the second optimization objective is transformed into a third sub-objective;

[0292] According to the third sub-objective and the sixth and seventh conditions in the second optimization constraint condition, a fourth optimization problem is obtained;

[0293] Solving the fourth optimization problem to obtain an optimized transmit beamforming matrix;

[0294] Specifically, transmit beamforming optimization problem: In this subproblem (the fourth optimization problem), the fractional structure of the original problem is first transformed using the Dinkelbach method, and then the Lagrangian dual method is used to obtain the optimal transmit beamforming matrix.

[0295] According to the optimized receive beamforming matrix, the optimized transmit beamforming matrix and the second optimization target, the target scheduling matrix is ​​optimized to obtain the second resource allocation information in the cooperative sensing mode.

[0296] That is, the target scheduling optimization problem: based on the optimized transmit and receive beamforming variables, the target scheduling variables are optimized according to the highest SINR.

[0297] The receive beamforming optimization sub-problem is described in detail below:

[0298] Given the transmit beamforming matrix and the target scheduling matrix, the receive beamforming matrix depends only on the target. Therefore, the receive beamforming matrix is ​​optimized by the following optimization criterion:

[0299]

[0300] Where B represents the direct channel interference from other targets and other base stations, and Q represents the covariance matrix of the downlink ISAC signal, which can be expressed as

[0301]

[0302]

[0303] From the expression of formula (32), it can be seen that the subproblem maximizes the minimum perceptual signal-to-noise ratio between targets, thereby obtaining the optimal receive beamforming solution. When the transmit beamforming and target scheduling are fixed, u j,i The optimization of u for the detection of target i does not affect the performance of other targets. Therefore, u for each target j,i The optimization of can be performed separately. On this basis, the maximum minimization optimization problem formula (32) is equivalent to the following problem:

[0304]

[0305] The transformed problem formula (32) is a generalized Rayleigh quotient, which can be obtained The optimal solution is:

[0306]

[0307] Among them, V max (Π) represents the eigenvector corresponding to the largest eigenvalue of the matrix H.

[0308] Transmit beamforming optimization subproblem:

[0309] Due to a j,i is a 0-1 integer variable. First, a j,i Relaxation is a continuous variable aj,i ∈[0,1]. Given the optimal receive beamforming matrix and target scheduling variables, the transmit beamforming subproblem can be expressed as:

[0310]

[0311]

[0312]

[0313] Among them, C j,i ,D j,i ,E j,i It can be expressed as follows:

[0314]

[0315]

[0316]

[0317] It can be seen that subproblem P3 is about transmit beamforming. The fractional structure of the objective function leads to a non-convex problem, which is difficult to solve directly. Inspired by Dinkelbach's method, the non-convexity of subproblem P3 can be transformed into a solvable problem. The optimal solution of subproblem P3 can be obtained if and only if the following equation holds.

[0318]

[0319] Among them, η ＊ is the optimal non-negative parameter, and Represent the numerator and denominator of the objective function in P3 respectively.

[0320]

[0321]

[0322] The proof of the above conclusion can be obtained from the generalized fractional programming theory. Therefore, problem P3 can be transformed into the following problem:

[0323]

[0324] By introducing the auxiliary variable γ, problem P4 can be transformed into:

[0325]

[0326]

[0327] Although P5 is more solvable, it is still a non-convex quadratic problem. In order to obtain the optimal solution, the Lagrangian dual method is used to analyze the optimal solution of the P5 problem with some strong duality conditions. The Lagrangian function can be expressed as:

[0328]

[0329] Where δ={δ j,i} j∈J,i∈I , and ζ={ζ j′} j′∈J is the dual variable.

[0330] Π is expressed as:

[0331]

[0332] Therefore, the dual function is expressed as:

[0333]

[0334] In order to ensure that the Lagrangian dual function formula (49) is not unbounded, it is obvious that Π ≥ 0 and Therefore, the dual problem is expressed as:

[0335]

[0336]

[0337]

[0338] For convex optimization problems, strong duality can always be guaranteed under the Slater condition, but it is more difficult for non-convex optimization problems. However, non-convex problems can also obtain strong duality under certain conditions. The following theorem clarifies the relationship between the existence of saddle points and strong duality.

[0339] Grange function The saddle point The following conditions should be met:

[0340]

[0341] Theorem 1: Let p ＊ ,d ＊ are the optimal values ​​of the non-convex quadratic problem and the dual problem respectively. If yes The strong duality d ＊ =p ＊ On the contrary, if d ＊ is finite, and (δ ＊ ,α ＊ ,ζ＊ )=argmax δ,α,ζ≥0 g(δ,α,ζ), the original problem is There is an optimal solution at yes saddle point.

[0342] Theorem 2: The condition for the existence of a saddle point is: there exists (δ ＊ ,α ＊ ,ζ ＊ ) makes is convex and has a minimum value that satisfies

[0343]

[0344]

[0345]

[0346]

[0347]

[0348]

[0349] The proof of the theorem can be easily generalized from the deduction of the prior art to multiple constraints.

[0350] Therefore, in order to prove the strong duality of the non-convex quadratic optimization problem and obtain the optimal solution by the dual method, it is necessary to find the saddle point that satisfies Theorem 2.

[0351] Due to the convexity of the dual problem P6 with respect to the Lagrange multiplier, its optimal solution (δ ＊ ,α ＊ ,ζ ＊ ) can be obtained by standard convex optimization techniques or interior point methods. Considering that Π is defined as a positive definite matrix in (50) to ensure the finiteness of the dual function formula (49), the Lagrangian function formula (49) is proved to be and γ are convex. Therefore, we only need to find:

[0352]

[0353] If Π>0, it means that Π is a full-rank matrix, and its eigenvalues ​​are all positive, so the null space of Π is the zero vector, is designed to minimize Equation (49). If Π≯0, then there is an eigenvalue that is 0. Therefore, Π is not a full rank matrix and has a non-zero null space. From this derivation, the null space of Π spans the unit norm eigenvector u Π, whose eigenvalue is 0 (i.e. Πu п. =0). Therefore, can be chosen to minimize formula (49) and can be calculated from formula (60).

[0354] Based on this, ＊ It can also be obtained from condition (56), that is:

[0355]

[0356] is equivalent to:

[0357]

[0358] Sub-problems of the target scheduling optimization problem:

[0359] Observe the Lagrangian function (49), a j,i Only the first term is relevant. Therefore, given the strong duality proved above, the Lagrangian function can be decomposed into multiple independent functions and solved independently:

[0360]

[0361] Therefore, the independent sub-problems can be expressed as:

[0362]

[0363] In order to minimize each The best a j,i It can be expressed as:

[0364]

[0365] Consider each goal The BS scheduling decisions are independent of each other, so the target scheduling problem can be decoupled into independent decisions for each target. This significantly simplifies the problem, allowing us to schedule each target independently based on the highest achievable SINR. This approach not only simplifies the computation process, but also complies with the max-min fairness criterion, ensuring that targets get enough resources even under the worst sensing conditions.

[0366] By iteratively calculating the above three sub-problems, the multi-base station collaborative sensing resource allocation results can be obtained.

[0367] This section proposes a multi-base station cooperative sensing and communication integrated (ISAC) resource scheduling method based on active and passive collaboration, in which multiple base stations simultaneously perform multi-target sensing and communication. By using downlink communication signals for sensing and combining receive beamforming, a signal-to-interference-plus-noise ratio (SINR) model for target sensing and downlink communication is mathematically established. The optimization problem is formulated as a joint optimization of multi-target scheduling, base station transmit beamforming, and receive beamforming, with the goal of maximizing the minimum radar sensing SINR. Given transmit beamforming and target scheduling, the optimal receive beamforming matrix is ​​obtained by solving the generalized Rayleigh entropy problem; the Dinkelbach method is used to process the fractional structure of the original problem, and then the Lagrange dual method is used to derive the optimal transmit beamforming matrix; based on the optimization of the transmit and receive beamforming variables, the target scheduling variables are optimized according to the highest SINR to ensure the fairness of multi-target sensing. The final resource optimization result is obtained through iterative calculation of the three sub-algorithms. The process is as follows: Figure 5 shown.

[0368] The base station obtains the prior location information of the unmanned aerial vehicle (UAV) and the UE, models the active and passive collaborative sensing signals, that is, the signal modeling in the collaborative sensing mode, optimizes the receiving beamforming, optimizes the transmitting beamforming, and optimizes the sensing target scheduling.

[0369] Considering that in the real environment, the traditional optimization algorithm is solidified, and it is difficult to select the perception mode (active perception and collaborative perception) according to the actual environment, which leads to inefficient resource allocation. In addition, the execution of the above resource allocation algorithm requires iterative calculation optimization, so it is difficult to adapt to the rapidly changing motion state of the perceived target, which in turn causes a time lag in resource allocation and affects the real-time performance of perception. Therefore, based on the idea of ​​dual-driven data and knowledge, a resource allocation method based on generative AI data-driven is designed in this section. According to the real-time perception requirements, the perception mode selection and computing resource allocation results can be calculated. Combined with the knowledge-driven method driven by the traditional optimization algorithm in the previous section, constraints on the data-driven method are implemented to ensure the accuracy of the data-driven method. Specifically, the generative AI includes a feature extraction module, a feature clustering module, a cross-domain invariant feature extraction module, and a domain feature module.

[0370] In an optional embodiment, the AI ​​resource optimization model is obtained by using the second communication perception related information and the second perception mode, and / or the second communication perception related information and the second resource allocation information for training, including:

[0371] Performing feature extraction on the second communication perception related information to obtain a first extracted feature;

[0372] Among them, the main purpose of the feature extraction module is to vector embed various resource constraints, communication perception (synaesthesia) indicator requirements, network topology and other time-value data in communication perception related information, and then splice and merge them to provide high-quality input features for subsequent generative AI models.

[0373] Specifically, the feature vector construction process:

[0374] Set different constraints and indicators in the system, including resource constraints, network topology, and synaesthesia indicator requirements. i The feature vector construction process is as follows:

[0375] Resource constraint vector (R i ):

[0376] Resource constraints may include power limits. Assuming there are M resource constraints, the resource constraint vector is:

[0377] R i =[R i1 ,R i2 ,…,R iM ]

[0378] Synaesthesia index vector (C i ):

[0379] The synaesthesia index includes the signal to interference and noise ratio (SINR). Assuming there are P synaesthesia indicators, the synaesthesia index vector is:

[0380] C i =[C i1 , C i2 , ..., C iP ]

[0381] Base station vector (B i ):

[0382] The base station vector describes the location of J base stations. Assuming there are Q base stations, the base station vector is

[0383] B i =[B i1 , B i2 , ..., B iQ ]

[0384] Communication user vector (U i ):

[0385] The communication user vector describes the positions of K communication users. Assuming there are R users, the communication user vector is:

[0386] U i =[U i1 , U i2 , ..., UiR ]

[0387] The perceived target vector (G i ):

[0388] The perception target vector describes the location of the I UAV perception target. Assuming there are S target features, the perception target vector is:

[0389] G i =[G i1 , G i2 , ..., G iS ]

[0390] Through these steps and formulas, we are able to construct the feature extraction module in detail, effectively vectorize and embed the data required by various resource constraints, network topology, and synaesthesia indicators, and provide reliable input features for the resource allocation algorithm of generative AI.

[0391] Vector Embedding:

[0392] Embed the above vector to get the corresponding embedding vector. Assuming the embedding function is φ, the embedding vector is expressed as

[0393] E R =φ R (R i ), E C =φ C (C i ), E B =φ B (B i ),

[0394] E U =φ U (U i ), E G =φ G (G i )

[0395] The embedding vectors of resource constraints, synaesthesia indicators, base stations, communication users, and perception targets are concatenated and merged to form the final feature vector (i.e., the first extracted feature):

[0396] F i =[E R ; E C ; E B ; E U ; E G ]

[0397] Among them, [;] represents the concatenation operation of vectors.

[0398] Performing feature clustering on the first extracted features to obtain multiple feature domains;

[0399] Among them, the goal of the feature clustering module is to aggregate similar cases to form a domain, so that the similarity of these cases in the feature space can be utilized to share feature information.

[0400] Commonly used clustering algorithms include K-means, Hierarchical Clustering, DBSCAN (Density-Based Spatial Clustering of Applications with Noise), etc. In this solution, the selection of a suitable clustering algorithm depends on the nature of the feature vector and the data scale.

[0401] K-means is a distance-based clustering algorithm that is suitable for most data sets. Its goal is to divide the data into K clusters so that the distance from each data point to the center of its cluster is minimized.

[0402] Hierarchical clustering does not require the number of clusters to be specified in advance. By building a tree-like hierarchy, the level of clustering can be selected as needed.

[0403] Assuming that the K-means clustering algorithm is selected, the specific process is as follows:

[0404] Feature matrix construction: The feature vectors F of all cases are i Composition feature matrix F:

[0405]

[0406] in, n is the number of cases, and d is the dimension of each feature vector.

[0407] Randomly select K initial cluster centers μ1, μ2, ..., μ K According to the distance between each eigenvector and the cluster center, the eigenvector is assigned to the nearest cluster center. Let the eigenvector F i To the cluster center μ k The distance is d(F i , μ k ), then the cluster label to which the i-th eigenvector belongs is: c i =argmin k d(F i , μ k ).

[0408] Calculate the new center of each cluster, which is the mean of all data points assigned to that cluster:

[0409]

[0410] Among them, C k is the set of the kth cluster, |C k | is the number of its data points. Iteration: Repeat the clustering step until the cluster center no longer changes or the maximum number of iterations is reached.

[0411] Through clustering, similar cases are aggregated into a feature domain, and the cases in each domain share feature information. Let the kth clustering domain be D k , the cases included are Then the feature vector of each domain is expressed as:

[0412]

[0413] Performing weighted processing on the extracted features within the same feature domain to obtain a first domain feature representation;

[0414] Specifically, the domain feature module further processes the features in each domain to adapt to the task requirements of the specific domain. Specifically, the attention mechanism and the fully connected layer are used for feature processing, and the cross-domain invariant features are combined for the final feature representation. The specific steps are as follows:

[0415] In-domain feature processing:

[0416] For the features in each domain, the attention mechanism is used for weighted processing, and then the domain feature representation (i.e., the first domain feature representation) is obtained through the fully connected layer.

[0417] Attention Mechanism:

[0418] Apply the attention mechanism to each case in each domain and calculate the attention weight. Let the input feature be F i , the attention weight is α i :

[0419]

[0420] Among them, W a is the parameter matrix of attention weights.

[0421] Weighted feature representation:

[0422] Apply the attention weights to the features to get the weighted feature representation:

[0423] F i，att =α i F i

[0424] Domain feature fusion:

[0425] All features in the domain processed by the attention mechanism are fused to obtain the domain feature representation:

[0426]

[0427] Fully connected layer:

[0428] The domain features are further processed through the fully connected layer to obtain the final in-domain feature representation (i.e., the first domain feature representation):

[0429] F d =W d F d +b d

[0430] Among them, W d and b d are the parameters of the fully connected layer.

[0431] Performing feature extraction and fusion on the extracted features in different feature domains to obtain a second domain feature representation;

[0432] Domain-invariant feature processing:

[0433] The goal of the cross-domain invariant feature extraction module is to extract consistent features from data in different domains so that these features can be shared and applied between different domains. This ensures that the generative AI model is consistent and robust when processing unseen data. The following are the specific steps and related formulas:

[0434] Feature extraction:

[0435] The feature vector of each domain is input into the cross-domain feature extraction model to extract the features that are invariant across domains. Assume that the input feature vector is F i , the cross-domain invariant feature is represented as F inv .

[0436] F i =[F Ri ; F Ti ; F Ci ; F Bi ; F Ui ; F Gi ]

[0437] Use deep neural network (DNN) to extract cross-domain invariant features. Let the cross-domain feature extraction network be ψ, then:

[0438] F inv =ψ(F i )

[0439] Here, it must be a multi-layer perceptron (MLP), consisting of multiple fully connected layers and activation functions:

[0440] F inv =ψ(F i ) = f L(W L (f L-1 (…f1(W1F i +b1)…)+b L-1 )+b L )

[0441] Among them, L represents the number of layers of the network, W l and b l are the weight matrix and bias vector of the lth layer, respectively, and f l is the activation function of the lth layer (such as ReLU, Sigmoid, etc.).

[0442] Combining all case features:

[0443] In order to obtain cross-domain invariant features, the features of all cases are fused. Assuming there are n cases, the fused cross-domain invariant feature representation (i.e., the second domain feature representation) is:

[0444]

[0445] This step ensures consistency between different domains and incorporates features from all cases.

[0446] Obtaining a target feature representation according to the first domain feature representation and the second domain feature representation;

[0447] Specifically, cross-domain feature combination:

[0448] The intra-domain feature representation and the cross-domain invariant features are concatenated or added to form the final feature representation (i.e., the target feature representation).

[0449] F final =[F d ; F inv ]

[0450] F final =F d +F inv

[0451] The AI ​​resource optimization model is obtained by training using the target feature representation and the second perception mode, and / or using the target feature representation and the second resource allocation information.

[0452] In an optional embodiment, training is performed using the target feature representation and the second resource allocation information to obtain the AI ​​resource optimization model, including:

[0453] Obtaining first training resource allocation information according to the target feature representation using a preset optimization method;

[0454] Among them, the preset optimization method is an existing traditional optimization algorithm, such as convex optimization, sequence approximation and other mathematical methods.

[0455] Using a preset initial AI model, obtaining second training resource allocation information according to the target feature representation;

[0456] Obtaining a loss function according to the first training resource allocation information and the second training resource allocation information;

[0457] The AI ​​resource optimization model is obtained by adjusting the model parameters in the initial AI model to reduce the loss function.

[0458] Specifically, the solution obtained by the traditional optimization algorithm is used as the label, and the mean square error (MSE) loss is used to optimize and solve the loss function.

[0459] Assume that the solution obtained by the traditional optimization algorithm (i.e., training resource allocation information) is R opt , the output of the generative AI model is R gen (i.e., the second training resource allocation information), then the loss function is:

[0460]

[0461] Use gradient descent or other optimization algorithms to update model parameters, minimize the loss function, and train the parameters of the generative AI model so that its output resource allocation results are as close as possible to the solution of the traditional optimization algorithm.

[0462] Through the above steps and formulas, the cross-domain invariant feature extraction module can effectively process and combine feature information from different domains, and the model can automatically select the perception mode according to the data characteristics, improving the accuracy and efficiency of the generative AI model in resource allocation. The generated model only needs to input data from the domain invariant feature module, and the model can decide the perception mode and resource allocation.

[0463] Combine the following Figure 6 , specifically describes the training process of the AI ​​resource optimization model provided by the embodiment of the present invention:

[0464] Feature extraction is performed on various resource constraints, communication perception (synaesthesia) index requirements, and network topology in the communication perception related information to obtain a feature vector (i.e., the first extracted feature). The feature vector is normalized and feature clustered, and an attention mechanism is introduced to obtain a first domain feature representation in the domain feature module, and a second domain feature representation is obtained in the cross-domain invariant feature module. A target feature representation is obtained based on the first domain feature representation and the second domain feature representation. The target feature representation is normalized, and the second training resource allocation information is obtained by a linear layer. The first training resource allocation information is obtained using a traditional optimization algorithm. A loss function is obtained based on the second training resource allocation information and the first training resource allocation information. The model parameters in the initial AI model are adjusted to reduce the loss function, and a direct correspondence between the communication resource related information and the training resource allocation information in the AI ​​resource optimization model is obtained.

[0465] It should be noted that the training process of the correspondence between communication resource related information and perception pattern in the AI ​​resource optimization model is basically consistent with the direct correspondence between the above-mentioned communication resource related information and training resource allocation information.

[0466] The embodiment of the present invention introduces generative AI, and the present invention realizes intelligent management and adaptive resource allocation of two perception modes. Generative AI can dynamically adjust the beam and power allocation strategy of the system according to environmental changes, the characteristics of the perception target, and the communication and perception QoS requirements. Generative AI first extracts in-domain features and out-of-domain features through training of historical data, and can learn and predict the perception and communication requirements in different scenarios in real time. The AI ​​model will select A send A receive or A send B receive mode according to the scenario, and automatically allocate beam direction and power resources. In different scenarios, AI autonomously determines which perception mode to use. When there are more communication tasks around the base station, AI will give priority to the A send B receive mode to reduce the burden of a single base station; when the accuracy requirement is high, the A send A receive mode is selected. AI ensures the best balance of communication and perception tasks under resource competition through optimization algorithms. Multi-objective optimization: When AI processes multiple perception targets, it combines the QoS requirements of different targets and optimizes through exponential utility functions to ensure that high-threat or high-priority targets obtain more resources, while low-priority targets maintain basic perception capabilities.

[0467] The present invention proposes a multi-base station beam and power allocation technology that adapts to different communication and perception QoS requirements. This solution is based on the base station deployment in the current cellular network. By introducing two perception modes, A transmit and A receive and A transmit and B receive, the generative AI is used to adaptively learn and optimize the environment, dynamically switch the perception mode according to the scene, and autonomously allocate beams and power resources according to different communication and perception requirements, so as to achieve efficient communication perception integration (ISAC) resource management. Specifically, firstly, multi-base station resource optimization schemes are designed for the active perception mode of A transmit and A receive and the active and passive collaborative perception mode of A transmit and B receive, and the condition information and resource allocation results of the two optimization schemes are used as training data for the generative AI model. Secondly, a generative AI model based on the transformer architecture is constructed. The model includes a domain-invariant feature extraction module and a cross-domain feature extraction module, which can effectively extract common feature information under the two perception modes and feature information under specific conditions, so as to achieve reliable resource allocation under unknown environmental conditions and accurate resource allocation under known environmental conditions.

[0468] Among them, the feature extraction module can effectively embed information such as various resource constraints, network topology, communication perception indicators (such as SINR), base station location, user location, and perception target location into a unified feature vector, ensuring the standardization and quantification of data in multi-base station collaborative communication and perception tasks. This module provides standardized input features for the generative AI model, ensuring the accuracy and consistency of resource allocation; the cross-domain invariant feature extraction module can identify and extract invariant features in different scenarios or domains, and embed these features into the feature representation of specific resource optimization scenarios to enhance the generalization and robustness of the model. This module ensures that the generative AI can maintain the stability and consistency of resource allocation decisions when the environment changes dynamically, and realize efficient communication and perception resource allocation in cross-domain scenarios; the feature fusion and connection module combines cross-domain invariant features with resource constraint features of specific scenarios through a deep neural network (DNN) to form a unified feature vector. Through multi-layer fully connected structure and activation function processing, the effective combination of multi-domain features is ensured, the accuracy of resource allocation of the system in complex and dynamic scenarios is improved, and the real-time resource scheduling optimization of communication and perception tasks is ensured; for the multi-base station A-transmit A-receive active perception system, the optimization problem modeling of differentiated QoS requirements is proposed. Resource allocation modeling is realized through exponential utility function to ensure that the perception accuracy and communication quality of different targets are given equal importance. This method can adapt to the changes in demand in different scenarios and is efficient and flexible.

[0469] like Figure 7 As shown, an embodiment of the present invention further provides a multi-base station collaborative sensing resource optimization device based on generative AI, comprising:

[0470] A first acquisition module 701 is used to acquire first communication perception related information of a first multi-base station coordinated communication perception integrated ISAC system;

[0471] A first processing module 702 is configured to input the first communication perception related information into a trained artificial intelligence (AI) resource optimization model to obtain first resource allocation information of a first multi-base station coordinated ISAC system and / or a first perception mode of a first multi-base station coordinated ISAC system output by the AI ​​resource optimization model;

[0472] The AI ​​resource optimization model is used to indicate at least one of the following:

[0473] The correspondence between communication perception related information and resource allocation information;

[0474] The correspondence between communication perception related information and perception patterns.

[0475] Optionally, the device further comprises:

[0476] A second acquisition module, used to acquire second communication perception related information, a second perception mode, and second resource allocation information of a second multi-base station coordinated ISAC system;

[0477] The second processing module is used to use the second communication perception related information and the second perception mode, and / or use the second communication perception related information and the second resource allocation information for training to obtain the AI ​​resource optimization model.

[0478] Optionally, the communication awareness related information includes at least one of the following:

[0479] Base station;

[0480] Communication users;

[0481] Perceived goals;

[0482] Resource constraint information;

[0483] Communication perception indicator information.

[0484] Optionally, the perception mode includes at least one of the following:

[0485] Active sensing mode;

[0486] Collaborative awareness model.

[0487] Optionally, the second acquisition module includes:

[0488] A first acquisition unit is used to acquire, in the active sensing mode, a downlink communication signal received by a communication user from a base station in the second multi-base station cooperative ISAC system, an echo signal reflected by a sensing target received by the base station, and beam power allocation information of the base station;

[0489] A first processing unit, configured to generate a first optimization target for resource allocation optimization in the active sensing mode according to an echo signal reflected by a sensing target received by the base station;

[0490] A second processing unit, configured to generate a first optimization constraint condition for performing resource allocation optimization in the active sensing mode according to a downlink communication signal received by the communication user from the base station and beam power allocation information of the base station;

[0491] The third processing unit is used to optimize the resource allocation of the second multi-base station collaborative ISAC system in the active sensing mode according to the first optimization target and the first optimization constraint, so as to obtain the second resource allocation information in the active sensing mode.

[0492] Optionally, the first processing unit is specifically configured to:

[0493] Performing separation and filtering processing on the echo signal reflected by the sensing target received by the base station to obtain a first filtered echo signal;

[0494] generating a Cramer-Rao bound CRLB of the perception performance of the perception target according to the first filtered echo signal;

[0495] Obtaining a first global positioning performance of the perception target according to the CRLB of the perception target, the echo arrival azimuth AoA of the perception target relative to the base station, and the echo arrival elevation angle ZoA of the perception target relative to the base station;

[0496] The first optimization objective is generated according to the first global positioning performance.

[0497] Optionally, the second processing unit is specifically configured to:

[0498] Obtaining a first signal-to-noise ratio (SINR) received by the communication user according to a downlink communication signal received by the communication user from a base station;

[0499] Taking the first SINR as greater than or equal to the first preset SINR threshold as a first condition;

[0500] Obtaining the transmit power of the base station and the number of beams supported by the base station according to the power allocation information of the base station and the beam power allocation information of the base station;

[0501] Taking the transmit power of the base station as less than or equal to the first preset power threshold as the second condition;

[0502] The number of beams supported by the base station is less than or equal to a first preset number as a third condition;

[0503] Generate a fourth condition according to a preset beam selection matrix;

[0504] The first optimization constraint condition is obtained according to the first condition, the second condition, the third condition and the fourth condition.

[0505] Optionally, the third processing unit is specifically configured to:

[0506] Performing an approximate convex relaxation process on the fourth condition in the first optimization constraint condition to obtain a fifth condition;

[0507] Decomposing the first optimization objective into a first sub-objective and a second sub-objective by using an initial beam selection matrix;

[0508] Obtaining, according to the initial beam selection matrix, numerator information and denominator information in a first SINR received by the communication user;

[0509] Converting the first condition in the first optimization constraint condition into a sixth condition according to the numerator information and the denominator information;

[0510] Generate a first optimization problem according to the first sub-goal, the sixth condition and the second condition in the first optimization constraint condition;

[0511] Generate a second optimization problem according to the second sub-objective, the sixth condition, and the third condition and the fifth condition in the first optimization constraint condition;

[0512] By alternately solving the first optimization problem and the second optimization problem, second resource allocation information in the active sensing mode is obtained.

[0513] Optionally, the second acquisition module includes:

[0514] A second acquisition unit is used to acquire, in the cooperative sensing mode, a transmission signal of a first base station in the second multi-base station cooperative ISAC system, a reception signal of a first communication user communicating with the first base station, and a reflection signal received by the first base station;

[0515] a fourth processing unit, configured to obtain a second SINR of the first communication user according to a transmission signal of the first base station and a reception signal of the first communication user;

[0516] a fifth processing unit, configured to obtain a third SINR of an output signal of a radar for sensing a target at the first base station according to the reflected signal received by the first base station;

[0517] A sixth processing unit, configured to generate a second optimization target according to the second SINR;

[0518] a seventh processing unit, configured to generate a second optimization constraint condition for performing resource allocation optimization in the collaborative sensing mode according to the third SINR and the power budget information of the first base station;

[0519] An eighth processing unit is used to optimize resource allocation of the second multi-base station cooperative ISAC system in the cooperative sensing mode according to the second optimization objective and the second optimization constraint condition, and obtain second resource allocation information in the cooperative sensing mode.

[0520] Optionally, the seventh processing unit is specifically configured to:

[0521] The third SINR is greater than or equal to the second preset SINR threshold as a seventh condition;

[0522] The transmit power of the first base station is less than or equal to the second preset power threshold indicated by the power budget information of the first base station as an eighth condition;

[0523] Generate the ninth condition according to the preset target scheduling matrix;

[0524] According to the seventh condition, the eighth condition and the ninth condition, the second optimization constraint condition is obtained.

[0525] Optionally, the eighth processing unit is specifically configured to:

[0526] According to the transmit beamforming matrix and the target scheduling matrix, converting the second optimization objective into a third optimization problem;

[0527] Using generalized Rayleigh entropy, solving the third optimization problem to obtain an optimized receive beamforming matrix;

[0528] Using the Dinkelbach method, the second optimization objective is transformed into a third sub-objective;

[0529] According to the third sub-objective and the sixth and seventh conditions in the second optimization constraint condition, a fourth optimization problem is obtained;

[0530] Using the Lagrange dual method, the fourth optimization problem is solved to obtain an optimized transmit beamforming matrix;

[0531] According to the optimized receive beamforming matrix, the optimized transmit beamforming matrix and the second optimization target, the target scheduling matrix is ​​optimized to obtain the second resource allocation information in the cooperative sensing mode.

[0532] Optionally, the second processing module includes:

[0533] A ninth processing unit, configured to perform feature extraction on the second communication perception related information to obtain a first extracted feature;

[0534] a tenth processing unit, configured to perform feature clustering on the first extracted features to obtain a plurality of feature domains;

[0535] an eleventh processing unit, configured to perform weighted processing on the extracted features within the same feature domain to obtain a first domain feature representation;

[0536] a twelfth processing unit, configured to extract and fuse the extracted features in different feature domains to obtain a second domain feature representation;

[0537] a thirteenth processing unit, configured to obtain a target feature representation according to the first domain feature representation and the second domain feature representation;

[0538] A fourteenth processing unit is used to obtain the AI ​​resource optimization model by using the target feature representation and the second perception mode, and / or by using the target feature representation and the second resource allocation information for training.

[0539] Optionally, the fourteenth processing unit is specifically configured to:

[0540] Obtaining first training resource allocation information according to the target feature representation using a preset optimization method;

[0541] Using a preset initial AI model, obtaining second training resource allocation information according to the target feature representation;

[0542] Obtaining a loss function according to the first training resource allocation information and the second training resource allocation information;

[0543] The AI ​​resource optimization model is obtained by adjusting the model parameters in the initial AI model to reduce the loss function.

[0544] It should be noted that the multi-base station collaborative sensing resource optimization device of the generative AI provided in the embodiment of the present invention is a device capable of executing the above-mentioned multi-base station collaborative sensing resource optimization method of the generative AI. Then all embodiments of the above-mentioned multi-base station collaborative sensing resource optimization method of the generative AI are applicable to the device and can achieve the same or similar technical effects.

[0545] like Figure 8 As shown, an embodiment of the present invention also provides a generative AI multi-base station collaborative perception resource optimization device, including: a processor 801; and a memory 803 connected to the processor 801 through a bus interface 802, the memory 803 is used to store the programs and data used by the processor 801 when performing operations, and the processor 801 calls and executes the programs and data stored in the memory 803.

[0546] The transceiver 804 is connected to the bus interface 802 and is used to receive and send data under the control of the processor 801. Specifically, the processor 801 is used to read the program in the memory 803 and execute the following process:

[0547] Acquire first communication perception related information of a first multi-base station coordinated communication perception integrated ISAC system;

[0548] Inputting the first communication perception related information into a trained artificial intelligence (AI) resource optimization model to obtain first resource allocation information of a first multi-base station collaborative ISAC system and / or a first perception mode of a first multi-base station collaborative ISAC system output by the AI ​​resource optimization model;

[0549] The AI ​​resource optimization model is used to indicate at least one of the following:

[0550] The correspondence between communication perception related information and resource allocation information;

[0551] The correspondence between communication perception related information and perception patterns.

[0552] Optionally, the processor 801 is further configured to:

[0553] Acquire second communication perception related information, a second perception mode, and second resource allocation information of a second multi-base station coordinated ISAC system;

[0554] The AI ​​resource optimization model is obtained by training using the second communication perception related information and the second perception mode, and / or using the second communication perception related information and the second resource allocation information.

[0555] Optionally, the communication awareness related information includes at least one of the following:

[0556] Base station;

[0557] Communication users;

[0558] Perceived goals;

[0559] Resource constraint information;

[0560] Communication perception indicator information.

[0561] Optionally, the perception mode includes at least one of the following:

[0562] Active sensing mode;

[0563] Collaborative awareness model.

[0564] Optionally, the processor 801 is specifically configured to:

[0565] Acquire, in the active sensing mode, a downlink communication signal received by a communication user from a base station in the second multi-base station collaborative ISAC system, an echo signal reflected by a sensing target received by the base station, and beam power allocation information of the base station;

[0566] Generating a first optimization target for resource allocation optimization in the active sensing mode according to an echo signal reflected by the sensing target received by the base station;

[0567] Generate a first optimization constraint condition for resource allocation optimization in the active sensing mode according to a downlink communication signal received by the communication user from the base station and beam power allocation information of the base station;

[0568] According to the first optimization target and the first optimization constraint, resource allocation optimization is performed on the second multi-base station cooperative ISAC system in the active sensing mode to obtain second resource allocation information in the active sensing mode.

[0569] Optionally, the processor 801 is specifically configured to:

[0570] Performing separation and filtering processing on the echo signal reflected by the sensing target received by the base station to obtain a first filtered echo signal;

[0571] generating a Cramer-Rao bound CRLB of the perception performance of the perception target according to the first filtered echo signal;

[0572] Obtaining a first global positioning performance of the perception target according to the CRLB of the perception target, the echo arrival azimuth AoA of the perception target relative to the base station, and the echo arrival elevation angle ZoA of the perception target relative to the base station;

[0573] The first optimization objective is generated according to the first global positioning performance.

[0574] Optionally, the processor 801 is specifically configured to:

[0575] Obtaining a first signal-to-noise ratio (SINR) received by the communication user according to a downlink communication signal received by the communication user from a base station;

[0576] Taking the first SINR as greater than or equal to the first preset SINR threshold as a first condition;

[0577] Obtaining the transmit power of the base station and the number of beams supported by the base station according to the power allocation information of the base station and the beam power allocation information of the base station;

[0578] Taking the transmit power of the base station as less than or equal to the first preset power threshold as the second condition;

[0579] The number of beams supported by the base station is less than or equal to a first preset number as a third condition;

[0580] Generate a fourth condition according to a preset beam selection matrix;

[0581] The first optimization constraint condition is obtained according to the first condition, the second condition, the third condition and the fourth condition.

[0582] Optionally, the processor 801 is specifically configured to:

[0583] Performing an approximate convex relaxation process on the fourth condition in the first optimization constraint condition to obtain a fifth condition;

[0584] Decomposing the first optimization objective into a first sub-objective and a second sub-objective by using an initial beam selection matrix;

[0585] Obtaining, according to the initial beam selection matrix, numerator information and denominator information in a first SINR received by the communication user;

[0586] Converting the first condition in the first optimization constraint condition into a sixth condition according to the numerator information and the denominator information;

[0587] Generate a first optimization problem according to the first sub-goal, the sixth condition and the second condition in the first optimization constraint condition;

[0588] Generate a second optimization problem according to the second sub-objective, the sixth condition, and the third condition and the fifth condition in the first optimization constraint condition;

[0589] By alternately solving the first optimization problem and the second optimization problem, second resource allocation information in the active sensing mode is obtained.

[0590] Optionally, the processor 801 is specifically configured to:

[0591] Acquire, in the cooperative sensing mode, a transmission signal of a first base station in the second multi-base station cooperative ISAC system, a reception signal of a first communication user communicating with the first base station, and a reflection signal received by the first base station;

[0592] Obtaining a second SINR of the first communication user according to a transmission signal of the first base station and a reception signal of the first communication user;

[0593] Obtaining, according to the reflected signal received by the first base station, a third SINR of an output signal of a radar for sensing a target at the first base station;

[0594] Generating a second optimization target according to the second SINR;

[0595] Generating a second optimization constraint condition for resource allocation optimization in the collaborative sensing mode according to the third SINR and the power budget information of the first base station;

[0596] According to the second optimization objective and the second optimization constraint, resource allocation optimization is performed on the second multi-base station cooperative ISAC system in the cooperative sensing mode to obtain second resource allocation information in the cooperative sensing mode.

[0597] Optionally, the processor 801 is specifically configured to:

[0598] The third SINR is greater than or equal to the second preset SINR threshold as a seventh condition;

[0599] The transmit power of the first base station is less than or equal to the second preset power threshold indicated by the power budget information of the first base station as an eighth condition;

[0600] Generate the ninth condition according to the preset target scheduling matrix;

[0601] According to the seventh condition, the eighth condition and the ninth condition, the second optimization constraint condition is obtained.

[0602] Optionally, the processor 801 is specifically configured to:

[0603] According to the transmit beamforming matrix and the target scheduling matrix, converting the second optimization objective into a third optimization problem;

[0604] Using generalized Rayleigh entropy, solving the third optimization problem to obtain an optimized receive beamforming matrix;

[0605] Using the Dinkelbach method, the second optimization objective is transformed into a third sub-objective;

[0606] According to the third sub-objective and the sixth and seventh conditions in the second optimization constraint condition, a fourth optimization problem is obtained;

[0607] Using the Lagrange dual method, the fourth optimization problem is solved to obtain an optimized transmit beamforming matrix;

[0608] According to the optimized receive beamforming matrix, the optimized transmit beamforming matrix and the second optimization target, the target scheduling matrix is ​​optimized to obtain the second resource allocation information in the cooperative sensing mode.

[0609] Optionally, the processor 801 is specifically configured to:

[0610] Performing feature extraction on the second communication perception related information to obtain a first extracted feature;

[0611] Performing feature clustering on the first extracted features to obtain multiple feature domains;

[0612] Performing weighted processing on the extracted features within the same feature domain to obtain a first domain feature representation;

[0613] Performing feature extraction and fusion on the extracted features in different feature domains to obtain a second domain feature representation;

[0614] Obtaining a target feature representation according to the first domain feature representation and the second domain feature representation;

[0615] The AI ​​resource optimization model is obtained by training using the target feature representation and the second perception mode, and / or using the target feature representation and the second resource allocation information.

[0616] Optionally, the processor 801 is specifically configured to:

[0617] Obtaining first training resource allocation information according to the target feature representation using a preset optimization method;

[0618] Using a preset initial AI model, obtaining second training resource allocation information according to the target feature representation;

[0619] Obtaining a loss function according to the first training resource allocation information and the second training resource allocation information;

[0620] The AI ​​resource optimization model is obtained by adjusting the model parameters in the initial AI model to reduce the loss function.

[0621] Among them, Figure 8In the embodiment, the bus architecture may include any number of interconnected buses and bridges, specifically linking various circuits of one or more processors represented by processor 801 and memory represented by memory 803. The bus architecture may also link various other circuits such as peripherals, voltage regulators, and power management circuits, which are well known in the art and are therefore not further described herein. The bus interface provides a user interface 805. The transceiver 804 may be a plurality of components, i.e., including a transmitter and a receiver, providing a unit for communicating with various other devices on a transmission medium. The processor 801 is responsible for managing the bus architecture and general processing, and the memory 803 may store data used by the processor 801 when performing operations.

[0622] In addition, a specific embodiment of the present invention also provides a readable storage medium on which a computer program is stored, wherein when the program is executed by a processor, the steps in the multi-base station collaborative sensing resource optimization method of the generative AI as described in any one of the above are implemented.

[0623] In the several embodiments provided in the present application, it should be understood that the disclosed methods and devices can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0624] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may be physically included separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of hardware plus software functional units.

[0625] The above-mentioned integrated unit implemented in the form of a software functional unit can be stored in a computer-readable storage medium. The above-mentioned software functional unit is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform some steps of the resource selection method described in each embodiment of the present invention, or to perform some steps of the information sending method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (Read-Only Memory, referred to as ROM), random access memory (Random Access Memory, referred to as RAM), disk or optical disk and other media that can store program codes.

[0626] A specific embodiment of the present invention further provides a computer program product, including computer instructions, which, when executed by a processor, implement the above Figure 1 The various processes of the method embodiment shown can achieve the same technical effect, and will not be described again here to avoid repetition.

[0627] The above is a preferred embodiment of the present invention. It should be pointed out that for ordinary personnel in this technical field, several improvements and modifications can be made without departing from the principles described in the present invention. These improvements and modifications are also within the scope of protection of the present invention.

Claims

1. A multi-base station collaborative sensing resource optimization method based on generative AI, characterized in that: include: Acquire first communication perception related information of a first multi-base station coordinated communication perception integrated ISAC system; Inputting the first communication perception related information into a trained artificial intelligence (AI) resource optimization model to obtain first resource allocation information of a first multi-base station collaborative ISAC system and / or a first perception mode of a first multi-base station collaborative ISAC system output by the AI ​​resource optimization model; The AI ​​resource optimization model is used to indicate at least one of the following: The correspondence between communication perception related information and resource allocation information; The correspondence between communication perception related information and perception patterns.

2. The method according to claim 1, characterized in that The method further comprises: Acquire second communication perception related information, a second perception mode, and second resource allocation information of a second multi-base station coordinated ISAC system; The AI ​​resource optimization model is obtained by training using the second communication perception related information and the second perception mode, and / or using the second communication perception related information and the second resource allocation information.

3. The method according to claim 1 or 2, characterized in that: The communication awareness related information includes at least one of the following: Base station; Communication users; Perceived goals; Resource constraint information; Communication perception indicator information.

4. The method according to claim 1 or 2, characterized in that: The sensing mode includes at least one of the following: Active sensing mode; Collaborative awareness model.

5. The method according to claim 2, characterized in that: Acquiring second resource allocation information of a second multi-base station coordinated ISAC system includes: Acquire, in the active sensing mode, a downlink communication signal received by a communication user from a base station in the second multi-base station collaborative ISAC system, an echo signal reflected by a sensing target received by the base station, and beam power allocation information of the base station; Generating a first optimization target for resource allocation optimization in the active sensing mode according to an echo signal reflected by the sensing target received by the base station; Generate a first optimization constraint condition for resource allocation optimization in the active sensing mode according to a downlink communication signal received by the communication user from the base station and beam power allocation information of the base station; According to the first optimization target and the first optimization constraint, resource allocation optimization is performed on the second multi-base station cooperative ISAC system in the active sensing mode to obtain second resource allocation information in the active sensing mode.

6. The method according to claim 5, characterized in that Generating a first optimization target for resource allocation optimization in the active sensing mode according to an echo signal reflected by the sensing target received by the base station includes: Performing separation and filtering processing on the echo signal reflected by the sensing target received by the base station to obtain a first filtered echo signal; generating a Cramer-Rao bound CRLB of the perception performance of the perception target according to the first filtered echo signal; Obtaining a first global positioning performance of the perception target according to the CRLB of the perception target, the echo arrival azimuth AoA of the perception target relative to the base station, and the echo arrival elevation angle ZoA of the perception target relative to the base station; The first optimization objective is generated according to the first global positioning performance.

7. The method according to claim 5, characterized in that Generating a first optimization constraint condition for resource allocation optimization in the active sensing mode according to a downlink communication signal received by the communication user from the base station and beam power allocation information of the base station, including: Obtaining a first signal-to-noise ratio (SINR) received by the communication user according to a downlink communication signal received by the communication user from a base station; Taking the first SINR as greater than or equal to the first preset SINR threshold as a first condition; Obtaining the transmit power of the base station and the number of beams supported by the base station according to the power allocation information of the base station and the beam power allocation information of the base station; Taking the transmit power of the base station as less than or equal to the first preset power threshold as the second condition; The number of beams supported by the base station is less than or equal to a first preset number as a third condition; Generate a fourth condition according to a preset beam selection matrix; The first optimization constraint condition is obtained according to the first condition, the second condition, the third condition and the fourth condition.

8. The method according to claim 5, characterized in that According to the first optimization objective and the first optimization constraint condition, resource allocation optimization is performed on the second multi-base station cooperative ISAC system in the active sensing mode to obtain second resource allocation information in the active sensing mode, including: Performing an approximate convex relaxation process on the fourth condition in the first optimization constraint condition to obtain a fifth condition; Decomposing the first optimization objective into a first sub-objective and a second sub-objective by using an initial beam selection matrix; Obtaining, according to the initial beam selection matrix, numerator information and denominator information in a first SINR received by the communication user; Converting a first condition in the first optimization constraint condition into a sixth condition according to the numerator information and the denominator information; Generate a first optimization problem according to the first sub-goal, the sixth condition and the second condition in the first optimization constraint condition; Generate a second optimization problem according to the second sub-objective, the sixth condition, and the third condition and the fifth condition in the first optimization constraint condition; By alternately solving the first optimization problem and the second optimization problem, second resource allocation information in the active sensing mode is obtained.

9. The method according to claim 2, characterized in that: Acquiring second resource allocation information of a second multi-base station coordinated ISAC system includes: Acquire, in the cooperative sensing mode, a transmission signal of a first base station in the second multi-base station cooperative ISAC system, a reception signal of a first communication user communicating with the first base station, and a reflection signal received by the first base station; Obtaining a second SINR of the first communication user according to a transmission signal of the first base station and a reception signal of the first communication user; Obtaining, according to the reflected signal received by the first base station, a third SINR of an output signal of a radar for sensing a target at the first base station; Generating a second optimization target according to the second SINR; Generating a second optimization constraint condition for resource allocation optimization in the collaborative sensing mode according to the third SINR and the power budget information of the first base station; According to the second optimization objective and the second optimization constraint, resource allocation optimization is performed on the second multi-base station cooperative ISAC system in the cooperative sensing mode to obtain second resource allocation information in the cooperative sensing mode.

10. The method according to claim 9, characterized in that Generating, according to the third SINR and the power budget information of the first base station, a second optimization constraint condition for performing resource allocation optimization in the collaborative sensing mode, including: The third SINR is greater than or equal to the second preset SINR threshold as a seventh condition; The transmit power of the first base station is less than or equal to the second preset power threshold indicated by the power budget information of the first base station as an eighth condition; Generate the ninth condition according to the preset target scheduling matrix; According to the seventh condition, the eighth condition and the ninth condition, the second optimization constraint condition is obtained.

11. The method according to claim 10, characterized in that According to the second optimization objective and the second optimization constraint condition, performing resource allocation optimization on the second multi-base station cooperative ISAC system in the cooperative sensing mode to obtain second resource allocation information in the cooperative sensing mode includes: According to the transmit beamforming matrix and the target scheduling matrix, converting the second optimization objective into a third optimization problem; Using generalized Rayleigh entropy, solving the third optimization problem to obtain an optimized receive beamforming matrix; Using the Dinkelbach method, the second optimization objective is transformed into a third sub-objective; According to the third sub-objective and the sixth and seventh conditions in the second optimization constraint condition, a fourth optimization problem is obtained; Using the Lagrange dual method, the fourth optimization problem is solved to obtain an optimized transmit beamforming matrix; According to the optimized receive beamforming matrix, the optimized transmit beamforming matrix and the second optimization target, the target scheduling matrix is ​​optimized to obtain the second resource allocation information in the cooperative sensing mode.

12. The method according to claim 2, characterized in that: Using the second communication perception related information and the second perception mode, and / or using the second communication perception related information and the second resource allocation information for training to obtain the AI ​​resource optimization model, includes: Performing feature extraction on the second communication perception related information to obtain a first extracted feature; Performing feature clustering on the first extracted features to obtain multiple feature domains; Performing weighted processing on the extracted features within the same feature domain to obtain a first domain feature representation; Performing feature extraction and fusion on the extracted features in different feature domains to obtain a second domain feature representation; Obtaining a target feature representation according to the first domain feature representation and the second domain feature representation; The AI ​​resource optimization model is obtained by training using the target feature representation and the second perception mode, and / or using the target feature representation and the second resource allocation information.

13. The method according to claim 12, characterized in that The target feature representation and the second resource allocation information are used for training to obtain the AI ​​resource optimization model, including: Obtaining first training resource allocation information according to the target feature representation using a preset optimization method; Using a preset initial AI model, obtaining second training resource allocation information according to the target feature representation; Obtaining a loss function according to the first training resource allocation information and the second training resource allocation information; The AI ​​resource optimization model is obtained by adjusting the model parameters in the initial AI model to reduce the loss function.

14. A multi-base station collaborative sensing resource optimization device based on generative AI, characterized in that: include: A first acquisition module, configured to acquire first communication perception related information of a first multi-base station coordinated communication perception integrated ISAC system; A first processing module, configured to input the first communication perception related information into a trained artificial intelligence (AI) resource optimization model, and obtain first resource allocation information of a first multi-base station collaborative ISAC system and / or a first perception mode of a first multi-base station collaborative ISAC system output by the AI ​​resource optimization model; The AI ​​resource optimization model is used to indicate at least one of the following: The correspondence between communication perception related information and resource allocation information; The correspondence between communication perception related information and perception patterns.

15. A multi-base station collaborative sensing resource optimization device based on generative AI, characterized in that: include: A processor, a memory, and a program stored in the memory and executable on the processor, wherein when the program is executed by the processor, the steps in the generative AI-based multi-base station collaborative sensing resource optimization method are implemented as described in any one of claims 1 to 13.

16. A readable storage medium, characterized in that: The readable storage medium stores a program, and when the program is executed by the processor, the steps in the generative AI-based multi-base station collaborative sensing resource optimization method as described in any one of claims 1 to 13 are implemented.

17. A computer program product, characterized in that It includes computer instructions, which, when executed by a processor, implement the steps in the multi-base station collaborative sensing resource optimization method based on generative AI as described in any one of claims 1 to 13.

Images