Wireless communication transmission method based on communication perception fusion
By improving the Gaussian mixture model and deep reinforcement learning methods, joint modeling and processing of communication users and sensing targets in multi-user millimeter-wave systems were achieved. This solved the problem of difficulty in balancing communication and sensing performance in existing technologies, improved spectrum utilization and system capacity, and enhanced beam pointing accuracy and link stability.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-18
- Publication Date
- 2026-03-13
AI Technical Summary
In existing multi-user millimeter-wave systems, the communication link and sensing link lack unified modeling and collaborative design in terms of beam design, power allocation and spectrum resource utilization, which makes it difficult to simultaneously achieve both communication performance and sensing performance, and results in insufficient joint clustering of users and targets and joint allocation of power and spectrum.
By establishing a set of communication users and a set of sensing targets, and employing an improved Gaussian mixture model clustering algorithm and a deep reinforcement learning policy network, joint modeling and processing of communication users and sensing targets are achieved, enabling unified clustering and power allocation, and optimizing beam design and spectrum resource allocation.
It achieves synergistic improvement in communication and sensing performance, increases spectrum utilization and system capacity, enhances beam energy concentration and sensing echo quality, and improves beam pointing accuracy and link stability in high-speed or complex scenarios.
Smart Images
Figure CN121665259A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of wireless communication technology, and in particular to a wireless communication transmission method based on communication-aware fusion. Background Technology
[0002] With the development of millimeter wave and massive MIMO technologies, integrated sensing and communication (ISAC) systems for 6G have gradually become a research hotspot. In existing multi-user millimeter wave systems, downlink transmission is usually carried out using orthogonal multiple access or simple NOMA. Sensing functions are mostly implemented independently in radar mode or passive sensing form. Communication links and sensing links are mostly optimized separately in beam design, power allocation and spectrum resource utilization, lacking unified modeling and collaborative design, making it difficult to make full use of space and spectrum resources.
[0003] Furthermore, in existing NOMA-ISAC schemes, user clustering is often based solely on communication channel gain or geometric location, without performing unified clustering modeling of sensing targets and communication users. This makes it difficult to simultaneously achieve both communication and sensing performance within the same beam. Power allocation is generally designed separately from the serial interference cancellation decoding sequence, failing to fully consider the joint constraints of multi-user interference and sensing accuracy. Therefore, existing technologies still have shortcomings in multi-user millimeter-wave NOMA-ISAC scenarios, particularly in joint clustering of users and targets, joint allocation of power and spectrum, and beam tracking and adaptive parameter updates. Summary of the Invention
[0004] To address the technical problems existing in the background art, this invention proposes a wireless communication transmission method based on communication-aware fusion.
[0005] The present invention proposes a wireless communication transmission method based on communication-aware fusion, comprising the following steps: S1. Establish the communication user set and the sensing target set, and set the carrier frequency, system bandwidth, time slot length, transmit power constraints and spectrum resource constraints; S2. In each time slot, the communication user sends an uplink pilot sequence, the base station sends a sensing and detection signal and receives the uplink pilot sequence and the echo signal vector generated by the sensing and detection signal, and obtains the channel state information set as well as the initial position information and motion information set. S3. Based on the channel state information set, the initial position information, and the motion information set, generate the first feature vector of the communication user and the second feature vector of the perceived target, and form a sample feature set; S4. Based on the improved Gaussian mixture model clustering algorithm, the sample feature set is divided into multiple clusters; S5. For each communication user in the cluster, determine the serial interference cancellation decoding order of power domain non-orthogonal multiple access based on the channel gain of the communication user in the cluster, and allocate power coefficients and spectrum resource units to the communication user in the cluster. S6. For each cluster, use a deep reinforcement learning policy network to obtain the cluster's beam direction and beamwidth; S7. The base station generates and transmits integrated communication sensing transmission signals for each cluster based on channel state information, power coefficient, spectrum resource unit, beam direction and beamwidth. S8. The base station receives the uplink feedback signal and the environmental echo signal vector, and updates the channel state information set, as well as the initial position information and motion information set.
[0006] Preferably, a set of communication users and a set of sensing targets are established, and constraints on carrier frequency, system bandwidth, time slot length, transmit power, and spectrum resources are set as follows: A millimeter-wave multi-antenna transmitting array and a millimeter-wave multi-antenna receiving array are configured on the base station side. The number of antenna elements in the millimeter-wave multi-antenna transmitting array is denoted as . The number of antenna elements in a millimeter-wave multi-antenna receiver array is denoted as ; Establish communication user set With the set of perceived targets ,in, For the first One communication user, For the first One perceived target; Set carrier frequency System bandwidth Time slot length transmit power constraint The total transmit power of the base station transmitting signals to all communication users and sensing targets in any time slot shall not exceed the transmit power constraint. ; Set spectrum resource constraints and establish a set of spectrum resource units ,in, For the first Each spectrum resource unit, the carrier frequency System bandwidth Time slot length Transmit power constraints Both spectrum resource constraints are set based on expert experience.
[0007] Preferably, in each time slot, the communication user sends an uplink pilot sequence, the base station sends a sensing and detection signal and receives the uplink pilot sequence and the echo signal vector generated by the sensing and detection signal, thus obtaining a set of channel state information, initial position information, and motion information, as follows: With a time slot length of Under the time slot structure, indexed by time slot. Mark the first In the time slot, at the time slot At the start of each time slot, from the set of communication users For each communication user The allocation length is uplink pilot sequence ; At the beginning of each time slot, each communication user transmits its corresponding uplink pilot sequence through its respective uplink channel. The base station-side receiving array receives uplink pilot signals from all communication users, forming a received pilot signal matrix. : ; in, , For communication users In the Uplink channel vector for each time slot, This refers to the number of antenna elements in a millimeter-wave multi-antenna receiver array. For the received noise matrix; The base station receives the pilot signal matrix. and each uplink pilot sequence Calculate the channel estimation vector for each communication user And constitute a set of channel state information. ; The base station transmits sensing and detection signals according to the transmit power constraint and receives echoes from the set of sensing targets, forming a received echo signal vector. Matched filtering and time-frequency processing are then performed on the received echo signal vector to obtain the time delay estimate for each sensing target. Doppler frequency shift estimation and initial azimuth angle The initial distance to the target is calculated based on time delay estimation and Doppler frequency shift estimation. and initial radial velocity : ; ; in, At the speed of light, For the operating wavelength, , For carrier frequency; The initial distance, initial azimuth angle, and initial radial velocity of each sensed target are used to construct the initial position and motion information of the sensed target. ; The initial position information and motion information of all sensed targets are combined into an initial position information and motion information set. .
[0008] Preferably, a first feature vector of the communication user and a second feature vector of the perceived target are generated based on the channel state information set, the initial position information, and the motion information set, and then a sample feature set is formed, as follows: Obtain the location coordinates of the communication users in each time slot And calculate the distance between the communication user and the base station. And the azimuth angle of the communication user relative to the base station : ; ; Based on the channel estimation vectors of each communication user in the channel state information set Obtaining channel gain : ; Obtaining communication user service requirements The first feature vector of a communication user is formed by combining the user's location coordinates, distance from the base station, azimuth angle relative to the base station, channel gain, and service requirements. ; Based on the initial position and motion information set, for each sensed target in each time slot, according to the initial distance... and initial azimuth angle Calculate the position coordinates of the perceived target : ; Assign importance weights to perceived targets The importance weights are set based on expert experience, and the position coordinates, initial distance, initial azimuth angle, initial radial velocity, and importance weights of the perceived target are combined to form the second feature vector of the perceived target: ; The sample feature set is composed of the first feature vectors of all communication users and the second feature vectors of all perceived targets. .
[0009] Preferably, the sample feature set is divided into multiple clusters based on the improved Gaussian mixture model clustering algorithm, as follows: Rearrange the sample feature set into a sample vector sequence. , And set the number of clusters The mixing weights of the Gaussian components are denoted as... The mean vector is denoted as The covariance matrix is denoted as , ; From each sample in the sample vector sequence Extract the corresponding azimuth angle And obtain the direction weights : ; in, , These are the weighting coefficients; In each iteration, for each sample and each Gaussian component The orientation penalty term is defined based on the cluster orientation angle: ; in, For the first During the nth iteration Cluster orientation angles corresponding to each Gaussian component; And calculate the sample Belongs to the The posterior probability of each Gaussian component : ; in, The penalty coefficient is... , It is a multidimensional Gaussian probability density function. It is a natural exponential function; In each iteration, based on the posterior probability Update Mixed Weights Mean vector Covariance Matrix : ; ; ; And update the cluster orientation angle based on the positional components of the mean vector: ; When the maximum number of iterations is reached Stop the iteration and obtain the cluster label for each sample. : ; The sample feature set is divided into multiple clusters based on the cluster label of each sample, and each cluster corresponds to a cluster index. .
[0010] Preferably, for each communication user within a cluster, the serial interference cancellation decoding order for power domain non-orthogonal multiple access is determined based on the channel gain of the communication user within the cluster, and power coefficients and spectrum resource units are allocated to the communication user within the cluster, as follows: The communication user set and the sensing target set are divided according to the cluster label of each sample to obtain the intra-cluster communication user set and the intra-cluster sensing target set; For each communication user in the cluster's communication user set, the serial interference cancellation decoding order of the power domain non-orthogonal multiple access used by the cluster in each time slot is determined in ascending order of channel gain; Under transmit power constraints Next, a cluster transmit power is allocated to each cluster, and the cluster transmit power is distributed among the communication users within the cluster according to a preset rule. Communication users with smaller channel gain receive a larger power coefficient, and communication users with larger channel gain receive a smaller power coefficient, thereby determining the actual transmit power of each communication user within the cluster. Based on the service requirements of each communication user, a minimum received signal-to-noise ratio (SNR) threshold is determined for each communication user within the cluster. Under the determined power coefficient and serial interference cancellation decoding order, the received SNR of each communication user within the cluster is constrained to be no less than the minimum received SNR threshold. Based on the set of spectrum resource units, a set of spectrum resource units is selected for each cluster and allocated to all communication users within the cluster for power domain non-orthogonal multiple access transmission. The spectrum resource units allocated between different clusters do not overlap.
[0011] Preferably, the sum of all power coefficients of all users communicating within the cluster is one.
[0012] Preferably, for each cluster, a deep reinforcement learning policy network is used to obtain the cluster's beam direction and beamwidth, as follows: Within each time slot, for each cluster, the beam direction, beamwidth, received signal-to-noise ratio of the communication users within the cluster, and echo power and position estimation residual of the sensed target within the cluster are read from the previous time slot cluster. The average received signal-to-noise ratio, average echo power, and average position estimation residual are calculated. Based on the beam direction, beamwidth, average echo power, average received signal-to-noise ratio, and average position estimation residual, the integrated communication sensing beam tracking state of the cluster in the current time slot is constructed. The integrated communication-aware beam tracking state of each cluster is input into the deep reinforcement learning policy network to obtain the beam direction adjustment and beamwidth adjustment of the cluster. Based on the beam direction and beamwidth of the previous time slot cluster, as well as the beam direction and beamwidth adjustment amounts of the current time slot, the updated beam direction and updated beamwidth of the cluster in the current time slot are obtained, and the updated beam direction and updated beamwidth are used as the beam direction and beamwidth of the current time slot cluster.
[0013] Preferably, the base station receives the uplink feedback signal and the environmental echo signal vector, and updates the channel state information set, as well as the initial position information and motion information set, as follows: The uplink feedback signals of each communication user are obtained to form a received feedback signal matrix. The instantaneous channel estimation vector of each communication user is obtained based on the received feedback signal matrix. The channel estimation vector of the current time slot is weighted and updated based on the instantaneous channel estimation vector and combined with the channel estimation vector of the current time slot to obtain the updated channel estimation vector. The updated channel state information set is then formed. Matched filtering and time-frequency processing are performed on the environmental echo signal vector to obtain the time delay estimate, Doppler frequency shift estimate and azimuth estimate of the sensed target in the current time slot. The updated distance, updated azimuth and updated radial velocity of the sensed target are also obtained to form an updated set of position and motion information. The updated set of channel state information, along with the updated set of position and motion information, will be used as the set of channel state information and the initial set of position and motion information for the next time slot.
[0014] The wireless communication transmission method based on communication-aware fusion proposed in this invention has the following beneficial technical effects: 1. The multi-user millimeter-wave communication sensing fusion method based on NOMA-ISAC proposed in this application achieves a synergistic improvement in communication and sensing performance by jointly modeling and processing communication users and sensing targets under a unified framework. An improved Gaussian mixture model with directional weights and penalty terms is used to perform unified clustering of communication users and sensing targets, so that users and targets within the same cluster have good aggregation in spatial direction. This is beneficial to improving beam energy concentration and sensing echo quality, and also meets the requirements of power domain non-orthogonal multiple access for channel gain difference and decoding order. Thus, it takes into account both multi-user transmission efficiency and target sensing accuracy within the same beam. The power allocation and decoding order designed based on channel gain and service requirement constraints enable each communication user to share cluster transmit power and spectrum resources under the premise of meeting the minimum received signal-to-noise ratio threshold, thereby improving spectrum utilization and system capacity.
[0015] 2. This application introduces a beam tracking mechanism based on deep reinforcement learning. The beam direction, beamwidth, signal-to-noise ratio of the communication receiver, and the residuals from the sensing echo and position estimation in the previous time slot are used as state variables. This allows for time-slot-level adaptive updates of the cluster beam direction and beamwidth, enabling the millimeter-wave multi-antenna transmitter array to perform rapid beam tracking in response to the dynamic changes of communication users and sensing targets. This improves beam pointing accuracy and link stability in high-speed or complex scenarios. Simultaneously, the uplink feedback signal from the communication user and the echo information from the sensing target are used to perform closed-loop updates of the channel state information and the sensing target state. This ensures that the generated user features and target features are continuously optimized over time slot iterations, forming a dynamic collaborative mechanism integrating communication and sensing. Overall, this improves the system's transmission reliability, sensing accuracy, and resource utilization efficiency. Attached Figure Description
[0016] Figure 1 This is a flowchart of a wireless communication transmission method based on communication-aware fusion according to the present invention. Detailed Implementation
[0017] Embodiments of the present invention are described in detail below. Examples of these embodiments are illustrated in the accompanying drawings, wherein the same or similar symbols denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.
[0018] like Figure 1 The wireless communication transmission method based on communication-aware fusion, as shown, includes the following steps: S1. Establish the communication user set and the sensing target set, and set the carrier frequency, system bandwidth, time slot length, transmit power constraints and spectrum resource constraints; In an optional embodiment, a set of communication users and a set of sensing targets are established, and carrier frequency, system bandwidth, time slot length, transmit power constraints, and spectrum resource constraints are set as follows: A millimeter-wave multi-antenna transmitting array and a millimeter-wave multi-antenna receiving array are configured on the base station side. The number of antenna elements in the millimeter-wave multi-antenna transmitting array is denoted as . The number of antenna elements in a millimeter-wave multi-antenna receiver array is denoted as ; Establish communication user set With the set of perceived targets ,in, For the first One communication user, For the first One perceived target; Set carrier frequency System bandwidth Time slot length transmit power constraint The total transmit power of the base station transmitting signals to all communication users and sensing targets in any time slot shall not exceed the transmit power constraint. ; Set spectrum resource constraints and establish a set of spectrum resource units ,in, For the first Each spectrum resource unit.
[0019] S2. In each time slot, the communication user sends an uplink pilot sequence, the base station sends a sensing and detection signal and receives the uplink pilot sequence and the echo signal vector generated by the sensing and detection signal, and obtains the channel state information set as well as the initial position information and motion information set. In an optional embodiment, in each time slot, the communication user sends an uplink pilot sequence, the base station sends a sensing and detection signal and receives the uplink pilot sequence and the echo signal vector generated by the sensing and detection signal, to obtain a set of channel state information, as well as a set of initial position information and motion information, as follows: With a time slot length of Under the time slot structure, indexed by time slot. Mark the first In the time slot, at the time slot At the start of each time slot, from the set of communication users For each communication user The allocation length is uplink pilot sequence ; At the beginning of each time slot, each communication user transmits its corresponding uplink pilot sequence through its respective uplink channel. The base station-side receiving array receives uplink pilot signals from all communication users, forming a received pilot signal matrix. : ; in, , For communication users In the Uplink channel vector for each time slot, This refers to the number of antenna elements in a millimeter-wave multi-antenna receiver array. For the received noise matrix; The base station receives the pilot signal matrix. and each uplink pilot sequence Calculate the channel estimation vector for each communication user And constitute a set of channel state information. ; The base station transmits sensing and detection signals according to the transmit power constraint and receives echoes from the set of sensing targets, forming a received echo signal vector. Matched filtering and time-frequency processing are then performed on the received echo signal vector to obtain the time delay estimate for each sensing target. Doppler frequency shift estimation and initial azimuth angle The initial distance to the target is calculated based on time delay estimation and Doppler frequency shift estimation. and initial radial velocity : ; ; in, At the speed of light, For the operating wavelength, , For carrier frequency; The initial distance, initial azimuth angle, and initial radial velocity of each sensed target are used to construct the initial position and motion information of the sensed target. ; The initial position information and motion information of all sensed targets are combined into an initial position information and motion information set. .
[0020] S3. Based on the channel state information set, the initial position information, and the motion information set, generate the first feature vector of the communication user and the second feature vector of the perceived target, and form a sample feature set; In an optional embodiment, a first feature vector of the communication user and a second feature vector of the perceived target are generated based on a set of channel state information, initial position information, and motion information, and then combined to form a sample feature set, as follows: Obtain the location coordinates of the communication users in each time slot And calculate the distance between the communication user and the base station. And the azimuth angle of the communication user relative to the base station : ; ; Based on the channel estimation vectors of each communication user in the channel state information set Obtaining channel gain : ; Obtaining communication user service requirements The first feature vector of a communication user is formed by combining the user's location coordinates, distance from the base station, azimuth angle relative to the base station, channel gain, and service requirements. ; Based on the initial position and motion information set, for each sensed target in each time slot, according to the initial distance... and initial azimuth angle Calculate the position coordinates of the perceived target : ; Assign importance weights to perceived targets The importance weights are set based on expert experience, and the position coordinates, initial distance, initial azimuth angle, initial radial velocity, and importance weights of the perceived target are combined to form the second feature vector of the perceived target: ; The sample feature set is composed of the first feature vectors of all communication users and the second feature vectors of all perceived targets. .
[0021] S4. Based on the improved Gaussian mixture model clustering algorithm, the sample feature set is divided into multiple clusters; In an optional embodiment, the sample feature set is divided into multiple clusters based on an improved Gaussian mixture model clustering algorithm, as follows: Rearrange the sample feature set into a sample vector sequence. , And set the number of clusters The mixing weights of the Gaussian components are denoted as... The mean vector is denoted as The covariance matrix is denoted as , ; From each sample in the sample vector sequence Extract the corresponding azimuth angle And obtain the direction weights : ; in, , These are the weighting coefficients; In each iteration, for each sample and each Gaussian component The orientation penalty term is defined based on the cluster orientation angle: ; in, For the first During the nth iteration Cluster orientation angles corresponding to each Gaussian component; And calculate the sample Belongs to the The posterior probability of each Gaussian component : ; in, The penalty coefficient is... , It is a multidimensional Gaussian probability density function. It is a natural exponential function; In each iteration, based on the posterior probability Update Mixed Weights Mean vector Covariance Matrix : ; ; ; And update the cluster orientation angle based on the positional components of the mean vector: ; When the maximum number of iterations is reached Stop the iteration and obtain the cluster label for each sample. : ; The sample feature set is divided into multiple clusters based on the cluster label of each sample, and each cluster corresponds to a cluster index. .
[0022] S5. For each communication user in the cluster, determine the serial interference cancellation decoding order of power domain non-orthogonal multiple access based on the channel gain of the communication user in the cluster, and allocate power coefficients and spectrum resource units to the communication user in the cluster. In an optional embodiment, for each communication user within a cluster, the serial interference cancellation decoding order for power domain non-orthogonal multiple access is determined based on the channel gain of the communication user within the cluster, and power coefficients and spectrum resource units are allocated to the communication user within the cluster, as follows: The communication user set and the sensing target set are divided according to the cluster label of each sample to obtain the intra-cluster communication user set and the intra-cluster sensing target set; For each communication user in the cluster's communication user set, determine the serial interference cancellation decoding order of the power domain non-orthogonal multiple access used by the cluster in each time slot according to the order of channel gain from smallest to largest; Under transmit power constraints Next, a cluster transmit power is allocated to each cluster, and the cluster transmit power is distributed among the communication users within the cluster according to a preset rule. Communication users with smaller channel gain receive a larger power coefficient, and communication users with larger channel gain receive a smaller power coefficient, thereby determining the actual transmit power of each communication user within the cluster. Based on the service requirements of each communication user, a minimum received signal-to-noise ratio (SNR) threshold is determined for each communication user within the cluster. Under the determined power coefficient and serial interference cancellation decoding order, the received SNR of each communication user within the cluster is constrained to be no less than the minimum received SNR threshold. Based on the set of spectrum resource units, a set of spectrum resource units is selected for each cluster and allocated to all communication users within the cluster for power domain non-orthogonal multiple access transmission. The spectrum resource units allocated between different clusters do not overlap.
[0023] In an optional embodiment, the sum of all power coefficients of intra-cluster communication users is one.
[0024] S6. For each cluster, use a deep reinforcement learning policy network to obtain the cluster's beam direction and beamwidth; In an optional embodiment, the beam direction and beamwidth of each cluster are obtained using a deep reinforcement learning policy network, as follows: Within each time slot, for each cluster, the beam direction, beamwidth, received signal-to-noise ratio of the communication users within the cluster, and echo power and position estimation residual of the sensed target within the cluster are read from the previous time slot cluster. The average received signal-to-noise ratio, average echo power, and average position estimation residual are calculated. Based on the beam direction, beamwidth, average echo power, average received signal-to-noise ratio, and average position estimation residual, the integrated communication sensing beam tracking state of the cluster in the current time slot is constructed. The integrated communication-aware beam tracking state of each cluster is input into the deep reinforcement learning policy network to obtain the beam direction adjustment and beamwidth adjustment of the cluster. Based on the beam direction and beamwidth of the previous time slot cluster, as well as the beam direction and beamwidth adjustment amounts of the current time slot, the updated beam direction and updated beamwidth of the cluster in the current time slot are obtained, and the updated beam direction and updated beamwidth are used as the beam direction and beamwidth of the current time slot cluster.
[0025] S7. The base station generates and transmits integrated communication sensing transmission signals for each cluster based on channel state information, power coefficient, spectrum resource unit, beam direction and beamwidth.
[0026] S8. The base station receives the uplink feedback signal and the environmental echo signal vector, and updates the channel state information set, as well as the initial position information and motion information set. In an optional embodiment, the base station receives the uplink feedback signal and the ambient echo signal vector, and updates the channel state information set, as well as the initial position information and motion information set, as follows: The uplink feedback signals of each communication user are obtained to form a received feedback signal matrix. The instantaneous channel estimation vector of each communication user is obtained based on the received feedback signal matrix. The channel estimation vector of the current time slot is weighted and updated based on the instantaneous channel estimation vector and combined with the channel estimation vector of the current time slot to obtain the updated channel estimation vector. The updated channel state information set is then formed. Matched filtering and time-frequency processing are performed on the environmental echo signal vector to obtain the time delay estimate, Doppler frequency shift estimate and azimuth estimate of the sensed target in the current time slot. The updated distance, updated azimuth and updated radial velocity of the sensed target are also obtained to form an updated set of position and motion information. The updated set of channel state information, along with the updated set of position and motion information, will be used as the set of channel state information and the initial set of position and motion information for the next time slot.
[0027] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.
Claims
1. A wireless communication transmission method based on communication-aware fusion, characterized in that, Includes the following steps: S1. Establish the communication user set and the sensing target set, and set the carrier frequency, system bandwidth, time slot length, transmit power constraints and spectrum resource constraints; S2. In each time slot, the communication user sends an uplink pilot sequence, the base station sends a sensing and detection signal and receives the uplink pilot sequence and the echo signal vector generated by the sensing and detection signal, and obtains the channel state information set as well as the initial position information and motion information set. S3. Based on the channel state information set, the initial position information, and the motion information set, generate the first feature vector of the communication user and the second feature vector of the perceived target, and form a sample feature set; S4. Based on the improved Gaussian mixture model clustering algorithm, the sample feature set is divided into multiple clusters; S5. For each communication user in the cluster, determine the serial interference cancellation decoding order of power domain non-orthogonal multiple access based on the channel gain of the communication user in the cluster, and allocate power coefficients and spectrum resource units to the communication user in the cluster. S6. For each cluster, use a deep reinforcement learning policy network to obtain the cluster's beam direction and beamwidth; S7. The base station generates and transmits integrated communication sensing transmission signals for each cluster based on channel state information, power coefficient, spectrum resource unit, beam direction and beamwidth. S8. The base station receives the uplink feedback signal and the environmental echo signal vector, and updates the channel state information set, as well as the initial position information and motion information set.
2. The wireless communication transmission method based on communication-aware fusion according to claim 1, characterized in that, Establish a set of communication users and a set of sensing targets, and set constraints on carrier frequency, system bandwidth, time slot length, transmit power, and spectrum resources as follows: Establish a set of communication users and a set of sensing targets; Set carrier frequency, system bandwidth, time slot length, and transmit power constraints, ensuring that the total transmit power of the base station to all communication users and sensing targets in any time slot does not exceed the transmit power constraints; Set spectrum resource constraints and establish a set of spectrum resource units.
3. The wireless communication transmission method based on communication-aware fusion according to claim 1, characterized in that, In each time slot, the communication user sends an uplink pilot sequence, and the base station sends a sensing and detection signal and receives the uplink pilot sequence and the echo signal vector generated by the sensing and detection signal, thus obtaining a set of channel state information, initial position information, and motion information, as follows: At the beginning of each time slot, each communication user sends its corresponding uplink pilot sequence through its own uplink channel. The base station receiving array receives the uplink pilot signals of all communication users and forms a receiving pilot signal matrix. The base station calculates the channel estimation vector for each communication user based on the received pilot signal matrix and each uplink pilot sequence, and forms a set of channel state information. The base station transmits sensing and detection signals according to the transmit power constraint and receives echoes from the set of sensing targets, forming a received echo signal vector. Matched filtering and time-frequency processing are performed on the received echo signal vector to obtain the time delay estimate, Doppler frequency shift estimate and initial azimuth angle of each sensing target. The initial distance and initial radial velocity of the sensing target are calculated based on the time delay estimate and Doppler frequency shift estimate. The initial distance, initial azimuth angle, and initial radial velocity of each sensed target are used to form the initial position and motion information of the sensed target. The initial position information and motion information of all sensed targets are combined into an initial position information and motion information set.
4. The wireless communication transmission method based on communication-aware fusion according to claim 1, characterized in that, Based on the channel state information set, initial position information, and motion information set, a first feature vector of the communication user and a second feature vector of the perceived target are generated, forming a sample feature set as follows: Obtain the location coordinates of the communication user in each time slot, and calculate the distance of the communication user relative to the base station and the azimuth angle of the communication user relative to the base station; The channel gain is obtained based on the channel estimation vector of each communication user in the channel state information set. The communication user's service requirements are obtained by combining the communication user's location coordinates, distance from the communication user to the base station, azimuth angle of the communication user relative to the base station, channel gain, and service requirements to form the first feature vector of the communication user. Based on the initial position information and motion information set, the position coordinates of each sensing target are calculated in each time slot according to the initial distance and initial azimuth angle; An importance weight is assigned to the sensing target, and the position coordinates, initial distance, initial azimuth angle, initial radial velocity, and importance weight of the sensing target are combined to form the second feature vector of the sensing target; The sample feature set is composed of the first feature vectors of all communication users and the second feature vectors of all perceived targets.
5. The wireless communication transmission method based on communication-aware fusion according to claim 1, characterized in that, The improved Gaussian mixture model clustering algorithm divides the sample feature set into multiple clusters, as follows: The sample feature set is rearranged into a sequence of sample vectors, and the number of clusters is set. Extract the corresponding azimuth angle and obtain the direction weight from each sample in the sample vector sequence; In each iteration, a direction penalty term is defined for each sample and each Gaussian component based on the cluster orientation angle, and the posterior probability of the sample is calculated. In each iteration, the mixture weights, mean vector, and covariance matrix are updated based on the posterior probability, and the cluster orientation angle is updated based on the positional components of the mean vector. When the maximum number of iterations is reached, stop iterating and obtain the cluster label for each sample; The sample feature set is divided into multiple clusters based on the cluster label of each sample.
6. The wireless communication transmission method based on communication-aware fusion according to claim 1, characterized in that, For each communication user within a cluster, the serial interference cancellation decoding order for power domain non-orthogonal multiple access is determined based on the channel gain of the communication user within the cluster. Power coefficients and spectrum resource units are then allocated to the communication users within the cluster, as follows: The communication user set and the sensing target set are divided according to the cluster label of each sample to obtain the intra-cluster communication user set and the intra-cluster sensing target set; For each communication user in the cluster's communication user set, the serial interference cancellation decoding order of the power domain non-orthogonal multiple access used by the cluster in each time slot is determined in ascending order of channel gain; Each cluster is assigned a cluster transmit power, and the cluster transmit power is distributed among the communication users within the cluster according to a preset rule to determine the actual transmit power of each communication user within the cluster. Based on the service requirements of each communication user, a minimum received signal-to-noise ratio (SNR) threshold is determined for each communication user within the cluster. Under the determined power coefficient and serial interference cancellation decoding order, the received SNR of each communication user within the cluster is constrained to be no less than the minimum received SNR threshold. Based on the set of spectrum resource units, a set of spectrum resource units is selected for each cluster and allocated to all communication users within the cluster for power domain non-orthogonal multiple access transmission. The spectrum resource units allocated between different clusters do not overlap.
7. The wireless communication transmission method based on communication-aware fusion according to claim 6, characterized in that, The sum of all power coefficients of all users communicating within a cluster is one.
8. The wireless communication transmission method based on communication-aware fusion according to claim 1, characterized in that, For each cluster, a deep reinforcement learning policy network is used to obtain the cluster's beam direction and beamwidth, as follows: Within each time slot, for each cluster, the beam direction, beamwidth, received signal-to-noise ratio of the communication users within the cluster, and echo power and position estimation residual of the sensed target within the cluster are read from the previous time slot cluster. The average received signal-to-noise ratio, average echo power, and average position estimation residual are calculated. Based on the beam direction, beamwidth, average echo power, average received signal-to-noise ratio, and average position estimation residual, the integrated communication sensing beam tracking state of the cluster in the current time slot is constructed. The integrated communication-aware beam tracking state of each cluster is input into the deep reinforcement learning policy network to obtain the beam direction adjustment and beamwidth adjustment of the cluster. Based on the beam direction and beamwidth of the previous time slot cluster, as well as the beam direction and beamwidth adjustment amounts of the current time slot, the updated beam direction and updated beamwidth of the cluster in the current time slot are obtained, and the updated beam direction and updated beamwidth are used as the beam direction and beamwidth of the current time slot cluster.
9. The wireless communication transmission method based on communication-aware fusion according to claim 1, characterized in that, The base station receives the uplink feedback signal and the environmental echo signal vector, and updates the channel state information set, as well as the initial position information and motion information set, as follows: The uplink feedback signals of each communication user are obtained to form a received feedback signal matrix. The instantaneous channel estimation vector of each communication user is obtained based on the received feedback signal matrix. The channel estimation vector of the current time slot is weighted and updated based on the instantaneous channel estimation vector and combined with the channel estimation vector of the current time slot to obtain the updated channel estimation vector. The updated channel state information set is then formed. Matched filtering and time-frequency processing are performed on the environmental echo signal vector to obtain the time delay estimate, Doppler frequency shift estimate and azimuth estimate of the sensed target in the current time slot. The updated distance, updated azimuth and updated radial velocity of the sensed target are also obtained to form an updated set of position and motion information. The updated set of channel state information, along with the updated set of position and motion information, will be used as the set of channel state information and the initial set of position and motion information for the next time slot.