A perception-aided millimeter wave channel adaptive tracking method

Abstract

This invention belongs to the field of integrated sensing technology for millimeter-wave communication, specifically relating to a sensing-assisted adaptive tracking method for millimeter-wave channels. The method includes: constructing a time-division duplex frame structure, divided into sensing time slots and communication time slots; in the downlink phase of each sensing time slot: updating the environmental scatterer set information through a multi-target tracking algorithm to construct the channel geometry, including determining multiple non-line-of-sight propagation paths and the path parameters of each path; constructing the geometry of each path based on its path parameters; in the uplink phase: acquiring the prior geometry of each path and the statistical prior of the complex gain of the current sensing time slot to obtain the updated complex gain of that path; in the uplink phase of each communication time slot: acquiring the prior geometry of each valid existing path and the statistical prior of the complex gain of the current communication time slot to obtain the updated complex gain of that path. This invention achieves adaptive channel tracking for dynamic scenarios with path creation and disappearance.

Description

Technical Field

[0001] This invention belongs to the field of integrated sensing technology for millimeter-wave communication, and more specifically, relates to a sensing-assisted millimeter-wave channel adaptive tracking method. Background Technology

[0002] With the development of millimeter-wave massive MIMO systems, high-speed mobile communications, and integrated communication and sensing technologies, higher requirements have been placed on the acquisition of channel state information.

[0003] Existing millimeter-wave wireless channel estimation techniques employ sparse parameterization modeling to characterize channel properties in parameter domains such as angle, delay, and Doppler. Based on the quasi-static assumption—that is, assuming a fixed number of paths and invariant geometry within the coherence time—time-varying channels in high-speed, large-scale millimeter-wave scenarios are estimated using methods such as compressed sensing, subspace decomposition, and tensor decomposition. Furthermore, millimeter-wave wireless channel tracking techniques utilize temporal correlation, recursively updating path gain or angle parameters to achieve cross-timeslot channel tracking. These methods reduce pilot overhead and have been widely applied in millimeter-wave vehicle-to-everything (V2X) communication scenarios. With the further development of Integrated Communication and Sensing (ISAC) technology, the sensing module of the ISAC system has become a new source of prior information, applicable to channel estimation and channel tracking, further reducing system pilot overhead. ISAC-assisted communication channel estimation and tracking has become a new research hotspot.

[0004] Existing millimeter-wave channel estimation methods are typically based on quasi-static assumptions, meaning the channel is assumed to remain constant within a training block. Channel tracking methods, on the other hand, are based on slowly time-varying assumptions, recursively updating channel parameters through state-space models. However, most of these methods assume the number of propagation paths remains constant, only estimating or updating parameters for existing paths. However, in time-varying scenarios, the set of propagation paths themselves can change significantly over time. Furthermore, when system sensors have hardware limitations or beam scheduling mismatches, false alarms and missed detections can occur. In these cases, the number of paths is no longer a fixed constant but becomes a random variable that evolves over time. When the path set changes, modeling methods based on fixed-dimensional state space or fixed support set assumptions may face the following problems: the need to re-execute the structure reconstruction process, increasing pilot overhead and computational complexity; mistakenly including vanished paths in the model, leading to state-dimensional redundancy; delayed response to newly formed paths, increasing estimation residuals; and a lack of reconstruction mechanisms under conditions of structure mismatch caused by path birth and death. Therefore, when the number of paths changes over time, it is necessary to continuously acquire path-aware parameter information through multi-target tracking. Summary of the Invention

[0005] To address the aforementioned deficiencies or improvement needs of existing technologies, this invention provides a perception-assisted millimeter-wave channel adaptive tracking method, which aims to achieve adaptive channel tracking in dynamic scenarios where path generation and disappearance occur.

[0006] To achieve the above objectives, according to one aspect of the present invention, a perception-assisted millimeter-wave channel adaptive tracking method is provided, comprising: A time-division duplex frame structure is constructed, and the frame structure is divided into a sensing time slot and a communication time slot, wherein multiple consecutive communication time slots are set between two sensing time slots; In the downlink phase of each sensing time slot: the base station transmits a sensing integrated signal and receives environmental echoes; based on the environmental echoes, it updates the environmental scatterer set information using a multi-target tracking algorithm; based on the updated scatterer set information, it constructs the channel geometry, including determining multiple non-line-of-sight propagation paths and path parameters for each path, where the path parameters include transmission angle, angle of arrival, time delay, and Doppler shift; based on the path parameters of each path, it constructs the geometry of that path. ; In the uplink phase of each sensing time slot: the base station receives pilot signals sent by user equipment to update the complex gain of each non-line-of-sight propagation path. The update method is to obtain the geometry of each path. The prior and statistical prior of the complex gain of the current synesthetic time slot are considered. If the path is a newly created path in the downlink phase of the current synesthetic time slot, the statistical prior of the complex gain of the path is a weak Gaussian prior. If the path is an existing path, the complex gain is constructed based on the time correlation of the complex gain obtained from the previous time slot update. The calculation is based on the geometric structure of each path. The prior and the statistical prior of complex gain under the current sensing time slot are combined with the pilot signal to obtain the updated complex gain of each path under the current sensing time slot; the geometric structure of each path and the updated complex gain constitute the channel of the current sensing time slot, and the adaptive tracking of the time slot is completed. In the downlink phase of each communication time slot: based on the mean and variance of the path complex gain after updates across multiple consecutive time slots, the validity of each existing path is determined, identifying the valid existing paths for the current communication time slot, and updating the set of paths participating in the complex gain update; the geometric structure of each valid existing path in this path set is then obtained. ; In the uplink phase of each communication time slot: the base station receives pilot signals sent by user equipment to update the complex gain of each valid legacy path. The update method is to obtain the geometric structure of each valid legacy path. The prior and the statistical prior of complex gain under the current communication time slot, wherein the statistical prior is constructed by using the complex gain updated in the previous time slot according to the time correlation of complex gain; based on the geometric structure of each valid old path. The prior and complex gain statistical prior, combined with the pilot signal, yield the updated complex gain of each valid old path in the current communication time slot; the geometric structure of each path and the updated complex gain constitute the channel of the current communication time slot, completing the adaptive tracking of the time slot.

[0007] Furthermore, the echo signal is formed by reflections from multiple scatterers in the environment; The method for updating the environmental scatterer set information based on environmental echoes using a multi-target tracking algorithm is as follows: The base station processes the received echo signal to obtain observation information related to the scatterer, including the distance, angle, and Doppler information of the scatterer. Based on the observation information, a multi-target tracking algorithm is used to estimate the state of multiple scatterers in the environment to obtain the scatterer set information at the current moment. The scatterer set information includes the position information, velocity information, and existence probability of each scatterer.

[0008] Furthermore, in the downlink phase of each communication slot, the geometry of each valid legacy path is obtained through prediction. The implementation method is as follows: The geometry of this effective legacy path, determined based on the most recent synesthesia time slot. Based on the kinematic model, the geometric structure of the valid old path in the current communication time slot is analyzed. Perform forecast updates.

[0009] Furthermore, the mean and variance of the updated complex gain for all paths in each time slot are expressed as follows: The complex gains to be updated for all paths in the current time slot are combined into a vector, denoted as . The mean vector corresponding to the updated complex gain vectors of all paths in the current time slot. Covariance Matrix for:

[0010]

[0011] in, Represents the LMMSE gain matrix. , , Let these represent the prior mean vector and prior covariance matrix required for updating the complex gain vector, respectively. , , Indicates the number of valid paths. This represents the vectorized pilot received signal matrix. This indicates that the observation dictionary matrix is ​​obtained by concatenating the equivalent observation vectors corresponding to all paths involved in the complex gain. Indicates the noise variance. express An identity matrix of dimension 1 Indicates the number of receiving antennas at the base station. Indicates the pilot length.

[0012] Furthermore, the method for determining the validity of each existing path is as follows: If the existing path is continuous before the included communication time slot Each time slot satisfies any of the following conditions: a) The square of the mean of the updated complex gain of the path is less than a preset threshold. b) The variance of the updated complex gain of this path is greater than a preset threshold. If the path is frozen, it will not participate in prior construction and complex gain update in subsequent communication time slots until the subsequent synesthesia time slot is confirmed by multi-target tracking.

[0013] Furthermore, during the uplink phase of each communication time slot, path creation detection is also performed, implemented as follows: Calculate the current communication time slot residual matrix In the formula, Indicates communication time slot The pilot received signal matrix below, , representing the channel matrix recovered after updating with the complex gain of the effective old path. These represent the number of base station receiving antennas and the number of user transmitting antennas, respectively. Represents the equivalent observation vector of all valid old paths. The constructed observation dictionary matrix, the first Equivalent observation vector of an existing valid path , Indicates the first The geometric structure of an existing valid path. Indicates communication time slot The pilot transmit signal matrix below, Indicates communication time slot The updated complex gain vector is derived from all valid existing paths. Based on the residual matrix Calculate residual energy. Indicates the noise variance. Indicates the known pilot length. Represents the Frobenius norm of the matrix; if the residual energy exceeds a threshold This indicates the existence of a new non-line-of-sight propagation path. The residual matrix is ​​represented in the angular domain using a two-dimensional fast Fourier transform, and peak detection is performed, retaining the path with the highest energy. Each corner domain component, according to The angle index corresponding to each angle component restores the arrival angle of the corresponding newborn non-line-of-sight propagation path. With departure angle , represented as , used to describe The complex gain of each angular domain component is solved using the least squares method, and the solution formula is: In the formula, Indicates communication time slot The complex gain of all newborn non-line-of-sight propagation paths is as follows. The complex gain vector formed by them This represents the receiver array steering vector of the base station. This represents the transmit array steering vector of the user equipment; Calculate the channel compensation term based on the complex gain of all newly generated non-line-of-sight propagation paths obtained from the solution. , The path geometry of this compensation item does not include any valid set of existing paths. The path in; The channel compensation term is combined with the channel matrix recovered after updating with the complex gain of the effective old path. Add them together to get the current communication time slot. The channel is used to complete adaptive tracking.

[0014] According to another aspect of the present invention, an electronic device is provided, including a memory and a processor, the memory storing a computer program, the processor executing the computer program to implement the steps of the method described above.

[0015] According to another aspect of the invention, a computer-readable storage medium is provided, the computer-readable storage medium including a stored computer program, wherein, when the computer program is run by a processor, it controls the device where the storage medium is located to perform the steps of the method described above.

[0016] According to another aspect of the invention, a computer program product is provided, comprising a computer program or instructions that, when executed by a processor, implement the steps of the method described above.

[0017] In summary, compared with the prior art, the technical solutions conceived by this invention have the following main advantages: 1. This invention proposes a sensing-assisted millimeter-wave channel adaptive tracking method, applied to an integrated communication and sensing system. Under a unified time-division duplex frame structure, sensing time slots and communication time slots are set: In the downlink phase of the sensing time slot, the base station transmits an integrated sensing signal and updates the scatterer set information based on the environmental echo using a multi-target tracking algorithm. The channel geometry is then constructed based on the scatterer set information to more accurately reflect the dynamic propagation characteristics of the millimeter-wave wireless channel. In the downlink phase of the communication time slot, the base station can perform regular data transmission. In the uplink phase of each time slot, the user transmits pilot signals, and the base station recursively updates the complex gain of each propagation path based on the pilot signals. During the uplink phase of the communication time slot, the propagation path status is determined through validity detection. Furthermore, in the complex gain update, this invention utilizes the propagation path geometry obtained from multi-target detection as prior information for the channel structure. Only a small number of pilot signals are needed to complete the complex gain update, avoiding frequent high-dimensional channel re-estimation, thereby reducing the computational complexity of channel estimation and pilot overhead. This method is suitable for applications in high-speed mobile and massive MIMO systems. In summary, this invention is based on a channel model. By separating the channel geometry update from the complex gain update, the computational complexity of channel estimation can be effectively reduced. Furthermore, the multi-target tracking algorithm is used to perform joint state estimation on multiple scatterers in the environment and to track the appearance and disappearance of scatterers. This allows for adaptive adjustment of the path structure when the propagation path appears or disappears, thus achieving adaptive tracking of the millimeter-wave channel.

[0018] 2. Further, in the downlink phase of each communication time slot, considering that the path parameters in the communication time slot may change due to the dynamic environment, the present invention proposes to predict and update the geometric structure of the path in the current communication time slot based on the geometric structure of the path estimated in the most recent sensing time slot and combined with the kinematic model, and to update the complex gain of each propagation path in the current communication time slot, so as to more accurately reflect the dynamic propagation characteristics of the millimeter wave wireless channel.

[0019] 3. This invention also proposes an adaptive channel tracking mechanism. First, the validity of the path is detected using statistical information of the complex gain, and the complex gain is updated within the set of valid paths. Then, path generation is detected based on the updated residual energy, and channel compensation is further performed. Through this mechanism, dynamic updating of the propagation path set can be achieved, thereby realizing adaptive tracking of millimeter-wave channels. Attached Figure Description

[0020] Figure 1 A flowchart of a perception-assisted millimeter-wave channel adaptive tracking method provided in an embodiment of the present invention; Figure 2This is a schematic diagram of the time axis and time slot structure in the integrated communication and sensing system provided in an embodiment of the present invention; Figure 3 This is a comparison chart of the NMSE performance of different channel estimation algorithms provided in the embodiments of the present invention under different signal-to-noise ratio conditions; Figure 4 The simulation results of the impact of pilot repetition period, time correlation coefficient and signal-to-noise ratio on channel estimation performance are provided in the embodiments of the present invention. Figure 5 This is a comparison chart of channel estimation performance before and after the introduction of the path structure adaptive update mechanism provided in an embodiment of the present invention. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. Furthermore, the technical features involved in the various embodiments of this invention described below can be combined with each other as long as they do not conflict with each other.

[0022] Example 1 A perception-assisted millimeter-wave channel adaptive tracking method, such as Figure 1 As shown, it includes: A time-division duplex frame structure is constructed, and the frame structure is divided into a sensing time slot and a communication time slot, wherein multiple consecutive communication time slots are set between two sensing time slots; In the downlink phase of each sensing time slot: the base station transmits a sensing integrated signal and receives environmental echoes. Based on the environmental echoes, the environmental scatterer set information is updated using a multi-target tracking algorithm. The channel geometry is constructed based on the updated scatterer set information, including determining multiple non-line-of-sight propagation paths and the path parameters for each path, where the path parameters include transmission angle, angle of arrival, time delay, and Doppler shift. The geometry of each path is constructed based on its path parameters. In the uplink phase of each sensing time slot: the base station receives pilot signals transmitted by user equipment to update the complex gain of each non-line-of-sight propagation path. The update method is as follows: The process involves: acquiring the prior geometric structure of each path and the statistical prior of the complex gain in the current sensing time slot; if the path is a newly created path in the downlink phase of the current sensing time slot, the statistical prior of the complex gain is a weak Gaussian prior; if the path is an existing path, the complex gain obtained from the previous time slot update is constructed based on the time correlation of the complex gain; based on the prior geometric structure of each path and the statistical prior of the complex gain in the current sensing time slot, combined with the pilot signal, the updated complex gain of each path in the current sensing time slot is obtained; the geometric structure of each path and the updated complex gain constitute the channel of the current sensing time slot, completing the adaptive tracking of the time slot; In the downlink phase of each communication time slot: based on the mean and variance of the path complex gain after updates in multiple consecutive time slots, the validity of each existing path is determined, the valid existing paths in the current communication time slot are identified, and the set of paths participating in the complex gain update is updated; the geometric structure of each valid existing path in this path set is obtained. In the uplink phase of each communication time slot: the base station receives pilot signals sent by user equipment to update the complex gain of each valid existing path. The update method is as follows: the prior geometric structure of each valid existing path and the statistical prior of complex gain in the current communication time slot are obtained. The statistical prior is constructed by using the complex gain obtained from the previous time slot update based on the time correlation of complex gain. Based on the prior geometric structure of each valid existing path and the statistical prior of complex gain, combined with the pilot signals, the updated complex gain of each valid existing path in the current communication time slot is obtained. The geometric structure of each path and the updated complex gain constitute the channel of the current communication time slot, completing the adaptive tracking of this time slot.

[0023] This embodiment applies the method to an integrated communication and sensing system. Under a unified time-division duplex frame structure, the system sets up sensing time slots and communication time slots: In the downlink phase of the sensing time slot, the base station transmits an integrated sensing signal and updates the scatterer set information based on the environmental echo using a multi-target tracking algorithm, and constructs the channel geometry based on the scatterer set information; In the downlink phase of the communication time slot, the base station can perform regular data transmission; In the uplink phase of each time slot, the user sends a pilot signal, and the base station recursively updates the complex gain of each propagation path based on the pilot signal, and determines the propagation path status through validity detection and residual detection. Based on the channel model , dimension ,in, These represent the number of base station receiving antennas and the number of user transmitting antennas, respectively. Represents the number of non-line-of-sight propagation paths, generally including . Representing the Complex gain of the path, Representing the The geometric structure of the path is represented as follows: in, Representing the The transmission angle, arrival angle, Doppler shift, and time delay of each path, and This represents the MIMO array transmit / receive steering vector. Indicates the carrier frequency of the transmitted signal. The time interval represents the distance between the two objects. This embodiment of the method effectively reduces the computational complexity of channel estimation by separating the channel geometry update from the complex gain update. Furthermore, the multi-target tracking algorithm is used to perform joint state estimation on multiple scatterers in the environment and to track the appearance and disappearance of the scatterers. This allows for adaptive adjustment of the path structure when the propagation path appears or disappears, thereby achieving adaptive tracking of the millimeter-wave channel.

[0024] Specifically, firstly, a unified time-division duplex frame structure is constructed within the integrated communication and sensing system. The system includes a base station and at least one user equipment, where the base station is equipped with a MIMO antenna array and possesses communication and environmental sensing capabilities. The time-division duplex frame structure comprises multiple time slots, divided into sensing time slots and communication time slots. In the sensing time slots, the base station achieves environmental sensing by transmitting integrated sensing signals; in the communication time slots, the base station performs conventional communication data transmission. Multiple communication time slots are configured between two sensing time slots, the number of which is determined according to actual needs.

[0025] See Figure 2 A unified time-division duplex frame structure is constructed, dividing the frame structure into sensing time slots and communication time slots. Specifically: in the sensing time slot, the base station achieves environmental perception by transmitting integrated sensing signals; in the communication time slot, the base station performs regular communication data transmission. This frame structure design enables the system to acquire environmental scatterer information while communicating. Since changes in environmental geometry are typically slower than the rapid fluctuations in complex gain, sensing time slots are periodically inserted into the communication time slot sequence at a lower frequency to update the state information of environmental scatterers. Between adjacent sensing time slots, the system uses a scatterer motion model to predict their state, thereby achieving continuous tracking of the propagation path geometry; and in the uplink phase of each time slot, the complex path gain is updated via pilot signals.

[0026] The index of the time slot is defined as:

[0027] in, The total number of time slots is [value], and the duration of each time slot is [duration]. And define the synesthetic time slot set. ,when During the downlink phase, the system performs environmental awareness to update the channel geometry, and during the uplink phase, it updates the complex gain of each propagation path; when During the uplink phase, the system updates the complex gain of each propagation path, but does not update the channel geometry during the downlink phase.

[0028] Through the frame structure design described above, the system can acquire environmental information while communicating. This method can more accurately describe the dynamic characteristics of millimeter-wave wireless channels. Unlike existing methods that assume a fixed number of propagation paths or use fixed-dimensional state space modeling, this embodiment considers the dynamic changes of the propagation path set. It continuously tracks the channel propagation paths through multi-target tracking and further acquires the parameter information of the propagation paths, thereby constructing a dynamically updated channel geometry structure that more accurately reflects the dynamic propagation characteristics of millimeter-wave wireless channels.

[0029] Secondly, during the downlink phase of each sensing time slot, a multi-target tracking algorithm is used to estimate the state of multiple scatterers in the environment, thereby obtaining the scatterer set information at the current moment. Based on the updated scatterer set information, a channel geometry is constructed (determining multiple non-line-of-sight propagation paths and the path parameters contained in each path). Specifically, the non-line-of-sight propagation path of the signal is determined according to the geometric relationship between the base station, user equipment, and scatterers. For each scatterer, a reflection propagation path from the base station to the scatterer and then to the user equipment can be constructed, and the structural parameters of the non-line-of-sight propagation path are calculated according to the geometric relationship. The structural parameters include, but are not limited to, transmission angle, angle of arrival, time delay, Doppler shift, etc.

[0030] In this embodiment, the propagation path set is estimated by observing the wireless propagation environment to obtain the existence status and path parameters of the propagation paths. The path set is a variable-size set, and its elements correspond to the effective propagation paths in the propagation environment. The estimation of the path set can be achieved using multi-target tracking algorithms, such as random finite set methods, multi-hypothesis tracking methods, or joint probability data association methods, thereby achieving unified processing of unknown path quantity, path parameter changes, and path appearance or disappearance. This embodiment is applied to a communication sensing integration scenario where the communication terminal is a cooperative user. A communication link is established between the base station and the communication terminal. The base station can obtain information such as the spatial location, motion state, and direct path (i.e., line-of-sight path) parameters between the base station and the terminal. In this scenario, the base station and the communication terminal form a bi-base geometric structure. The method in this embodiment mainly focuses on modeling and dynamically updating non-direct paths (i.e., non-line-of-sight paths: base station-scatterer-terminal) generated by environmental scatterers, while the line-of-sight path can be used as known information to participate in channel reconstruction. During system operation, by processing the sensing echo signals received in the downlink phase of the sensing time slot, the set of non-line-of-sight propagation paths and their corresponding path information can be obtained, including parameters such as angle of arrival, Doppler frequency shift, transmission angle, propagation delay, and path existence probability. Based on the above path parameters, the channel geometry can be constructed.

[0031] In practical implementations, for example, a multi-target tracking algorithm based on RFS-LMB (Random Finite Set Tag Multi-Bernoulli Filtering) can be used to obtain information on the presence or absence of scatterers and their state vectors. These state vectors can then be used to construct path parameters. In cooperative user scenarios, the state vectors of the base station and the user are known to the base station.

[0032] Assuming we are currently at the th There are 1 communication time slot, and the most recent synaptic time slot is defined as:

[0033] exist In the downlink phase of a single sensing time slot, the set of scatterers obtained by the multi-target tracking algorithm is represented as follows:

[0034] in, The labels for the corresponding scatterers are used to maintain the consistency of the scatterers in the time dimension, based on the state vector of the scatterers. Then, based on the known state vectors of the base stations and users... They are defined as follows:

[0035] It represents the current number The states of the scatterers, where... Let x and y represent the coordinates and velocity of the scatterer in the spatial coordinate system, respectively. The known physical states of the base station and the user are also represented as follows:

[0036]

[0037] in, These represent the base station's coordinates in the x and y directions and its velocity in the spatial coordinate system, respectively. These represent the user's coordinates in the x and y directions and their velocity in the spatial coordinate system, respectively.

[0038] It should be noted that the base station is stationary in space, while the user and the scattering body are in motion, and the value of the state vector changes under different sensing time slots.

[0039] The state vectors obtained above can be used to construct the path parameters for the propagation path: First, define the position difference vector and velocity difference vector of the scatterer relative to the base station as follows:

[0040]

[0041] in, This indicates the spatial position of the scatterer relative to the base station. This represents the relative velocity between the two. Based on the above definition, we have:

[0042]

[0043] in, Indicates the distance between the base station and the scattering object. This represents the departure angle. Then, the position difference vector and velocity difference vector of the scatterer relative to the user are further defined as follows:

[0044]

[0045] Therefore:

[0046]

[0047]

[0048]

[0049] in, For carrier wavelength, At the speed of light, These represent the distance from the base station to the scatterer and then to the user, the time delay, the angle of arrival, and the Doppler shift, respectively.

[0050] Therefore, the first MIMO channel can be obtained. The path parameters for this path are: The geometric structure of the path can then be estimated from the path parameters. .

[0051] The method in this embodiment further performs complex gain updates during the uplink phase of each synaptic time slot.

[0052] In millimeter-wave propagation environments, channel evolution is primarily driven by two factors: the geometry of the propagation path and the path complex gain. Geometric parameters such as angle and time delay are determined by the spatial positions of the scatterer and communication nodes, and their changes are typically relatively slow. In contrast, the path complex gain is affected by small-scale fading, residual Doppler, and hardware noise, and changes more rapidly. Therefore, for the... Each time slot (whether for sensing or communication) can represent the channel as a combination of geometric structure and path complex gain, defined as follows:

[0053] in, Representing the The first time slot The complex gain of the path is caused by receiver hardware noise, residual Doppler, and small-scale fading. Representing the The first time slot The geometric structure of the path.

[0054] Because the geometry of the path evolves relatively slowly, its information can be acquired through multi-target tracking over a longer update interval (i.e., the interval between two sensing time slots). Therefore, channel tracking can focus on estimating the complex gain, which changes rapidly over time. Next, the temporal correlation of the complex gain can be utilized, i.e., it can be obtained through the... The a posteriori of the complex gain of the time slot is related to the _th The complex gain prior of the time slot is recursively calculated, and this recursive process is usually modeled as a first-order Gaussian-Markov process. For the The path, at the In each time slot, it can be represented as:

[0055] in, It is process noise, subject to It is used to absorb random effects such as residual Doppler, small-scale fading, and receiver hardware noise. This represents the variance of process noise. , representing the time correlation coefficient, where For zero-order Bessel functions, this model reflects the change in performance as... and As the value increases, the time dependence of the complex gain decreases.

[0056] Based on this, the uplink phase of the synesthesia time slot is divided into two stages: prior acquisition and complex gain update.

[0057] First, prior information is constructed. In the proposed perception-assisted channel tracing framework, the channel prior consists of two parts: one part is the geometric structure of the propagation path obtained in the previous time slot. The other part is the time correlation of the path complex gain. Based on the aforementioned channel model, and according to these two types of priors, the channel update problem under subsequent pilot observations can be transformed into a low-dimensional recursive estimation problem of the path complex gain.

[0058] Regarding the time correlation of path complex gain, a statistical prior for complex gain is constructed, and two cases are discussed in the synesthesia time slot: (1) For each existing path, based on the time correlation of the complex gain defined above, the prior of the complex gain obtained from the previous time slot update (the complex gain after the previous time slot update) can be used, combined with the time correlation of the complex gain, to construct the prior required for the current time slot complex gain update. First, the prior mean of the current time slot complex gain is expressed as:

[0059] The prior covariance of the current time slot complex gain is expressed as:

[0060] Therefore, in the first The statistical prior of the complex gain over each time slot (i.e., the current time slot) is:

[0061] (2) For new paths, it is necessary to initialize their geometry and the priors required for complex gain updates. For geometry initialization, in the current synesthetic time slot... New paths constructed based on multi-object tracking algorithms The geometric structure is as follows:

[0062] in This indicates that when path generation is detected in the syn-sensory time slot, the Doppler frequency shift does not cause a change in the phase term.

[0063] For the prior initialization required for complex gain update, due to the lack of historical complex gain information, the prior required for complex gain update can be expressed as:

[0064] Among them, the prior mean required for the complex gain update of the new path is 0, and the variance is... Setting it to a large value indicates a lack of trust in the current prior information, which is called a weak Gaussian prior.

[0065] At this point, the channel prior for the current time slot is complete before entering the uplink phase of the current sensing time slot. Then, the complex gain update process is performed within the uplink phase of the current time slot.

[0066] After receiving the pilot signal, the base station updates the path complex gain using the geometry-based linear minimum mean square error (GB-LMMSE) method in one exemplary implementation.

[0067] Specifically, the pilot received signal matrix can be represented as:

[0068] in, The dimension is , This represents the user's pilot transmission signal matrix, with dimensions of... , Indicates the number of transmit antennas for the user. This indicates the pilot length. In one exemplary configuration, the pilot length is not less than the number of transmit antennas of the user (i.e., the communication terminal), that is... This ensures that the matrix is ​​not rank deficient. The matrix is ​​an additive white Gaussian noise matrix, with each column being independent and identically distributed, following the formula... .

[0069] The current channel contains There are 10 valid paths. Then, the complex gains of all current paths to be updated are combined into a vector as follows:

[0070] Therefore, based on the prior knowledge required for the complex gain update of each path, the prior mean vector and prior covariance matrix required to construct this complex gain vector update are respectively expressed as:

[0071]

[0072] In the formula, This represents a diagonal matrix.

[0073] Furthermore, the pilot received signal matrix is ​​vectorized, that is:

[0074] Then we can obtain:

[0075] For the A path, determined by its geometry and pilot transmit signal matrix The resulting equivalent observation vector is:

[0076] Then, the equivalent observation vectors corresponding to all paths are concatenated to obtain the observation dictionary matrix:

[0077] noise vector obey ,in, This represents the noise variance.

[0078] Next, using the LMMSE method, the mean vector and covariance matrix corresponding to the updated complex gain vector in the current time slot can be expressed as follows:

[0079]

[0080] The LMMSE gain matrix can be expressed as:

[0081] In summary, compared with existing methods, the proposed method simultaneously utilizes the propagation path geometry information provided by multi-target tracking and the temporal correlation of complex gain. Specifically, it divides the channel estimation problem into two parts: path geometry update and complex gain update. The path geometry update in the sensing time slot is represented by the latest path geometry directly provided by multi-target tracking, while in the communication time slot, it is predicted by combining the path geometry updated in the most recent sensing time slot with a kinematic model. For the complex gain update, the necessary complex gain prior is first constructed using the temporal correlation of the complex gain, and then the complex gain is updated by combining it with the pilot signal. Since the sensing time slot is much smaller than the communication time slot, the overhead for performing additional sensing is smaller; furthermore, since the required channel matrix dimension is... Traditionally, directly estimating the channel matrix is ​​computationally very complex. The proposed method, based on the path geometry provided by multi-target tracking under sensing time slots, transforms channel estimation into estimation of the complex gain vector. (dimension is) The update of the channel estimation significantly reduces the computational complexity and pilot overhead.

[0082] It should be noted that during path set updates in a synesthetic time slot, when the system enters a new synesthetic time slot, multi-target tracking updates the propagation path set and reconstructs the channel geometry. This can specifically include the following three types of operations: a) Path extinction and deletion When a path is detected to have disappeared, it is removed from the path set, and its corresponding path label, geometry, and complex path gain are deleted.

[0083] b) Path creation and introduction When a new path is detected, it is introduced into the path set and the corresponding geometry is introduced to initialize the complex path gain state.

[0084] c) Existence path and updates For persistent paths, the geometry is updated based on their path labels, while maintaining the continuous evolution of the complex path gain state without re-initialization.

[0085] Through the above structural update process, the system can periodically correct the path set and geometric structure, thereby restoring the consistency between the channel model and the real environment, and providing a reliable structural prior for channel tracing in subsequent communication slots.

[0086] In a preferred implementation, during the downlink phase of each communication slot, the geometry of each valid legacy path is obtained through prediction. The implementation method is as follows: The geometry of this effective legacy path, determined based on the most recent synesthesia time slot. Based on the kinematic model, the geometric structure of the valid old path in the current communication time slot is analyzed. Perform forecast updates.

[0087] In the downlink phase of each communication time slot, considering that path parameters in the communication time slot may change due to dynamic environments, as a preferred implementation, the geometry of the path estimated from the most recent sensing time slot can be further considered. And combine the kinematic model to analyze the geometric structure of the path in the current communication time slot. Perform prediction updates and update the complex gain of each propagation path in the current communication time slot.

[0088] In practical implementation, between two adjacent sensing time slots, it is assumed that the user and all scatterers are moving at approximately uniform linear speeds. Therefore, for each communication time slot between two sensing time slots, the most recent sensing time slot can be used. Estimated scatterer state vector The scatterer state vector of the current communication time slot Perform recursion. For the... For the scatterer, the kinematic model (i.e., uniform linear motion) is used to analyze the current scatterer. The prediction of the scatterer state vector under a communication time slot can be expressed as:

[0089] in,

[0090] Let represent the transition matrix for linear motion, and let have:

[0091] Therefore, based on the parameters of the most recent sensing time slot, the path parameters of each non-line-of-sight propagation path in the current communication time slot can be further predicted as follows: This parameter information can then be used to define the channel geometry.

[0092] This method further describes the dynamic characteristics of millimeter-wave wireless channels more accurately. Unlike existing methods that assume a fixed number of propagation paths or use fixed-dimensional state space modeling, this preferred method considers the dynamic changes of the propagation path set. It continuously tracks and recursively predicts the channel propagation path through multi-target tracking and further obtains the parameter information of the propagation path, thereby constructing a dynamically updated channel geometry structure that more accurately reflects the dynamic propagation characteristics of millimeter-wave wireless channels.

[0093] As a preferred implementation, the complex gain can be updated using the GB-LMMSE method, and the mean and variance of the updated complex gain for all paths in each time slot are specifically expressed as follows: The complex gains to be updated for all paths in the current time slot are combined into a vector, denoted as . The mean vector corresponding to the updated complex gain vectors of all paths in the current time slot. Covariance Matrix for:

[0094]

[0095] in, Represents the LMMSE gain matrix. , , Let these represent the prior mean vector and prior covariance matrix required for updating the complex gain vector, respectively. , , Indicates the number of valid paths. This represents the vectorized pilot received signal matrix. This indicates that the observation dictionary matrix is ​​obtained by concatenating the equivalent observation vectors corresponding to all paths involved in the complex gain. Indicates the noise variance. express An identity matrix of dimension 1 Indicates the number of receiving antennas at the base station. Indicates the pilot length.

[0096] Alternatively, as a preferred implementation, the method for determining the validity of each existing path is as follows: Including the continuous period preceding this communication time slot Each time slot meets one of the following conditions: a) The square of the mean of the updated complex gain of the path is less than a preset threshold. b) The variance of the updated complex gain of this path is greater than a preset threshold. If the path is frozen, it will not participate in prior construction and complex gain update in subsequent communication time slots until the subsequent synesthesia time slot is confirmed by multi-target tracking.

[0097] In this embodiment, path presence state detection is performed during the downlink phase of each communication time slot. Considering that the propagation path may change in a dynamic propagation environment due to factors such as scatterer obstruction, target emergence, or sensing errors, a temporary mismatch may occur between the constructed channel geometry and the actual channel. When the geometry cannot accurately interpret the current observations, directly performing channel tracking based on the original model may lead to unstable path complex gain estimation, or even estimation divergence. Therefore, to ensure the robustness of the channel tracking process, an adaptive tracking method is further proposed.

[0098] In specific implementation, for example in the first... Within a single communication time slot, some propagation paths may gradually become invalid due to obstruction or environmental changes. Retaining these paths in the channel model would consume limited pilot energy and reduce estimation stability. Therefore, the method in this embodiment is based on continuous propagation paths preceding the current communication time slot. The mean and variance of the path complex gain updated within each time slot are used to determine the path validity. Specifically, when the... The path is Within a communication time slot, one of the following conditions must be met: a) The square of the mean complex gain of the path is less than a small threshold. b) The variance of the complex gain of this path is greater than a certain threshold. , that is

[0099] If the above conditions are met, the path is frozen. The frozen path will not participate in channel prior construction and complex gain update in the following communication time slots, but its state will not be deleted immediately until the subsequent sensing time slot is confirmed by multi-target tracking.

[0100] After determining if a path has disappeared, the complex gain update operation is performed only on valid paths. By removing frozen paths, redundant or failed paths are avoided from diluting the limited pilot energy, thus ensuring that the complex gain update process only applies to the statistically reliable set of paths. Define the... The set of valid paths for each communication time slot is First, as mentioned earlier, we construct the path geometry prior and the complex gain prior required for complex gain updates.

[0101] For the set of valid paths First, the path geometry in the current communication time slot is predicted using the path geometry output from the most recent multi-target tracking in the communication time slot. .

[0102] Then, for the complex gain, using the current... The complex gain vector is reconstructed by taking the prior mean and variance of the complex gain for all paths. , and its prior mean vector and covariance matrix .

[0103] Then, based on the pilot reception matrix, the corresponding set of valid paths is obtained. LMMSE gain matrix in Observation dictionary matrix and set up a valid path set. Complex gain update within:

[0104]

[0105] By limiting updates to only the set of valid paths, we can avoid redundant paths diluting pilot energy, thereby improving the stability of channel estimation.

[0106] Alternatively, as a preferred implementation, path creation detection is also performed during the uplink phase of each communication time slot, implemented as follows: Calculate the current communication time slot residual matrix In the formula, Indicates communication time slot The pilot received signal matrix below, , representing the channel matrix recovered after updating with the complex gain of the effective old path. These represent the number of base station receiving antennas and the number of user transmitting antennas, respectively. Represents the equivalent observation vector of all valid old paths. The constructed observation dictionary matrix, the first Equivalent observation vector of an existing valid path , Indicates the first The geometric structure of an existing valid path. Indicates communication time slot The pilot transmit signal matrix below, Indicates communication time slot The updated complex gain vector is derived from all valid existing paths. Based on the residual matrix Calculate residual energy , Indicates the noise variance. Indicates the known pilot length. Represents the Frobenius norm of the matrix; if the residual energy exceeds a threshold This indicates the existence of a new non-line-of-sight propagation path. The residual matrix is ​​represented in the angular domain using a two-dimensional Fast Fourier Transform (2D-FFT), and peak detection is performed, retaining the path with the highest energy. Each corner domain component, according to The angle index corresponding to each angle component restores the arrival angle of the corresponding newborn non-line-of-sight propagation path. With departure angle , represented as , used to describe The complex gain of each angular domain component is solved using the least squares method, and the solution formula is: In the formula, Indicates communication time slot The complex gain of all newborn non-line-of-sight propagation paths is as follows. The complex gain vector formed by them This represents the receiver array steering vector of the base station. This represents the transmit array steering vector of the user equipment; Calculate the channel compensation term based on the complex gain of all newly generated non-line-of-sight propagation paths obtained from the solution. , The path geometry of this compensation item does not include any valid set of existing paths. The path in; The channel compensation term is combined with the channel matrix recovered after updating with the complex gain of the effective old path. Add them together to get the current communication time slot. The channel is used to complete adaptive tracking.

[0107] After completing the complex gain update of the effective path, the residual is calculated. This is used to detect model mismatch. The residual energy is defined as:

[0108] Once the residual energy exceeds the threshold , that is This indicates the possible existence of unmodeled or newly formed non-line-of-sight propagation paths. When residual detection is triggered, further sparse recovery is performed on the residual matrix, but this sparse recovery is only used for the current communication slot and does not participate in the prediction of path geometry and the construction of complex gain priors for subsequent communication slots. Considering that unmodeled energy is usually sparse in the angular domain, in one exemplary implementation, the channel residual can be represented in the angular domain by a two-dimensional fast Fourier transform, retaining only the angular domain components with the largest energy, and solving for the complex gain only on these components using the least squares method to achieve channel compensation. This compensation term is denoted as... It should be noted that the path geometry of this channel compensation term does not include any previous valid path sets. The path in the code. The above process can be expressed by the following formula. a) 2D-FFT corner domain representation and peak detection:

[0109] in These are standard FFT matrices, with dimensions of [dimensions to be filled in]. , , Represents the residual matrix The matrix representation in the angular domain, and the output of the stronger energy through peak detection. Each component, and by reconstructing the arrival and departure angles of the corresponding path based on its corresponding angle field index, can be represented as:

[0110] c) Solving for complex gain least squares: That is, based on the known angle of the new path and the pilot transmission signal, the complex gain of each new path is estimated using the least squares method. .

[0111] d) Residual channel recovery:

[0112] The channel compensation term has now been obtained. This can correct model mismatch caused by path generation, thereby maintaining the stability of channel estimation.

[0113] This preferred approach employs an adaptive mechanism based on validity detection and residual detection. Specifically, during the update of the complex gain of the non-line-of-sight propagation path, validity detection first determines whether the existence probability and prior mean of the propagation path are below a preset threshold; residual detection calculates the energy residual between the predicted channel and the actual observed channel. When the validity detection result indicates that a propagation path no longer meets the preset conditions, the propagation path is determined to have disappeared and is removed from the propagation path set; when the residual detection result indicates the existence of a new propagation path, the propagation path is added to the propagation path set. Through this mechanism, the propagation path set can be dynamically updated, thereby achieving adaptive tracking of the millimeter-wave channel.

[0114] See Figure 3This paper presents a comparison of the Normalized Mean Squared Error (NMSE) performance of different algorithms under various signal-to-noise ratio (SNR) conditions. It shows that the traditional Full-Dimensional Least Squares (FD-LS) method, due to its lack of utilization of channel geometry information, exhibits the worst performance across the entire SNR range. Pilot-less prediction methods, lacking observation updates, maintain a relatively high NMSE level. Sparse reconstruction methods such as Orthogonal Matching Pursuit (OMP) and Geometry-Based Sparse Bayesian Learning (GB-SBL) show improved performance in the mid-to-high SNR region, but the error reduction slows down in the high SNR region. In contrast, the proposed GB-LMMSE method exhibits superior NMSE performance across the entire SNR range, effectively suppressing noise effects, especially under low SNR conditions, and maintaining high estimation accuracy under high SNR conditions, thus verifying the effectiveness and robustness of the proposed method in channel estimation.

[0115] See Figure 4 It presents the impact of pilot repetition period, time correlation coefficient, and signal-to-noise ratio on the NMSE performance of the system. Figure 4 The figure above shows that under the closed-loop update mechanism, NMSE gradually accumulates within the pilot update interval and is corrected in the pilot time slot. When the pilot period increases, the error accumulation becomes more obvious, while the error remains at a high level when there is no pilot update. Figure 4 The intermediate plot shows that as the time correlation coefficient increases, the channel variation tends to level off, and the NMSE of each scheme decreases. Figure 4 The figure below shows that as the SNR increases, the overall NMSE of the system decreases, and the performance differences between different pilot period schemes gradually become apparent. These results verify that the proposed method has good stability and adaptability under different channel conditions.

[0116] See Figure 5 It provides a performance comparison before and after the introduction of the path structure adaptive update mechanism. Figure 5 The figure above shows the changes in NMSE before and after the effective path set update and channel compensation. It can be seen that after introducing the effective path set update and channel compensation mechanism, the channel estimation NMSE is significantly reduced, improving by about 7 dB. This indicates that the mechanism can effectively alleviate the model mismatch problem caused by path disappearance or new path emergence. Figure 5The figure below shows a comparison of the gain estimation results for each path. It can be seen that the gain estimates for the stable paths (ID1–ID5) remain largely consistent, while the vanished path ID6 and the newly formed path ID7 can be correctly suppressed or recovered after compensation. These results demonstrate that the proposed adaptive path structure update method can maintain the stability and robustness of channel estimation in dynamic propagation environments.

[0117] In summary, the method of this embodiment can solve the following technical problems: (1) Tracking dynamic path sets Existing channel estimation and tracking methods typically assume that the number of paths is fixed within the estimation window, consider only the quasi-static support set, and do not account for the impact of the random evolution of the path set. This embodiment considers the situation where the path set changes over time, probabilistically describing the appearance, existence, and disappearance of paths, and then introduces multi-target tracking to continuously track path parameter information.

[0118] (2) Reduce pilot overhead and computational complexity In large-scale millimeter-wave scenarios, frequent full channel structure estimation will lead to increased pilot resource consumption and computational burden. This embodiment proposes to introduce a perception-assisted low-complexity channel tracking method, which reduces pilot overhead and computational complexity by utilizing the sparse structure of the millimeter-wave channel, time correlation, and information provided by multi-target tracking.

[0119] (3) Improve system robustness when the path set changes. When propagation paths appear or disappear, the effective dimensions of the channel model need to be adjusted accordingly. Existing methods do not consider the structural changes caused by path creation and destruction, and lack corresponding robust handling mechanisms. This embodiment introduces an adaptive update method for channel path geometry, enabling the channel model dimensions to be dynamically adjusted according to the state of the path set and corrected in the event of structural mismatch, thereby improving system robustness.

[0120] Example 2 This application also relates to an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the method described above.

[0121] The electronic device can be a desktop computer, laptop, handheld computer, or cloud server, etc. The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The memory can be used to store computer programs and / or modules. The processor performs various functions of the electronic device by running or executing the computer programs and / or modules stored in the memory, and by accessing data stored in the memory.

[0122] The relevant technical solutions are the same as above, and will not be repeated here.

[0123] Example 3 This application also relates to a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method described above.

[0124] Specifically, the memory may include high-speed random access memory, as well as non-volatile memory, such as hard disks, RAM, plug-in hard disks, smart media cards (SMC), secure digital (SD) cards, flash cards, at least one disk storage device, flash memory device, or other volatile solid-state storage devices.

[0125] The relevant technical solutions are the same as above, and will not be repeated here.

[0126] Example 4 This application provides a computer program product or computer program that includes computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the steps of the method described in the above embodiments of this application.

[0127] The relevant technical solutions are the same as above, and will not be repeated here.

[0128] Those skilled in the art will readily understand that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A sensing-assisted millimeter-wave channel adaptive tracking method, characterized in that, include: A time-division duplex frame structure is constructed, and the frame structure is divided into a sensing time slot and a communication time slot, wherein multiple consecutive communication time slots are set between two sensing time slots; In the downlink phase of each sensing time slot: the base station transmits a sensing integrated signal and receives environmental echoes; based on the environmental echoes, it updates the environmental scatterer set information using a multi-target tracking algorithm; based on the updated scatterer set information, it constructs the channel geometry, including determining multiple non-line-of-sight propagation paths and path parameters for each path, wherein the path parameters include transmission angle, angle of arrival, time delay, and Doppler frequency shift; based on the path parameters of each path, it constructs the geometry of that path. In the uplink phase of each sensing time slot: the base station receives pilot signals sent by user equipment to update the complex gain of each non-line-of-sight propagation path. The update method is as follows: the geometric structure prior of each path and the statistical prior of the complex gain of the current sensing time slot are obtained. If the path is a newly created path in the downlink phase of the current sensing time slot, the statistical prior of the complex gain of the path is a weak Gaussian prior. If the path is an existing path, the complex gain obtained from the previous time slot is constructed based on the time correlation of the complex gain. Based on the geometric structure prior of each path and the statistical prior of the complex gain in the current sensing time slot, combined with the pilot signal, the updated complex gain of each path in the current sensing time slot is obtained. The geometric structure of each path and the updated complex gain constitute the channel of the current sensing time slot, completing the adaptive tracking of the time slot. In the downlink phase of each communication time slot: the validity of each old path is determined based on the mean and variance of the path complex gain after updating in multiple consecutive time slots, the valid old path of the current communication time slot is determined, and the geometric structure of each valid old path is obtained; In the uplink phase of each communication time slot: the base station receives pilot signals sent by user equipment to update the complex gain of each valid legacy path. The update method is as follows: the geometric structure prior of each valid legacy path and the complex gain statistical prior under the current communication time slot are obtained. The complex gain statistical prior is constructed by using the complex gain updated in the previous time slot based on the complex gain time correlation. Based on the geometric structure prior and the complex gain statistical prior of each valid legacy path, combined with the pilot signals, the updated complex gain of each valid legacy path under the current communication time slot is obtained. The geometric structure of each path and the updated complex gain constitute the channel of the current communication time slot, completing the adaptive tracking of the time slot.

2. The millimeter-wave channel adaptive tracking method as described in claim 1, characterized in that, The echo signal is formed by reflections from multiple scatterers in the environment; The method for updating the environmental scatterer set information based on environmental echoes using a multi-target tracking algorithm is as follows: The base station processes the received echo signal to obtain observation information related to the scatterer, including the distance, angle and Doppler information of the scatterer; Based on observation information, the state of multiple scatterers in the environment is estimated by a multi-target tracking algorithm to obtain the scatterer set information of the current time slot. The scatterer set information includes the position information and velocity information of each scatterer.

3. The millimeter-wave channel adaptive tracking method as described in claim 1, characterized in that, In the downlink phase of each communication slot, the geometry of each valid legacy path is obtained through prediction. The implementation method is as follows: The geometry of this effective legacy path, determined based on the most recent synesthesia time slot. Based on the kinematic model, the geometric structure of the valid old path in the current communication time slot is analyzed. Perform forecast updates.

4. The millimeter-wave channel adaptive tracking method as described in claim 1, characterized in that, The mean vector and covariance matrix of the updated complex gain for all paths in each time slot are specifically represented as follows: The complex gains to be updated for all paths in the current time slot are combined into a vector, denoted as . The mean vector corresponding to the updated complex gain vectors of all paths in the current time slot. Covariance Matrix for: in, Represents the LMMSE gain matrix. , , Let these represent the prior mean vector and prior covariance matrix required for updating the complex gain vector, respectively. , , Indicates the number of valid paths. This represents the vectorized pilot received signal matrix. This indicates that the observation dictionary matrix is ​​obtained by concatenating the equivalent observation vectors corresponding to all paths participating in subsequent complex gain updates. Indicates the noise variance. express An identity matrix of dimension 1 Indicates the number of receiving antennas at the base station. Indicates the pilot length.

5. The millimeter-wave channel adaptive tracking method as described in claim 1, characterized in that, The implementation method for determining the validity of each existing path is as follows: If the existing path is continuous before the included communication time slot Within a time slot, one of the following conditions must be met: a) The square of the mean of the updated complex gain of the path is less than a preset threshold. b) The variance of the updated complex gain of this path is greater than a preset threshold. If the path is frozen, it will not participate in prior construction and complex gain update in subsequent communication time slots until the subsequent synesthesia time slot is confirmed by multi-target tracking.

6. The millimeter-wave channel adaptive tracking method as described in claim 1, characterized in that, In the uplink phase of each communication time slot, path creation detection is also performed, implemented as follows: Calculate the current communication time slot Channel residual matrix In the formula, Indicates communication time slot The pilot received signal matrix below, , representing the channel matrix recovered after updating with the complex gain of the effective old path. These represent the number of base station receiving antennas and the number of user transmitting antennas, respectively. Represents the equivalent observation vector of all valid old paths. The constructed observation dictionary matrix, the first Equivalent observation vector of an existing valid path , Indicates the first The geometric structure of an existing valid path. Indicates communication time slot The pilot transmit signal matrix below, Indicates communication time slot The updated complex gain vector is derived from all valid existing paths. Based on the residual matrix Calculate residual energy , Indicates the noise variance. Indicates the known pilot length. Represents the Frobenius norm of the matrix; if the residual energy exceeds a threshold This indicates the existence of a new non-line-of-sight propagation path. The residual matrix is ​​represented in the angular domain using a two-dimensional fast Fourier transform, and peak detection is performed, retaining the path with the highest energy. Each corner domain component, according to The angle index corresponding to each angle component restores the arrival angle of the corresponding newborn non-line-of-sight propagation path. With departure angle , represented as , used to describe The complex gain of each angular domain component is solved using the least squares method, and the solution formula is: In the formula, Indicates communication time slot The complex gain of all newborn non-line-of-sight propagation paths is as follows. The complex gain vector formed by them This represents the receiver array steering vector of the base station. This represents the transmit array steering vector of the user equipment; Calculate the channel compensation term based on the complex gain of all newly generated non-line-of-sight propagation paths obtained from the solution. , The path geometry of this compensation item does not include any valid set of existing paths. The path in; Channel compensation terms The channel matrix recovered after updating with the complex gain of the effective old path Add them together to get the current communication time slot. The channel is used to complete adaptive tracking.

7. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method as described in any one of claims 1 to 6.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored computer program, wherein the computer program, when executed by a processor, controls the device on which the storage medium is located to perform the steps of the method as described in any one of claims 1 to 6.

9. A computer program product, comprising a computer program or instructions, characterized in that, When the computer program or instructions are executed by a processor, they implement the steps of the method as described in any one of claims 1 to 6.

Images