Six-dimensional movable antenna pose prediction and updating method based on stgcn

Abstract

The embodiment of the present disclosure relates to a six-dimensional movable antenna pose prediction and updating method based on STGCN, comprising: acquiring user motion state information and channel state information; constructing a user state vector according to the motion state information and the channel state information, and constructing a dynamic space graph structure based on the user state vector; performing feature extraction processing on the dynamic space graph structure by using a multi-layer graph convolution network to obtain a spatial feature; inputting the spatial feature into a gated recurrent unit network to output a predicted target pose; and determining a response time according to the predicted target pose and a current actual pose of the antenna, so as to update the antenna pose according to the response time and the predicted target pose when a yield constraint is met. The embodiment of the present disclosure realizes accurate decision of a predicted target pose at a future time, can adapt to a dynamic user position in advance to maintain a high-performance communication service, and guarantees communication continuity and stability of an intelligent antenna system.

Description

Technical Field

[0001] This disclosure relates to the field of 6G wireless communication smart antenna technology, and in particular to a method for predicting and updating the pose of a six-dimensional movable antenna based on STGCN. Background Technology

[0002] A six-dimensional (6D) movable antenna is a smart antenna system with six degrees of freedom. Its core lies in dynamically adapting to the user distribution in a wireless communication network by flexibly adjusting the spatial attitude of the distributed antenna surfaces, thereby improving network capacity and communication performance. This smart antenna system mainly consists of distributed antenna surfaces, a central processing unit, a retractable rotating rod, and motors. Each antenna surface can be a uniform planar array or other shapes, containing multiple directional antenna elements. It is connected to the central processing unit via the retractable rotating rod, which houses flexible wires for transmitting power and radio frequency signals. The motors at both ends are controlled by the central processing unit to achieve independent adjustment of the antenna surface's three-dimensional position and three-dimensional rotation angle.

[0003] Compared to traditional fixed-position antennas or fluid antennas that can only be adjusted in position, 6D movable antennas fully utilize spatial degrees of freedom by leveraging the six-dimensional controllability of each antenna surface. This significantly improves average user speed and system capacity in scenarios with non-uniform user distribution, providing an efficient solution for 6G wireless communication networks to adapt to dynamic user distribution. In a 6D movable antenna-assisted communication system, each antenna surface needs to dynamically adjust its three-dimensional position and three-dimensional rotation angle to adapt to user distribution.

[0004] Existing technologies for 6D mobile antenna-assisted communication typically employ a passively triggered movement judgment criterion, meaning that panel adjustments are only initiated when pilot resources are insufficient or the channel gain difference reaches a specific threshold. Due to the non-negligible response time of the physical mechanical movement of the motor, this passive reactive mechanism often causes antenna panel position updates to lag behind the actual migration of user hotspots. This fails to meet the real-time coverage requirements of 6D mobile antenna systems in highly dynamic environments, easily leading to service interruptions or sudden performance degradation.

[0005] Therefore, it is necessary to provide a new technical solution to improve one or more of the problems existing in the above solutions.

[0006] It should be noted that the information disclosed in the background section above is only used to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0007] The purpose of this disclosure is to provide a method for predicting and updating the pose of a six-dimensional movable antenna based on STGCN, accurately determining the predicted target pose at future moments, and ensuring communication continuity and the stability of the smart antenna system.

[0008] According to embodiments of this disclosure, a method for predicting and updating the pose of a six-dimensional movable antenna based on STGCN is provided, including: The user's motion state information and channel state information are acquired. The motion state information includes spatial position, motion parameters, and the deflection angle parameter of the user's current spatial position relative to the target antenna beam. The channel state information includes channel gain. A user state vector is constructed based on the motion state information and the channel state information, and a dynamic spatial graph structure is constructed based on the user state vector; The dynamic spatial graph structure is processed by a multi-layer graph convolutional network to extract spatial features. The spatial features are input into a gated recurrent unit network, which outputs the predicted target pose. The response time is determined based on the predicted target pose and the current actual pose of the antenna, so that the antenna pose is updated based on the response time and the predicted target pose when the revenue constraint is satisfied.

[0009] In an exemplary embodiment of this disclosure, the step of constructing a dynamic spatial graph structure based on the user state vector includes: Construct the node set and edge set of the dynamic spatial graph structure. The node set includes a central control node and multiple user nodes corresponding to the user state vector. The edge set is used to connect the node pairs in the node set. The spatial position and deflection angle parameters of each node in the node set are obtained from the user state vector, and an adjacency matrix is ​​constructed based on the spatial position and deflection angle parameters of the nodes. Based on the adjacency matrix, a degree matrix is ​​constructed, and the degree matrix is ​​used to normalize the adjacency matrix to obtain the dynamic spatial graph structure.

[0010] In an exemplary embodiment of this disclosure, the step of constructing an adjacency matrix based on the spatial location and deflection angle parameters corresponding to the nodes includes: The distance correlation component is determined based on the spatial location of the node, and the interference correlation component is determined based on the deflection angle parameter of the node. An adjacency matrix is ​​generated based on the distance correlation component and the interference correlation component.

[0011] In an exemplary embodiment of this disclosure, the adjacency matrix is ​​represented as: (1) in, Represents the adjacency matrix. , Indicates the user ID. Indicates a time index. Represents the distance correlation component. Represents the distance normalization factor. Indicates the number is users in Spatial location at a given time Indicates the number is users in Spatial location at a given time Indicates the interference correlation component. Indicates the number is users in The angle parameter of the spatial position at a given time relative to the target antenna beam. Indicates the number is users in The angle parameter of the spatial position at a given time relative to the target antenna beam.

[0012] In an exemplary embodiment of this disclosure, the user state vector is represented as: (2) in, Indicates the user ID. Indicates a time index. Indicates the number is users in Spatial location at a given time Indicates the speed of motion. Indicates the direction angle of motion. Indicates channel gain. This parameter represents the angle of deflection of the user's current spatial position relative to the target antenna beam.

[0013] In an exemplary embodiment of this disclosure, the step of inputting the spatial features into a gated recurrent unit network and outputting a predicted target pose includes: Based on the spatial characteristics at the current moment and the final hidden state at the previous moment, calculate the update gate and reset gate of the gated cyclic unit network; Obtain the absolute value of the time-series difference of the channel gain, and calculate the sensing gate of the gated recurrent unit network based on the absolute value of the time-series difference; The candidate hidden state of the gated recurrent unit network is modulated using the perception gate, and the final hidden state of the previous time step is fused with the modulated candidate hidden state using the update gate to obtain the final hidden state of the current time step. The final hidden state at the current moment is input into the fully connected layer of the gated recurrent unit network for processing, and the predicted target pose is obtained by filtering.

[0014] In an exemplary embodiment of this disclosure, in the step of modulating the candidate hidden state of the gated recurrent unit network using the perception gate, the candidate hidden state is modulated using formula (3): (3) in, This represents the modulated candidate hidden state. Indicates the attenuation amplification factor. This indicates that the door is being reset. This represents the Hadamard product. express The final hidden state of time. Indicates a sensor gate. This represents the bias vector. Represents the weight matrix. express Spatial characteristics at any given time.

[0015] In an exemplary embodiment of this disclosure, the sensing gate is represented as: (4) in, Indicates a sensor gate. This represents the Sigmoid activation function. Represents the weight matrix. The absolute value of the timing difference representing the channel gain. express The global spatial feature vector at time step [time]. express The global spatial feature vector at time step [time]. express Spatial characteristics of time, express The final hidden state of time. This represents the bias vector.

[0016] In an exemplary embodiment of this disclosure, determining the response time based on the predicted target pose and the current actual pose of the antenna includes: Calculate the spatial displacement and rotational displacement based on the predicted target pose and the current actual pose of the antenna; The first time consumption is determined based on the spatial displacement and the translational speed of the antenna, and the second time consumption is determined based on the rotational displacement and the angular velocity of the antenna. The maximum value between the first time consumption and the second time consumption is taken as the response time consumption.

[0017] In an exemplary embodiment of this disclosure, the benefit constraint is: (5) in, express Predict the target pose at any given time. This represents the predicted revenue corresponding to the predicted target pose. Indicates that the antenna is in The actual pose at any given moment. This represents the actual benefit corresponding to the actual pose. This represents the threshold.

[0018] The technical solution provided in this disclosure may include the following beneficial effects: In the embodiments of this disclosure, user spatial location, motion parameters, yaw angle parameters, and channel state information are integrated to construct a dynamic spatial graph structure adapted to the characteristics of a six-dimensional antenna, accurately capturing user spatial topological associations and temporal migration patterns. A multi-layer graph convolutional network is used to extract features from the dynamic spatial graph structure, aggregating features from adjacent nodes to accurately extract spatial features. Simultaneously, a gated recurrent unit network is used to capture the temporal migration patterns of hotspot areas from spatial features, accurately determining the predicted target pose for future moments. Furthermore, antenna pose updates are pre-controlled based on response time, and the rationality of pose update decisions is managed through revenue constraints. This allows for pre-adaptation to dynamic user locations to maintain high-efficiency communication services, ensuring communication continuity and the stability of the smart antenna system.

[0019] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this disclosure. Attached Figure Description

[0020] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this disclosure and, together with the description, serve to explain the principles of this disclosure. It is obvious that the drawings described below are merely some embodiments of this disclosure, and those skilled in the art can obtain other drawings based on these drawings without any inventive effort.

[0021] Figure 1 A flowchart illustrating the steps of a six-dimensional movable antenna pose prediction and update method based on STGCN in an exemplary embodiment of this disclosure is shown. Figure 2 This diagram illustrates a smart antenna system in an exemplary embodiment of the present disclosure. Figure 3 This diagram illustrates a spatiotemporal graph convolutional network in an exemplary embodiment of the present disclosure. Detailed Implementation

[0022] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, they are provided so that this disclosure will be more comprehensive and complete, and will fully convey the concept of the exemplary embodiments to those skilled in the art. The described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.

[0023] Furthermore, the accompanying drawings are merely illustrative of this disclosure and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and therefore repeated descriptions of them will be omitted. Some block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities may be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor devices and / or microcontroller devices.

[0024] This example implementation first provides a method for predicting and updating the pose of a six-dimensional movable antenna based on STGCN, referencing... Figure 1 As shown, the method may include the following steps: Step S101: Obtain user motion state information and channel state information. Motion state information includes spatial location, motion parameters, and the angle parameter of the user's current spatial location relative to the target antenna beam. Channel state information includes channel gain.

[0025] It should be noted that in step S101, user motion state information and channel state information can be obtained in real time from the smart antenna system with six degrees of freedom.

[0026] Step S102: Construct a user state vector based on the motion state information and the channel state information, and construct a dynamic spatial graph structure based on the user state vector.

[0027] It should be noted that the dynamic spatial graph structure is used to depict the spatial topological coupling relationships between users and between users and antenna panels.

[0028] Step S103: Use a multi-layer graph convolutional network to extract features from the dynamic spatial graph structure to obtain spatial features.

[0029] It should be noted that Graph convolutional networks (GCNs) are a type of efficient convolutional neural network suitable for graph structures. They combine the topological structure of a graph with node features and extract the relationships and feature information between nodes through graph convolution operations.

[0030] Step S104: Input the spatial features into the gated recurrent unit network and output the predicted target pose.

[0031] It should be noted that the gated recurrent unit (GRU) network is a commonly used gated recurrent neural network. It controls the flow of information through learnable gates, enabling it to better capture dependencies with large time step distances in a time series. This embodiment uses the aforementioned dynamic spatial graph structure, multi-layer graph convolutional network, and gated recurrent unit network to construct a spatiotemporal graph convolutional network for accurate prediction of the pose of a six-dimensional movable antenna.

[0032] Step S105: Determine the response time based on the predicted target pose and the current actual pose of the antenna, so as to update the antenna pose based on the response time and the predicted target pose when the revenue constraint is satisfied.

[0033] In the embodiments of this disclosure, a prediction algorithm based on Spatio-Temporal Graph Convolutional Networks (STGCN) is executed to fuse user spatial location, motion parameters, skew angle parameters, and channel state information to construct a dynamic spatial graph structure adapted to the characteristics of a six-dimensional antenna, accurately capturing user spatial topological associations and temporal migration patterns. A multi-layer graph convolutional network is used to extract features from the dynamic spatial graph structure, aggregating features of adjacent nodes to accurately extract spatial features. Simultaneously, a gated recurrent unit network is used to capture the migration patterns of hotspot areas in the temporal dimension from the spatial features, accurately determining the predicted target pose for future moments. Furthermore, antenna pose updates are pre-controlled based on response time, and the rationality of pose update decisions is managed through revenue constraints. This allows for pre-adaptation of dynamic user locations to maintain high-efficiency communication services, ensuring communication continuity and the stability of the smart antenna system.

[0034] The steps of the method described above in this example implementation will now be explained in more detail.

[0035] refer to Figure 2 As shown, the smart antenna system in this embodiment includes M movable antenna panels with six degrees of freedom. Each panel can independently control its position translation and three-dimensional rotation angle in three-dimensional space. The smart antenna system simultaneously serves N users in a dynamic moving state. After the smart antenna system is started, the central processing unit can collect the spatial position, motion parameters, channel gain, and relative angle of the user's current spatial position to the central axis of the main lobe of the serving panel beam in real time within a continuous time window. This data is used to construct a user state vector with deeply fused 6D physical features and a dynamic spatial graph structure with spatiotemporal correlation attributes.

[0036] In one embodiment, the motion parameters in step S101 include motion speed and motion direction angle. In step S101, after the smart antenna system is started, the central processing unit can collect the spatial position, motion speed, motion direction angle, channel gain, and relative deviation angle of the main lobe center axis of the user's current spatial position service panel beam in real time within a continuous time window. This data is used to construct a user state vector with deeply fused 6D physical features and a dynamic spatial graph structure with spatiotemporal correlation attributes.

[0037] In one embodiment, in the step of constructing a user state vector based on the motion state information and the channel state information, the user state vector is represented as: (2) in, Represents the user state vector. Indicates the user ID. Indicates a time index. Indicates the number is users in Spatial location at a given time Indicates the speed of motion. Indicates the direction angle of motion. Indicates channel gain. This parameter represents the angle of deflection of the user's current spatial position relative to the target antenna beam.

[0038] Understandably, in dynamic environments such as six-dimensional movable antenna scenarios, the highly directional nature of the six-dimensional antenna beam makes it impossible for conventional spatial diagram structures to accurately describe the nonlinear impact of antenna pose on the user channel. Therefore, this embodiment constructs a user state vector that deeply integrates 6D physical features. This user state vector not only includes spatial geometric features but also incorporates channel gain and angular characteristics representing beam alignment quality to accurately describe the nonlinear impact of antenna pose on the user channel, while also providing a decision-making basis for overcoming channel abrupt changes in dynamic environments.

[0039] In one embodiment, step S102, constructing a dynamic spatial graph structure based on the user state vector, includes the following steps: Step S1021: Construct the node set and edge set of the dynamic spatial graph structure; the node set includes the central control node and multiple user nodes corresponding to the user state vector; the edge set is used to connect the node pairs in the node set and describe the channel gain relationship between the nodes; Step S1022: Obtain the spatial position and deflection angle parameters of each node in the node set from the user state vector, and construct the adjacency matrix based on the spatial position and deflection angle parameters of the nodes; Step S1023: Construct a degree matrix based on the adjacency matrix, and normalize the adjacency matrix using the degree matrix to obtain the dynamic spatial graph structure.

[0040] It should be explained that the dynamic spatial graph structure in this embodiment can be represented as follows: , Represents a set of nodes. Represents the set of edges.

[0041] Furthermore, in step S1022, an adjacency matrix is ​​constructed based on the spatial location and deflection angle parameters corresponding to the nodes, including the following steps: The distance correlation component is determined based on the spatial location of the node, and the interference correlation component is determined based on the deflection angle parameter of the node. An adjacency matrix is ​​generated based on the distance correlation component and the interference correlation component.

[0042] The adjacency matrix generated based on the distance correlation component and the interference correlation component can be expressed as follows: (1) in, Represents the adjacency matrix. , Indicates the user ID. Indicates a time index. Represents the distance correlation component. Represents the distance normalization factor. Indicates the number is users in Spatial location at a given time Indicates the number is users in Spatial location at a given time Indicates the interference correlation component. Indicates the number is users in The angle parameter of the spatial position at a given time relative to the target antenna beam. Indicates the number is users in The angle parameter of the spatial position at a given time relative to the target antenna beam It should be noted that traditional graph construction methods only build adjacency matrices based on user physical distance, which cannot accurately characterize the interference coupling relationship between users under the beam coverage of a 6D directional antenna, and are difficult to adapt to the topological dynamics brought about by changes in antenna rotation angle, resulting in insufficient accuracy in spatial feature extraction. In this embodiment, unlike traditional graph construction methods based solely on distance, this embodiment uses an adjacency matrix construction rule based on beam-distance joint weighting to construct a dynamic spatial graph structure. It simultaneously characterizes the correlation between user physical spatial distance and the interference correlation of beam coverage, and then performs symmetric normalization processing through the degree matrix, so that the dynamic spatial graph structure accurately reflects the beam coverage overlap and interference relationship under a specific rotation angle of the six-dimensional antenna.

[0043] Furthermore, in step S1023, a degree matrix is ​​constructed based on the adjacency matrix, resulting in the following degree matrix: (6) in, Degree matrix, , Indicates the user ID. Indicates a time index. This represents the adjacency matrix.

[0044] In step S1023, the adjacency matrix is ​​normalized using the degree matrix, and the resulting dynamic space graph structure can be represented as follows: .

[0045] In one embodiment, in step S103, a two-layer graph convolutional network can be used to extract features from the dynamic spatial graph structure and output a spatial feature matrix: (7) in, Represents the spatial characteristic matrix, This indicates a modified linear unit activation function; The node feature matrix is ​​represented by the user state vector. Combining ; and This represents the trainable weight matrix of a spatiotemporal graph convolutional network. , The aforementioned weight matrix is ​​iteratively optimized through backpropagation during network training to adaptively learn the correlation between user space motion features and antenna channel features.

[0046] It should be noted that in step S103, the spatial features obtained by performing feature extraction on the dynamic spatial graph structure through a multi-layer graph convolutional network are the spatial feature matrix shown in formula (7). Example, see reference. Figure 3As shown, the graph convolutional network is the spatial feature extraction unit corresponding to step S103. Taking the dynamic spatial graph structure and user node feature matrix as input, it undergoes message passing, feature aggregation, and regularization processing to fuse and extract user spatial topology, motion, and channel features under the six-dimensional antenna beam coverage. The spatial feature matrix is ​​then output to the gated recurrent unit network, providing spatial feature support for antenna pose timing prediction. This embodiment uses the above feature extraction method to modify the linear unit activation function for nonlinear mapping of the aggregated spatial features, enhancing the fitting ability to dynamic and complex channel environments. It utilizes a two-layer graph convolutional network to extract features from the node feature matrix, aggregating features from adjacent nodes to capture inter-user spatial interference features limited by the 6D beam coverage range, overcoming the limitation of traditional methods that rely solely on isolated channel gain differences for judgment.

[0047] In one embodiment, step S104, inputting spatial features into a gated recurrent unit network and outputting a predicted target pose, includes the following steps: Step S1041: Based on the spatial features at the current time and the final hidden state at the previous time, calculate the update gate and reset gate of the gated recurrent unit network; Step S1042: Obtain the time-series differential absolute value of the channel gain, and activate the sensing gate of the gated recurrent unit network based on the time-series differential absolute value of the channel gain. Step S1043: Modulate the candidate hidden state of the gated recurrent unit network using a perception gate, and fuse the final hidden state of the previous time step with the modulated candidate hidden state using an update gate to obtain the final hidden state of the current time step. Step S1044: Input the final hidden state at the current moment into the fully connected layer of the gated recurrent unit network for processing, and filter to obtain the predicted target pose at the next moment.

[0048] It should be noted that in traditional gated recurrent units (GRUs), the update and reset gates of the GRU are equally sensitive to changes in all dimensions of the sequence when processing time series. In a 6D mobile antenna scenario, once a user crosses the beam boundary, the channel quality drops drastically. Traditional GRUs cannot respond quickly to these abrupt channel changes, leading to pose prediction lag and failing to meet the needs of highly dynamic user tracking. Therefore, this embodiment of the GRU network introduces a sensing gate for channel fading awareness, based on the original update and reset gates. The absolute value of the temporal difference of the channel gain is used as a high-priority input signal. This allows the GRU network to adaptively modulate the candidate hidden state through the sensing gate when the channel experiences severe fading, effectively solving the pose prediction lag problem caused by the inability of traditional GRUs to respond quickly to abrupt channel changes.

[0049] Furthermore, prior to step S1041, the spatial features of T consecutive time steps are... Input the gated recurrent unit network. In step S1041, calculate the update gate and reset gate of the gated recurrent unit network, where: The update gate is represented as , Indicates an update to the door. This represents the Sigmoid activation function. Represents the weight matrix. This represents the bias vector. express Spatial characteristics of time, express The final hidden state of the moment.

[0050] Reset door is indicated as , This indicates that the door is being reset. Represents the weight matrix. This represents the bias vector.

[0051] Furthermore, in step S1042, the absolute value of the timing difference of the channel gain is calculated. The absolute value of the timing difference of the channel gain As a high-priority input signal, the perception gate of the gated recurrent unit network is calculated, and this perception gate is represented as: (4) in, Indicates a sensor gate. This represents the Sigmoid activation function. Represents the weight matrix; The absolute value of the timing difference representing the channel gain. express The global spatial feature vector at time step [time]. express The global spatial feature vector at time step [time]. The spatial characteristic matrix output at time t Obtained by global average pooling based on user dimension. Depend on The spatial feature matrix output at each time step Obtained by global average pooling based on user dimensions; express Spatial characteristics of time, Represents the final hidden state at time [time]. This represents the bias vector. It should be noted that when the channel experiences severe fading, the sensing gate... The value approaches 1.

[0052] Furthermore, in step S1043, in the step of modulating the candidate hidden state of the gated recurrent unit network using the perception gate, the candidate hidden state is modulated using formula (3): (3) in, This represents the modulated candidate hidden state. Indicates the attenuation amplification factor. This indicates that the door is being reset. This represents the Hadamard product. express The final hidden state of time. Indicates a sensor gate. This represents the bias vector. Represents the weight matrix. express Spatial characteristics at any given time.

[0053] It should be explained that in step S1043, the reset gate is used to selectively filter historical hidden states. As a pre-screening step, it provides more accurate historical information input for the generation of candidate hidden states.

[0054] It should be noted that, This represents the hyperbolic tangent activation function. The mathematical expression is The hyperbolic tangent activation function Used to compress the calculated candidate hidden states to Within the interval, while ensuring numerical stability, nonlinear modeling capability is introduced into the gated cyclic unit network.

[0055] Furthermore, in step S1043, the final hidden state of the previous time step is fused with the modulated candidate hidden state using the update gate to obtain the final hidden state of the current time step: (8) in, This represents the Hadamard product. express The final hidden state of time. express The final hidden state of time. This represents the modulated candidate hidden state. This indicates an update to the door.

[0056] Furthermore, in step S1044, the final hidden state at the current moment is input into the fully connected layer of the gated recurrent unit network for processing, and the probability distribution of each antenna panel at the next moment (time t+1) is output through the Softmax function. , express The final hidden state of time. Represents the weight matrix. Represent the bias vector; and based on the probability distribution The predicted target pose is obtained through filtering. It should be noted that, to achieve discrete and controllable prediction of the panel pose, a six-dimensional pose codebook is preset in the model. Output the probability distribution of each antenna panel on the codebook at the next time step. Then, the codeword with the highest probability is selected as the target pose for prediction. .

[0057] The above embodiment predicts the target pose by introducing a gated recurrent unit network after the sensing gate, which can adaptively reduce the historical old topology state when the channel deteriorates sharply. The weights are adjusted to force the model to rely more on the latest extracted spatial features, thereby effectively overcoming the high dynamic tracking lag problem faced by the 6D movable antenna during position movement.

[0058] In one embodiment, step S105, determining the response time based on the predicted target pose and the current actual pose of the antenna, includes the following steps: Step S1051: Calculate the spatial displacement and rotational displacement based on the predicted target pose and the current actual pose of the antenna at the next moment. Step S1052: Determine the first time consumption based on the spatial displacement and the translational speed of the antenna, determine the second time consumption based on the rotational displacement and the angular velocity of the antenna, and take the maximum value of the first time consumption and the second time consumption as the response time consumption.

[0059] During the pose execution phase, this embodiment determines the mechanical delay by calculating the response time, so as to drive the motor in advance and achieve dynamic coverage with zero delay.

[0060] For example, let the current actual pose of the antenna be... The predicted target pose at the next moment is The spatial displacement is The rotational displacement is Based on the maximum translational speed of the motor used to adjust the posture. and maximum angular velocity Calculate the first time consumption Second time The response time is taken as: .

[0061] It should also be noted that the start time of pose adjustment can be calculated backward from the response time. , Indicates the current system time to the predicted target time The time interval. When the smart antenna system reaches this start time, it sends a drive command in advance, enabling the antenna panel to complete the pose update before the target user group arrives, achieving zero-latency coverage.

[0062] In one embodiment, in step S105, the revenue constraint is: (5) in, express Predict the target pose at any given time. This represents the predicted revenue corresponding to the predicted target pose. Indicates that the antenna is in The actual pose at any given moment. This represents the actual benefit corresponding to the actual pose. This represents the threshold.

[0063] Only when the benefit from the predicted target pose update meets the benefit constraint will it be implemented at startup time. Sending drive commands enables the motor to adjust the antenna panel's pose in advance, ensuring the rationality of movement decisions, as well as communication continuity, the stability of the smart antenna system, and energy efficiency.

[0064] It should be noted that after the antenna panel is in position, the user's main service panel is updated according to the principle of maximum channel gain. The smart antenna system feeds back new user motion state information and channel state information, and updates the user state vector to achieve closed-loop zero-lag coverage tracking.

[0065] It should also be noted that the advantages of the proposed method in high-dynamic scenarios can be demonstrated more significantly through the system resultant rate of the antenna system. The formula for calculating the system resultant rate is as follows: , This represents the signal-to-interference-plus-noise ratio for the k-th user. The predicted revenue in formula (5) This can be specifically understood as the combined rate of the system.

[0066] With target time Using the system composite rate as the objective function, the proposed method achieves average performance improvements of 10.6%, 21.0%, and 38.3% compared to traditional GCN+GRU, pure GRU, and passively triggered methods in low-dynamic scenarios (user movement speed ≤ 1 m / s), respectively. In high-dynamic scenarios (user movement speed 5~10 m / s), it achieves average performance improvements of 34.5%, 62.4%, and 130.1% compared to the aforementioned three traditional methods, respectively. The results demonstrate that the proposed method achieves early adaptation of the antenna pose at the target time through accurate pose prediction, outperforming traditional schemes in system composite rate, with particularly significant advantages in high-dynamic scenarios.

[0067] Regarding the system in the above embodiments, the specific manner in which each unit performs operations has been described in detail in the embodiments related to the method, and will not be elaborated here.

[0068] It should be noted that although several units of the system for executing actions have been mentioned in the detailed description above, this division is not mandatory. In fact, according to embodiments of this disclosure, the features and functions of two or more units described above can be embodied in one unit. Conversely, the features and functions of one unit described above can be further divided and embodied by multiple units. Some or all of the units can be selected to achieve the purpose of this disclosure according to actual needs. Those skilled in the art can understand and implement this without any inventive effort.

[0069] Other embodiments of this disclosure will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not disclosed herein. The specification and examples are to be considered exemplary only, and the true scope and spirit of this disclosure are indicated by the appended claims.

Claims

1. A method for predicting and updating the pose of a six-dimensional movable antenna based on STGCN, characterized in that, include: The user's motion state information and channel state information are acquired. The motion state information includes spatial position, motion parameters, and the deflection angle parameter of the user's current spatial position relative to the target antenna beam. The channel state information includes channel gain. A user state vector is constructed based on the motion state information and the channel state information, and a dynamic spatial graph structure is constructed based on the user state vector; The dynamic spatial graph structure is processed by a multi-layer graph convolutional network to extract spatial features. The spatial features are input into a gated recurrent unit network, which outputs the predicted target pose. The response time is determined based on the predicted target pose and the current actual pose of the antenna, so that the antenna pose is updated based on the response time and the predicted target pose when the revenue constraint is satisfied.

2. The method for predicting and updating the pose of a six-dimensional movable antenna based on STGCN according to claim 1, characterized in that, The construction of the dynamic spatial graph structure based on the user state vector includes: Construct the node set and edge set of the dynamic spatial graph structure. The node set includes a central control node and multiple user nodes corresponding to the user state vector. The edge set is used to connect the node pairs in the node set. The spatial position and deflection angle parameters of each node in the node set are obtained from the user state vector, and an adjacency matrix is ​​constructed based on the spatial position and deflection angle parameters of the nodes. Based on the adjacency matrix, a degree matrix is ​​constructed, and the degree matrix is ​​used to normalize the adjacency matrix to obtain the dynamic spatial graph structure.

3. The method for predicting and updating the pose of a six-dimensional movable antenna based on STGCN according to claim 2, characterized in that, The step of constructing the adjacency matrix based on the spatial location and deflection angle parameters of the nodes includes: The distance correlation component is determined based on the spatial location of the node, and the interference correlation component is determined based on the deflection angle parameter of the node. An adjacency matrix is ​​generated based on the distance correlation component and the interference correlation component.

4. The method for predicting and updating the pose of a six-dimensional movable antenna based on STGCN according to claim 2 or 3, characterized in that, The adjacency matrix is ​​represented as: （1） in, Represents the adjacency matrix. , Indicates the user ID. Indicates a time index. Represents the distance correlation component. Represents the distance normalization factor. Indicates the number is users in Spatial location at a given time Indicates the number is users in Spatial location at a given time Indicates the interference correlation component. Indicates the number is users in The angle parameter of the spatial position at a given time relative to the target antenna beam. Indicates the number is users in The angle parameter of the spatial position at a given time relative to the target antenna beam.

5. The method for predicting and updating the pose of a six-dimensional movable antenna based on STGCN according to claim 1, characterized in that, The user state vector is represented as follows: （2） in, Indicates the user ID. Indicates a time index. Indicates the number is users in Spatial location at a given time Indicates the speed of motion. Indicates the direction angle of motion. Indicates channel gain. This parameter represents the angle of deflection of the user's current spatial position relative to the target antenna beam.

6. The method for predicting and updating the pose of a six-dimensional movable antenna based on STGCN according to claim 1, characterized in that, The step of inputting the spatial features into a gated recurrent unit network and outputting the predicted target pose includes: Based on the spatial characteristics at the current moment and the final hidden state at the previous moment, calculate the update gate and reset gate of the gated cyclic unit network; Obtain the absolute value of the time-series difference of the channel gain, and calculate the sensing gate of the gated recurrent unit network based on the absolute value of the time-series difference; The candidate hidden state of the gated recurrent unit network is modulated using the perception gate, and the final hidden state of the previous time step is fused with the modulated candidate hidden state using the update gate to obtain the final hidden state of the current time step. The final hidden state at the current moment is input into the fully connected layer of the gated recurrent unit network for processing, and the predicted target pose is obtained by filtering.

7. The method for predicting and updating the pose of a six-dimensional movable antenna based on STGCN according to claim 6, characterized in that, In the step of modulating the candidate hidden state of the gated recurrent unit network using the perception gate, the candidate hidden state is modulated using formula (3): （3） in, This represents the modulated candidate hidden state. Indicates the attenuation amplification factor. This indicates that the door is being reset. This represents the Hadamard product. express The final hidden state of time. Indicates a sensor gate. This represents the bias vector. Represents the weight matrix. express Spatial characteristics at any given time.

8. The method for predicting and updating the pose of a six-dimensional movable antenna based on STGCN according to claim 6, characterized in that, The sensing gate is represented as follows: （4） in, Indicates a sensor gate. This represents the Sigmoid activation function. Represents the weight matrix. The absolute value of the timing difference representing the channel gain. express The global spatial feature vector at time step [time]. express The global spatial feature vector at time step [time]. express Spatial characteristics of time, express The final hidden state of time. This represents the bias vector.

9. The method for predicting and updating the pose of a six-dimensional movable antenna based on STGCN according to claim 1, characterized in that, The step of determining the response time based on the predicted target pose and the current actual pose of the antenna includes: Calculate the spatial displacement and rotational displacement based on the predicted target pose and the current actual pose of the antenna; The first time consumption is determined based on the spatial displacement and the translational speed of the antenna, and the second time consumption is determined based on the rotational displacement and the angular velocity of the antenna. The maximum value between the first time consumption and the second time consumption is taken as the response time consumption.

10. The method for predicting and updating the pose of a six-dimensional movable antenna based on STGCN according to claim 1, characterized in that, The revenue constraint is: （5） in, express Predict the target pose at any given time. This represents the predicted revenue corresponding to the predicted target pose. Indicates that the antenna is in The actual pose at any given moment. This represents the actual benefit corresponding to the actual pose. This represents the threshold.

Images