Positioning method, electronic equipment and storage medium
By receiving and analyzing channel measurement data from direct links and relay reflected signals, an end-to-end mapping from channel characteristics to spatial location is established, solving the problems of high cost and low accuracy in wireless positioning technology and achieving efficient and reliable user positioning.
Patent Information
- Application Number
- CN202610018382.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-01-07
- Publication Date
- 2026-04-21
AI Technical Summary
Existing wireless positioning technologies and methods have high hardware deployment and data collection costs, making them difficult to promote on a large scale, and their positioning accuracy and reliability are difficult to guarantee.
By receiving complete channel measurement data, including both direct link signals and signals reflected by relay reflection devices, multidimensional features are extracted. Then, an end-to-end mapping relationship between channel features and spatial location coordinates is established using a positioning model. This reduces the deployment and maintenance costs of relay reflection devices and improves positioning accuracy and reliability.
It achieves high-precision user positioning, reduces hardware deployment and maintenance costs, and improves the reliability and accuracy of positioning.
Smart Images
Figure CN121899746A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of wireless positioning technology, and in particular to a positioning method, electronic device and storage medium. Background Technology
[0002] Driven by the new wave of information technologies such as the Internet of Things and smart cities, high-precision indoor positioning technology has become a critical infrastructure. From personnel and asset management and autonomous robot navigation to augmented reality and emergency rescue services, the demand for stable and reliable positioning capabilities is becoming increasingly urgent.
[0003] Currently, wireless signal-based positioning typically employs the following methods: Geometric positioning based on channel state information, which increases the observable reflection paths in space by deploying multiple physically reconfigurable intelligent surfaces (RIS), and uses geometric parameters such as angle of arrival and time of arrival for calculation; and fingerprint positioning based on received signal strength, which obtains multiple sets of received signal strength values by changing the phase state of the RIS, and constructs a fingerprint database for matching.
[0004] However, the above methods are too costly in terms of hardware deployment and data acquisition, making it difficult to promote and apply them on a large scale in real-world scenarios. Summary of the Invention
[0005] This application provides a positioning method, electronic device, and storage medium for achieving high-precision positioning of users, and can also effectively reduce hardware deployment and data acquisition costs.
[0006] In a first aspect, this application provides a positioning method, comprising: acquiring channel measurement data of a user to be positioned; wherein the channel measurement data includes direct link signals and reflected link signals; wherein the reflected link signals include signals reflected via relay reflection equipment; obtaining multidimensional features of the channel based on the channel measurement data; wherein the multidimensional features of the channel are used to reflect the physical environment information of signal propagation; obtaining the spatial location coordinates of the user to be positioned based on the multidimensional features of the channel and a positioning model; wherein the positioning model is used to establish an end-to-end mapping relationship between the multidimensional features and the spatial location coordinates.
[0007] The technical solution provided in this application offers at least the following advantages: By receiving complete channel measurement data, including both direct link signals and signals reflected via relay reflection devices, and extracting multi-dimensional features that reflect the physical environment of signal propagation, an end-to-end mapping relationship from channel features to spatial coordinates is established using a positioning model, thereby achieving accurate positioning of the user to be located. This method can reduce the deployment and maintenance costs of multiple relay reflection devices and also improve the positioning accuracy and reliability of the user to be located.
[0008] One possible implementation is that the relay reflection device includes any of the following: a single physical relay reflection device; a single physical relay reflection device and multiple virtual relay reflection devices; or multiple physical relay reflection devices.
[0009] Another possible implementation involves generating multiple virtual relay reflectors by adjusting the phase of a single physical relay reflector and equivalently generating multiple virtual relay reflectors within a preset range around the single physical relay reflector; wherein the multiple virtual relay reflectors operate in different time slots.
[0010] Another possible implementation method is to obtain the spatial coordinates of the user to be located based on the multidimensional features of the channel and the positioning model, including: stacking the multidimensional features of the channel in a fixed channel order to form a channel feature tensor; inputting the channel feature tensor into the positioning model to obtain the spatial coordinates of the user to be located.
[0011] Another possible implementation, the localization model, includes: an input layer, a convolutional layer, a fully connected layer, and an output layer; the input layer is used to receive the channel feature tensor; the convolutional layer is used to extract a high-dimensional feature map of the channel feature tensor; the high-dimensional feature map is used to reflect the spatial correlation and path structure information of the channel feature tensor; the fully connected layer is used to process the high-dimensional feature map and reduce its dimensionality to a low-dimensional feature vector; the output layer is used to map the low-dimensional feature vector to spatial location coordinates.
[0012] Another possible implementation is to locate the hyperparameters of the model as preferred hyperparameters; the preferred hyperparameters are the parameters optimized using a greedy search strategy.
[0013] Another possible implementation involves hyperparameters including at least one of the following: number of convolutional layers, kernel size, number of fully connected layer nodes, batch size, learning rate, dropout rate, and number of virtual relay reflection devices.
[0014] Another possible implementation involves the channel's multidimensional characteristics including at least one of the following: received signal strength, path direction angle, path type, phase information, and propagation distance.
[0015] Secondly, this application provides a positioning device, comprising: an acquisition unit and a processing unit; the acquisition unit is configured to acquire channel measurement data of a user to be positioned; wherein the channel measurement data includes direct link signals and reflected link signals; wherein the reflected link signals include signals reflected via relay reflection equipment; the processing unit is configured to obtain multi-dimensional features of the channel based on the channel measurement data; wherein the multi-dimensional features of the channel are used to reflect the physical environment information of signal propagation; the processing unit is configured to obtain the spatial location coordinates of the user to be positioned based on the multi-dimensional features of the channel and a positioning model; wherein the positioning model is used to establish an end-to-end mapping relationship between the multi-dimensional features and the spatial location coordinates.
[0016] One possible implementation is an acquisition unit that has the ability to adjust the phase of a single physical relay reflector and to generate multiple virtual relay reflectors equivalently within a preset range around the single physical relay reflector; wherein the multiple virtual relay reflectors operate in different time slots.
[0017] Another possible implementation is a processing unit that stacks the multidimensional features of the channel into a channel feature tensor in a fixed channel order; inputs the channel feature tensor into a positioning model to obtain the spatial location coordinates of the user to be located.
[0018] Thirdly, this application provides an electronic device comprising: a processor and a memory; the memory storing processor-executable instructions; when the processor is configured to execute the instructions, causing the electronic device to implement the method of the first aspect described above.
[0019] Fourthly, this application provides a computer-readable storage medium comprising: computer software instructions; which, when executed in an electronic device, cause the electronic device to implement the method described in the first aspect.
[0020] Fifthly, this application provides a computer program product comprising a computer program; when the computer program is run in an electronic device, the electronic device performs the method described in the first aspect.
[0021] The beneficial effects of the second to fifth aspects mentioned above are described in the corresponding description of the first aspect and will not be repeated here. Attached Figure Description
[0022] Figure 1 A schematic diagram of the system architecture of a positioning method provided in this application; Figure 2 A schematic diagram of signal propagation provided for this application; Figure 3 A schematic diagram of another signal propagation method provided for this application; Figure 4A flowchart of a positioning method provided in this application; Figure 5 A schematic diagram illustrating the composition of a positioning model provided in this application; Figure 6 A flowchart of a training localization model provided in this application; Figure 7 A flowchart of another positioning method provided in this application; Figure 8 A schematic diagram of the composition of a positioning device provided in this application; Figure 9 This is a schematic diagram of the composition of an electronic device provided in this application. Detailed Implementation
[0023] The positioning method, electronic device, and storage medium provided in this application will now be described in detail with reference to the accompanying drawings.
[0024] In this article, the term "and / or" is merely a description of the relationship between related objects, indicating that there can be three relationships. For example, A and / or B can represent three situations: A exists alone, A and B exist simultaneously, and B exists alone.
[0025] The terms "first" and "second," etc., used in the specification and drawings of this application are used to distinguish different objects or to distinguish different treatments of the same object, rather than to describe a specific order of objects.
[0026] Furthermore, the terms "comprising" and "having," and any variations thereof, used in the description of this application are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not limited to the steps or units listed, but may optionally include other steps or units not listed, or may optionally include other steps or units inherent to such process, method, product, or apparatus.
[0027] It should be noted that in the embodiments of this application, the words "exemplary" or "for example" are used to indicate examples, illustrations, or explanations. Any embodiment or design scheme described as "exemplary" or "for example" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design schemes. Specifically, the use of the words "exemplary" or "for example" is intended to present the relevant concepts in a specific manner.
[0028] To facilitate a clear description of the technical solutions of the embodiments of this application, the terms "first" and "second" are used in the embodiments of this application to distinguish the same or similar items with essentially the same function and effect. Those skilled in the art can understand that the terms "first" and "second" are not intended to limit the quantity or execution order.
[0029] In the description of this application, unless otherwise stated, "a plurality of" means two or more.
[0030] With the acceleration of the Internet of Things and digitalization, wireless positioning technology can be applied to the following scenarios: real-time tracking and scheduling of robots, automated guided vehicles (AGVs), and high-value assets in smart factories; precise management of personnel and goods in large warehouses; indoor navigation and location services in shopping malls, airports, and museums; location management of medical equipment and patients in hospitals; and emergency rescue and personnel positioning in environments such as underground parking lots, mines, and tunnels.
[0031] Currently, fingerprint-based localization methods based on received signal strength involve several approaches: First, a dense grid of reference points is established within the target area. At each reference point, signal strength values from multiple fixed access points are collected through on-site measurements or channel modeling, constructing a fingerprint database corresponding to location and signal strength vectors. During online localization, the device to be located measures the signal strength received from each access point in real time, matches this measurement vector with records in the fingerprint database, and uses the nearest neighbor algorithm to find the most similar reference point, thus obtaining the final location estimate. Geometric localization methods based on channel state information involve deploying multiple physical RIS devices to acquire multi-dimensional channel characteristics of the signal, such as angle of arrival, time difference of arrival, or phase information. By measuring signal parameters from at least three known location reference nodes, and combining the geometric triangulation principle or maximum likelihood estimation algorithm, the spatial coordinates of the target are calculated. Deep learning-based localization methods input wireless signal features into a deep learning model. By training the model, a nonlinear mapping relationship from signal features to location coordinates is established. The trained model can directly infer location information from new signal measurements.
[0032] However, the aforementioned positioning methods have several drawbacks: First, fingerprint positioning requires significant manpower and time for extensive on-site data collection during the offline phase, resulting in high construction and maintenance costs. Furthermore, changes in environmental layout, personnel movement, or electromagnetic interference can quickly render the existing fingerprint database ineffective, leading to a sharp decline in positioning performance. Second, geometric positioning typically requires the deployment of numerous physical RIS devices and the construction of a sufficient number of reflection paths, resulting in excessively high hardware costs and deployment complexity, thus limiting large-scale application. Third, the generalization ability and robustness of deep learning-based positioning models are difficult to guarantee, making it challenging to ensure the accuracy of predicted positioning.
[0033] To address the aforementioned technical problems, this application provides a positioning method. The method involves receiving complete channel measurement data, including both direct-link signals and signals reflected via relay reflection devices, extracting multi-dimensional features reflecting the physical environment of signal propagation, and then using a positioning model to establish an end-to-end mapping relationship from channel features to spatial coordinates, thereby achieving accurate positioning of the user to be located. This method can reduce the deployment and maintenance costs of multiple relay reflection devices and improve the positioning accuracy and reliability of the user.
[0034] The embodiments provided in this application will now be described in detail with reference to the accompanying drawings.
[0035] The positioning method provided in this application can be applied to, for example... Figure 1 In the system architecture shown. For example... Figure 1 As shown, the system architecture includes: access point device 10, relay reflection device 20, and electronic device 30. Access point device 10 and relay reflection device 20 are interconnected; relay reflection device 20 and electronic device 30 are interconnected.
[0036] In some embodiments, the access point device 10 is used to collect channel measurement data. For example, the access point device 10 can be a fixed node device providing access services in a wireless communication network, responsible for transmitting probe signals and receiving echo signals.
[0037] For example, the acquired channel measurement data includes the first echo signal; such as Figure 2 As shown, the first echo signal is the direct link signal from the target detection signal to the access point device 10 after passing through the user to be located. Wherein, as... Figure 2 As shown in (a), the target detection signal is a detection signal transmitted by access point device 10 and reflected to the user to be located by relay reflection device; or, as ... Figure 2 As shown in (b), the target detection signal is a detection signal transmitted by the access point device 10 to the user to be located.
[0038] For example, the acquired channel measurement data includes a second echo signal; wherein, as Figure 3 As shown, the second echo signal is the reflected link signal of the target detection signal after passing through the user to be located, reflected by the relay reflection device to the access point device 101. Among them, as... Figure 3 As shown in (a), the target detection signal is a detection signal transmitted by access point device 10 to the user to be located; or, as... Figure 3 As shown in (b), the target detection signal is a detection signal transmitted by the access point device 10 and reflected to the user to be located by the relay reflection device.
[0039] For example, the access point device 10 may be a wireless local area network access point, a cellular communication base station, or an anchor node dedicated to positioning, etc. This application embodiment does not limit the specific device form of the access point device 10.
[0040] In some embodiments, the relay reflection device 20 is used to wirelessly receive and forward signals.
[0041] For example, the relay reflection device 20 can be a RIS device, which includes a two-dimensional array composed of a large number of subwavelength electromagnetic units. The reflection amplitude and / or reflection phase of each electromagnetic unit can be independently and dynamically adjusted by an external control signal, thereby enabling programmable beamforming, signal enhancement, or path reconstruction of the incident wireless signal. This application does not limit the specific device form of the relay reflection device 20.
[0042] In some embodiments, the electronic device 30 is used to receive channel measurement data collected by the access point device 10; then, it extracts and characterizes multidimensional features of the physical environment of signal propagation from the channel measurement data; finally, it inputs the multidimensional feature tensor into the positioning model to obtain the spatial location coordinates of the user to be located.
[0043] In some embodiments, the electronic device 30 is further configured to generate a phase modulation command for the relay reflection device 20 according to a preset phase configuration strategy or real-time channel state feedback; and send the phase modulation command to the relay reflection device 20 to dynamically adjust the phase of the reflection unit of the relay reflection device 20.
[0044] It should be noted that the system architecture described in the embodiments of this application is for the purpose of more clearly illustrating the technical solutions of the embodiments of this application, and does not constitute a limitation on the technical solutions provided in the embodiments of this application. As those skilled in the art will know, with the evolution of system architecture, the technical solutions provided in the embodiments of this application are also applicable to similar technical problems.
[0045] See Figure 4 This is a flowchart of a positioning method provided in this application. Figure 4 As shown, the positioning method provided in this application specifically includes the following steps S201~S203.
[0046] S201. Obtain channel measurement data of the user to be located.
[0047] The channel measurement data includes direct link signals and reflected link signals. Reflected link signals include signals reflected by relay reflection equipment.
[0048] In some embodiments, the relay reflection device includes any of the following: a single physical relay reflection device; a single physical relay reflection device and multiple virtual relay reflection devices; or multiple physical relay reflection devices.
[0049] In this context, a single physical relay reflection device refers to a physical device actually deployed in physical space. A single physical relay reflection device and multiple virtual relay reflection devices refer to a relay reflection device actually deployed in physical space, where the reflected phase is adjusted to generate equivalent reflected signals at different spatial locations at the signal level. Multiple physical relay reflection devices refer to two or more physical relay reflection devices actually deployed in physical space.
[0050] In some embodiments, multiple virtual relay reflection devices are generated by adjusting the phase of a single physical relay reflection device to equivalently generate multiple virtual relay reflection devices within a preset range around the single physical relay reflection device.
[0051] Multiple virtual relay reflection devices operate in different time slots.
[0052] One feasible approach is to establish a direct link model between the user to be located and the access point (AP) device, and obtain the direct link signal. This includes: controlling the access point (AP) device to emit a detection signal, having the user to be located receive the detection signal from the AP device, and measuring and recording at least one of the following: received power, path loss, phase information, and time of arrival, to obtain the direct link signal.
[0053] For example, let the location of the AP device be... The location of the user to be located is The Euclidean distance between the AP device and the user to be located is According to wireless propagation theory, the channel gain of this direct-fire link can be modeled as the product of the path loss factor and the propagation phase, thus obtaining the propagation signal of the direct-fire link.
[0054] As one possible approach, by adjusting the phase of its reflecting unit, multiple virtual RIS devices are equivalently generated within its surrounding radius to obtain the reflected link signal. This includes: controlling the relay reflecting device (such as the RIS device) to sequentially switch the phase of the reflecting unit within a continuous time window according to a preset phase configuration codebook; for each phase of the reflecting unit, controlling the access point device to transmit a detection signal, which is reflected by the relay reflecting device and received by the user to be located; the user to be located measures the received reflected signal each time and records at least one of the received power, equivalent path phase offset, and channel impulse response under each phase of the reflecting unit to obtain the reflected link signal corresponding to the phase of different reflecting units.
[0055] For example, suppose the location of a single physical RIS device is... In its surrounding radius k virtual RIS points are generated internally by uniformly distributing angles. Each virtual RIS point is controlled by adjusting the phase factor. At the same time, the channel parameters between each virtual RIS point and the user to be located and the AP device are calculated on their respective propagation paths to obtain the signal reflected by the relay reflection device.
[0056] S202. Based on channel measurement data, obtain the multidimensional characteristics of the channel.
[0057] Among them, the multidimensional characteristics of the channel are used to reflect the physical environment information of signal propagation.
[0058] In some embodiments, the multidimensional characteristics of the channel include at least one of the following: received signal strength, path direction angle, path type, phase information, and propagation distance.
[0059] Received signal strength refers to the power level of the wireless signal when it reaches the receiver during propagation. It reflects the degree of signal attenuation and is closely related to factors such as propagation distance, obstacle obstruction, and reflective surface material. Path direction angle indicates the geometric direction of signal propagation. By measuring the difference in direction angles for different paths, spatial orientation constraints can be provided for positioning. Path type is used to distinguish reflection paths from different relay reflection devices (or different virtual configurations). Phase information provides precise positioning information at the subwavelength level. Propagation distance refers to the actual distance the signal travels from the transmitter to the receiver.
[0060] In some embodiments, after obtaining the direct link signal and the reflected link signals of each virtual RIS, multidimensional features are calculated for each signal link.
[0061] For example, for each signal link (such as 1 direct path and k reflection paths), the received signal strength is obtained by converting the received power of each path into a value in dBm.
[0062] S203. Based on the multi-dimensional features of the channel and the positioning model, the spatial coordinates of the user to be located are obtained.
[0063] Among them, the localization model is used to establish an end-to-end mapping relationship between multidimensional features and spatial location coordinates.
[0064] In some embodiments, the multidimensional features of the channel are stacked in a fixed channel order to form a channel feature tensor; the channel feature tensor is input into the positioning model to obtain the spatial location coordinates of the user to be located.
[0065] In some embodiments, such as Figure 5As shown, the localization model includes: an input layer, a convolutional layer, a fully connected layer, and an output layer.
[0066] In some embodiments, the input layer is used to receive channel feature tensors.
[0067] For example, the input layer of the localization model standardizes or normalizes multidimensional features to ensure that features of different dimensions are within the same numerical range.
[0068] In some embodiments, the convolutional layer is used to extract a high-dimensional feature map of the channel feature tensor; the high-dimensional feature map is used to reflect the spatial correlation and path structure information of the channel feature tensor.
[0069] For example, a multi-layer convolutional neural network structure is used to extract high-dimensional feature maps of multi-dimensional features. Each convolutional layer performs sliding calculations on the multi-dimensional features through multiple convolutional kernels to generate new feature maps. Through the stacking of multiple convolutional layers, the localization model can progressively extract higher-level abstract information from the raw channel features at lower levels (e.g., lower-level convolutions may extract local features related to path loss, while higher-level convolutions may capture the correlation between different virtual RIS reflection links).
[0070] For example, after the convolutional layer, a non-linear activation function (such as Leaky ReLU) can be introduced to increase the representational power of the localization model.
[0071] For example, Leaky ReLU satisfies the following formula:
[0072] in, The output value after activation; Input value; It is a small constant (e.g., =0.01).
[0073] In some embodiments, a fully connected layer is used to process high-dimensional feature maps and reduce them to low-dimensional feature vectors.
[0074] For example, a fully connected layer flattens a high-dimensional feature map, compressing its spatial dimension into a one-dimensional vector. Fully connected layers can integrate feature information scattered across different paths and channels.
[0075] In some embodiments, the output layer is used to map low-dimensional feature vectors to spatial location coordinates.
[0076] For example, the output layer can simultaneously utilize integrated information from multiple paths to perform location prediction for the user to be located based on a one-dimensional vector.
[0077] In some embodiments, the hyperparameters of the localization model are preferred hyperparameters; the preferred hyperparameters are parameters optimized using a greedy search strategy.
[0078] In some embodiments, hyperparameters include at least one of the following: number of convolutional layers, kernel size, number of fully connected layer nodes, batch size, learning rate, dropout rate, and number of virtual relay reflection devices.
[0079] Based on the above embodiments, by receiving complete channel measurement data, including direct link signals and signals reflected via relay reflection devices, and extracting multi-dimensional features that reflect the physical environment of signal propagation, an end-to-end mapping relationship from channel features to spatial coordinates is established using a positioning model, thereby achieving accurate positioning of the user to be located. This method can reduce the deployment and maintenance costs of multiple relay reflection devices and also improve the positioning accuracy and reliability of the user to be located.
[0080] In some embodiments, such as Figure 6 As shown, the training process for the localization model includes the following steps S401~S402: S401. Based on the multidimensional features of the channel and the spatial location information of the user, construct training samples.
[0081] In some embodiments, multiple user location samples are randomly generated throughout the entire area, channel measurement data is obtained using the method described above, and then the multidimensional features and their true coordinates corresponding to each sample are obtained, thereby constructing training samples for training the positioning model, which are divided into training datasets and test datasets.
[0082] S402. Train the localization model to be trained based on the training samples to obtain the trained localization model.
[0083] In some embodiments, during the training of the localization model, a small batch of samples is selected from the training dataset each time, and their channel feature tensors are input into the model for forward propagation to obtain predicted coordinates; the Huber loss function value between the predicted coordinates and the true coordinate labels is calculated; and the weights of the localization model are updated using the backpropagation algorithm. After each round of training, the localization accuracy of the current model is evaluated on the validation subset (e.g., calculating the average localization error or 95th percentile error). When the validation set error no longer decreases over multiple consecutive training rounds, an early stopping mechanism is triggered, and the weights of the model with the best validation performance are saved.
[0084] For example, the training samples are standardized or normalized to ensure that features of different dimensions can be trained within the same numerical range, avoiding the problems of gradient vanishing or gradient exploding during the localization model training process due to excessively large or small values.
[0085] For example, after extracting high-dimensional feature maps, the convolutional layers in the localization model introduce the non-linear activation function LeakyReLU. This maintains a small negative slope when x < 0, avoiding the problem of gradients being completely zero. This helps improve the training stability of the localization model, enabling it to fit complex non-linear mapping relationships. Furthermore, batch normalization or layer normalization operations can be inserted between convolutional layers to accelerate training convergence and improve the stability of the localization model.
[0086] For example, the weight parameters of the fully connected layers are updated through backpropagation during training, enabling the localization model to learn the optimal mapping relationship based on multi-dimensional features. Meanwhile, to improve the generalization ability of the localization model, a dropout operation can be introduced between fully connected layers during training to reduce the risk of overfitting the localization model to the training data by disabling some neurons.
[0087] For example, the localization model uses the Huber Loss function as the optimization objective. The Huber loss approximates a squared loss when the error is small, and an absolute value loss when the error is large. This ensures the localization model's convergence speed while also improving its robustness to outliers or anomalies. During training, the Adam optimizer iteratively updates the localization model parameters, dynamically adjusting the learning rate (ReduceLROnPlateau) to achieve fast and stable convergence. ReduceLROnPlateau is an adaptive learning rate adjustment strategy used to dynamically reduce the learning rate during training to prevent the model from stagnating at local optima.
[0088] For example, the ReduceLROnPlateau adjustment strategy satisfies the following expression:
[0089] in, This represents the learning rate in the nth iteration; This represents the learning rate in the nth iteration; This represents the learning rate decay factor, with a commonly used value of; This represents the patience value, which is the number of rounds after which performance degradation is triggered.
[0090] For example, when validating the performance metrics of the dataset (such as loss or p95 error) in continuous If no improvement is achieved within a certain number of epochs, the scheduler will multiply the learning rate by a factor. (generally, This reduces the learning rate, helping the model converge better to the global or a better local optimum.
[0091] For example, during the training of a localization model, multiple iterations are performed on the training dataset, each iteration being called an epoch. After each epoch, the localization accuracy metric of the current localization model, such as 95% localization error (p95), is calculated on the validation dataset, and a convergence curve is plotted to observe the training progress of the localization model. When the performance on the validation set no longer significantly improves over multiple epochs, training is terminated early to avoid overfitting.
[0092] For example, after training is complete, the localization model parameters are automatically saved, with the file name including hyperparameter settings and a timestamp to prevent results from overlapping between different experiments. Subsequently, during the testing or deployment phase, users can directly load the pre-trained localization model for localization inference without retraining, improving usability.
[0093] In some embodiments, before training the localization model to be trained based on training samples, the method further includes: obtaining the initial value and adjustable range of each hyperparameter of the localization model; and using a greedy search strategy to optimize each hyperparameter sequentially based on the initial value and adjustable range of each hyperparameter to obtain the optimized hyperparameter.
[0094] For example, before training begins, reasonable initial values and adjustable ranges are set for each hyperparameter. The initial values of the hyperparameters can be determined based on prior experience or small-scale trials; for instance, the number of convolutional layers can be set to 2-4 layers, and the learning rate can be set to... arrive Within the specified range, the dropout rate can be adjusted between 0.3 and 0.6. If the hyperparameters are not designed properly, the localization model may treat noise as a location feature.
[0095] For example, a manual greedy search strategy is employed, fixing the remaining hyperparameters each time and adjusting only a single hyperparameter to observe its impact on positioning accuracy metrics (e.g., 95% positioning error p95). Specifically, in each optimization round, a complete positioning model training and validation is performed using the current hyperparameter combination as a baseline; the target hyperparameter is adjusted, gradually trying different values, recording the p95 error and convergence speed for each value; the optimal value of this hyperparameter is selected, and the next set of hyperparameters is optimized based on this. Through this iterative approach, the optimized hyperparameter combination is selected.
[0096] For example, a greedy search strategy can be used to adjust and optimize the learning rate. When the performance of the validation set does not improve within p consecutive epochs, the current learning rate is reduced by a specified factor γ. This adjustment can automatically reduce the learning rate when the localization model is close to convergence, helping the localization model to escape local optima and improve the final localization accuracy.
[0097] For example, when all hyperparameters have been optimized and the performance metrics on the validation dataset reach the preset accuracy requirements or the performance metrics no longer improve significantly after multiple rounds of optimization, the greedy search process is stopped.
[0098] For example, the localization model is trained based on the optimized hyperparameter combination, and this optimized hyperparameter combination is used as the unified configuration for subsequent comparative experiments to ensure the comparability and repeatability of experimental results. After determining the final hyperparameter combination, multiple random seed experiments are conducted to evaluate the performance fluctuation of the hyperparameter combination under different initial conditions, ensuring the robustness and stability of the model under random sampling and different user distribution conditions.
[0099] Based on the above embodiments, by extracting multi-dimensional features of the channel and user spatial location information from the direct link signal and the transmitted link signal, training samples can be constructed to obtain training samples covering the entire positioning area and including various channel conditions and user locations. Subsequently, the positioning model is trained using the Huber loss function and a manual greedy search strategy. The trained positioning model possesses end-to-end direct inference capabilities, improving the accuracy and reliability of the positioning model in real-world complex environments, and also meeting the needs of real-time positioning applications.
[0100] The positioning method of this application embodiment is described below with reference to a specific example. The specific implementation process of this method is as follows: Figure 7 As shown.
[0101] S501, Obtain the user's direct link signal and transmit link signal.
[0102] S502. Extract multi-dimensional features of the channel from the direct link signal and the transmitted link signal.
[0103] S503. Construct training samples based on multidimensional features and the corresponding user spatial location information.
[0104] S504. Train the localization model based on the training samples, Huber loss function, and human greedy search strategy.
[0105] S505. Stack the multidimensional features of the user to be located into a channel feature tensor in a fixed channel order.
[0106] S506. Input the channel feature tensor into the trained positioning model to obtain the spatial location coordinates of the user to be located.
[0107] As can be seen, the above mainly describes the solutions provided by the embodiments of this application from a methodological perspective. To achieve the above functions, the embodiments of this application provide corresponding hardware structures and / or software modules for executing each function. Those skilled in the art should readily recognize that, in conjunction with the modules and algorithm steps of the various examples described in the embodiments disclosed herein, the embodiments of this application can be implemented in hardware or a combination of hardware and computer software. Whether a function is executed by hardware or by computer software driving hardware depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this invention.
[0108] This application embodiment can divide the positioning device into functional modules according to the above method example. For example, each function can be divided into its own functional module, or two or more functions can be integrated into one processing module. The integrated module can be implemented in hardware or as a software functional module. Optionally, the module division in this application embodiment is illustrative and only represents one logical functional division; other division methods may be used in actual implementation.
[0109] In some embodiments, this application provides a positioning device. The positioning device may include one or more functional units for implementing the positioning method described in the above embodiments.
[0110] For example, Figure 8 This is a schematic diagram illustrating the composition of a positioning device provided in this application. Figure 8 As shown, the positioning device 600 includes an acquisition unit 601 and a processing unit 602.
[0111] Acquisition unit 601 is used to acquire channel measurement data of the user to be located; wherein, the channel measurement data includes direct link signal and reflected link signal; wherein, the reflected link signal includes the signal reflected by the relay reflection device; processing unit 602 is used to obtain multi-dimensional features of the channel based on the channel measurement data; wherein, the multi-dimensional features of the channel are used to reflect the physical environment information of signal propagation; processing unit 602 is used to obtain the spatial location coordinates of the user to be located based on the multi-dimensional features of the channel and the positioning model; wherein, the positioning model is used to establish an end-to-end mapping relationship from multi-dimensional features to spatial location coordinates.
[0112] In some embodiments, the acquisition unit 601 has the function of adjusting the phase of a single physical relay reflector and equivalently generating multiple virtual relay reflectors within a preset range around the single physical relay reflector; wherein the multiple virtual relay reflectors operate in different time slots.
[0113] In other embodiments, the processing unit 602 has the function of stacking the multidimensional features of the channel into a channel feature tensor in a fixed channel order; inputting the channel feature tensor into a positioning model to obtain the spatial location coordinates of the user to be located.
[0114] When implementing the functions of the integrated modules described above in hardware, this embodiment of the invention provides a possible schematic diagram of the electronic device involved in the above embodiments. For example... Figure 9 As shown, the electronic device 700 includes: a processor 702, a communication interface 703, and a bus 704. Optionally, the electronic device 700 may also include a memory 701.
[0115] Processor 702 may implement or execute various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure of this application. Processor 702 may be a central processing unit, a general-purpose processor, a digital signal processor, an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It may implement or execute various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure of this application. Processor 702 may also be a combination that implements computing functions, such as including one or more microprocessor combinations, a combination of a DSP and a microprocessor, etc.
[0116] The communication interface 703 is used to connect to other devices via a communication network. This communication network can be Ethernet, wireless access network, wireless local area network (WLAN), etc.
[0117] The memory 701 may be a read-only memory (ROM) or other type of static storage device capable of storing static information and instructions, random access memory (RAM) or other type of dynamic storage device capable of storing information and instructions, or electrically erasable programmable read-only memory (EEPROM), disk storage medium or other magnetic storage device, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but is not limited thereto.
[0118] In one possible implementation, the memory 701 can exist independently of the processor 702. The memory 701 can be connected to the processor 702 via a bus 704 and is used to store instructions or program code. When the processor 702 calls and executes the instructions or program code stored in the memory 701, it can implement the positioning method provided in this embodiment of the invention.
[0119] In another possible implementation, the memory 701 can also be integrated with the processor 702.
[0120] The 704 bus can be an extended industry standard architecture (EISA) bus, etc. The 704 bus can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 9 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.
[0121] Through the above description of the implementation methods, those skilled in the art can clearly understand that, for the sake of convenience and brevity, only the division of the above functional modules is used as an example. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the service calling device can be divided into different functional modules to complete all or part of the functions described above.
[0122] This application also provides a computer-readable storage medium. All or part of the processes in the above method embodiments can be executed by computer instructions instructing related hardware. The program can be stored in the aforementioned computer-readable storage medium, and when executed, it can include the processes of the above method embodiments. The computer-readable storage medium can be any of the foregoing embodiments or memory. The aforementioned computer-readable storage medium can also be an external storage device of the aforementioned service invocation device, such as a plug-in hard drive, smart media card (SMC), secure digital (SD) card, flash card, etc., equipped on the aforementioned service invocation device. Further, the aforementioned computer-readable storage medium can include both internal storage units of the aforementioned service invocation device and external storage devices. The aforementioned computer-readable storage medium is used to store the aforementioned computer program and other programs and data required by the aforementioned service invocation device. The aforementioned computer-readable storage medium can also be used to temporarily store data that has been output or will be output.
[0123] This application also provides a computer program product comprising a computer program that, when run on a computer, causes the computer to perform any of the positioning methods provided in the above embodiments.
[0124] The above are merely specific embodiments of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.
Claims
1. A positioning method, characterized in that, include: Acquire channel measurement data of the user to be located; wherein the channel measurement data includes direct link signals and reflected link signals; wherein the reflected link signals include signals reflected via relay reflection equipment; Based on the channel measurement data, multidimensional features of the channel are obtained; wherein, the multidimensional features of the channel are used to reflect the physical environment information of signal propagation; Based on the multidimensional features of the channel and the positioning model, the spatial coordinates of the user to be located are obtained; wherein, the positioning model is used to establish an end-to-end mapping relationship between the multidimensional features and the spatial coordinates.
2. The method according to claim 1, characterized in that, The relay reflection device includes any one of the following: Single physical relay reflection device; A single physical relay reflector and multiple virtual relay reflectors; Multiple physical relay reflection devices.
3. The method according to claim 2, characterized in that, The plurality of virtual relay reflection devices are generated in the following manner: The phase of the single physical relay reflector is adjusted to generate multiple virtual relay reflectors equivalently within a preset range around the single physical relay reflector; wherein the multiple virtual relay reflectors operate in different time slots.
4. The method according to claim 1, characterized in that, The spatial location coordinates of the user to be located are obtained based on the multidimensional features and positioning model of the channel, including: The multidimensional features of the channel are stacked into a channel feature tensor in a fixed channel order; The channel feature tensor is input into the positioning model to obtain the spatial location coordinates of the user to be located.
5. The method according to claim 4, characterized in that, The localization model includes: an input layer, a convolutional layer, a fully connected layer, and an output layer; The input layer is used to receive the channel feature tensor; The convolutional layer is used to extract a high-dimensional feature map of the channel feature tensor; the high-dimensional feature map is used to reflect the spatial correlation and path structure information of the channel feature tensor. The fully connected layer is used to process the high-dimensional feature map and reduce its dimensionality to a low-dimensional feature vector. The output layer is used to map the low-dimensional feature vector into spatial location coordinates.
6. The method according to claim 1, characterized in that, The hyperparameters of the localization model are preferred hyperparameters; the preferred hyperparameters are parameters optimized using a greedy search strategy.
7. The method according to claim 6, characterized in that, The hyperparameters include at least one of the following: Number of convolutional layers, kernel size, number of fully connected layer nodes, batch size, learning rate, dropout rate, and number of virtual relay reflection devices.
8. The method according to claim 1, characterized in that, The multidimensional characteristics of the channel include at least one of the following: Received signal strength, path direction angle, path type, phase information, and propagation distance.
9. An electronic device, characterized in that, It includes a processor and a memory, the processor being coupled to the memory; the memory is used to store computer instructions, which are loaded and executed by the processor to enable the computer device to perform the positioning method as described in any one of claims 1 to 8.
10. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes computer-executable instructions that, when executed on a computer, cause the computer to perform the positioning method as described in any one of claims 1 to 8.