Wireless positioning method and system, chip, storage medium and terminal equipment

Abstract

The invention discloses a wireless positioning method and system, a chip, a storage medium and an electronic device, and relates to the technical field of Internet of Things and artificial intelligence, and the method comprises the steps: receiving positioning reference data sent by a plurality of base stations, inputting the positioning reference data into a parameter estimation model, estimating the distance and confidence between each base station and a terminal device, and determining the distance and confidence between each base station and the terminal device; the confidence coefficient is used for representing the interference degree of a transmission path between the base station and the terminal equipment, the parameter estimation model is a pre-trained neural network model, a weight is distributed for the distance according to the confidence coefficient, and the current position of the terminal equipment is calculated according to the weight and the distance. The distance between each base station and the terminal equipment and the confidence coefficient are estimated through a deep learning model, and high-precision and low-delay positioning is realized based on the estimated distance and confidence coefficient in cooperation with weighted multilateral positioning.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of Internet of Things and artificial intelligence, and in particular to a wireless positioning method and system, a chip, a storage medium and an electronic device. BACKGROUND

[0002] Wireless positioning technology is a key technology for obtaining user equipment position information, and has wide application in the fields of Internet of Things and artificial intelligence. The basic principle is to analyze the wireless signal data between user equipment and base stations, and calculate the position of the user equipment in combination with the coordinate information of the base stations, to provide data support for various location-based applications.

[0003] Current wireless positioning technology mainly falls into two categories: one is the traditional geometric positioning method, which takes signal features and geometric relationships as the core, measures the signal strength, time difference and other data between user equipment and base stations, estimates the distance in combination with the free space path loss model, and then determines the coordinates of the user equipment by using the geometric intersection of at least three base stations. The other is an improved scheme based on machine learning or deep learning, which trains random forest, support vector regression, convolutional neural network and other models, learns the communication data between user equipment and base stations, and the mapping relationship between user equipment position, and directly outputs the position information in the inference stage.

[0004] However, the traditional geometric positioning method relies on a fixed propagation model, and is easily affected by multipath effects, obstacle shielding and other factors in complex environments, resulting in large distance estimation deviation, and the reliability of the transmission path between the base station and the user equipment cannot be dynamically evaluated. The static weight allocation strategy is difficult to adapt to the dynamic changes of the path, and the positioning accuracy and robustness are low. The method based on machine learning or deep learning needs to integrate multi-base station multi-dimensional data to form high-dimensional input, resulting in large model calculation amount, and a large amount of labeled data is needed to train the model, and the calculation complexity is high, which is not suitable for resource-limited scenarios. SUMMARY

[0005] Therefore, the present application provides a wireless positioning method, system, chip, storage medium and electronic device, which solves the deficiencies of wireless positioning technology in terms of accuracy, efficiency, anti-interference ability, etc., and realizes high-precision, low-delay and end-side scenario suitable positioning.

[0006] In a first aspect, the present application provides a wireless positioning method applied to a terminal device, comprising:

[0007] receiving positioning reference data sent by a plurality of base stations;

[0008] inputting the positioning reference data into a parameter estimation model to estimate a distance between the base station and the terminal device and a confidence level; the confidence level is used to represent an interference degree of a transmission path between the base station and the terminal device, and the parameter estimation model is a pre-trained neural network model;

[0009] assigning a weight to the distance according to the confidence level;

[0010] calculating a current position of the terminal device according to the weight and the distance.

[0011] In some embodiments of the present disclosure, training the parameter estimation model comprises:

[0012] obtaining historical actual distances between the base station and the terminal device and historical positioning reference data;

[0013] constructing a Gaussian distribution; a mean of the Gaussian distribution represents the distance output by the parameter estimation model, and a standard deviation of the Gaussian distribution is negatively correlated with the confidence level output by the parameter estimation model;

[0014] using a negative log maximum likelihood of the Gaussian distribution as a loss function to train the parameter estimation model based on the historical actual distances and the historical positioning reference data.

[0015] In some embodiments of the present disclosure, the positioning reference data transmitted by multiple base stations is received at a preset period.

[0016] In some embodiments of the present disclosure, the distances and the confidence levels corresponding to multiple continuous periods are subjected to an exponential moving average processing.

[0017] In some embodiments of the present disclosure, after the step of inputting the positioning reference data into the parameter estimation model to estimate the distance between the base station and the terminal device and the confidence level, the method further comprises:

[0018] obtaining a confidence level threshold;

[0019] in response to the confidence level being less than or equal to the confidence level threshold, discarding the confidence level and the distance;

[0020] in response to the confidence level being greater than the confidence level threshold, retaining the confidence level and the distance.

[0021] In some embodiments of the present disclosure, after the step of inputting the positioning reference data into the parameter estimation model to estimate the distance between the base station and the terminal device and the confidence level, the method further comprises:

[0022] obtaining a confidence level threshold;

[0023] detecting a number of confidence levels greater than the confidence level threshold;

[0024] In response to the quantity of the confidence degrees being less than or equal to the target quantity, the confidence degrees are sorted in descending order, and the first target quantity of the confidence degrees and the distances are retained.

[0025] In some embodiments of the present disclosure, the assigning the distances with weights according to the confidence degrees comprises:

[0026] The weight is a ratio of the confidence degree to a sum of all the confidence degrees.

[0027] The distance is configured with the weight.

[0028] In some embodiments of the present disclosure, the calculating the current position of the terminal device according to the weight and the distance comprises:

[0029] The base station position coordinates of the base station are obtained.

[0030] A circle equation is constructed according to the base station position coordinates and the distance.

[0031] The circle equation is converted into a linear equation.

[0032] A weight matrix is constructed according to the weight.

[0033] The current position is calculated based on the linear equation and the weight matrix by using a least square method.

[0034] In a second aspect, the present application provides a wireless positioning system applied to a terminal device, comprising:

[0035] A parameter estimation module is configured to receive positioning reference data sent by a plurality of base stations, input the positioning reference data into a parameter estimation model, and estimate distances and confidence degrees between the base stations and the terminal device. The confidence degree is used to represent an interference degree of a transmission path between the base station and the terminal device, and the parameter estimation model is a pre-trained neural network model.

[0036] A positioning module is configured to assign weights to the distances according to the confidence degrees, and calculate a current position of the terminal device according to the weights and the distances.

[0037] In a third aspect, the present application provides a computer readable storage medium having a computer program stored thereon, wherein the computer program is executed by a processor to implement the method of the first aspect.

[0038] In a fourth aspect, the present application provides a terminal device comprising a storage medium, a processor, and a computer program stored in the storage medium and executable on the processor, wherein the processor implements the method of the first aspect when executing the computer program.

[0039] In a fifth aspect, the present application provides a computer program product, which stores a computer program and is executed by a processor to implement the method of the first aspect.

[0040] By means of the technical scheme, the wireless positioning method, system, chip, storage medium and electronic device provided by the present application are provided, wherein the method comprises: receiving positioning reference data sent by a plurality of base stations, inputting the positioning reference data into a parameter estimation model, estimating the distance between each base station and a terminal device and the confidence, the confidence being used to represent the interference degree of the transmission path between the base station and the terminal device, the parameter estimation model being a pre-trained neural network model, assigning a weight to the distance according to the confidence, and calculating the current position of the terminal device according to the weight and the distance. The distance between each base station and the terminal device and the confidence are estimated by means of the deep learning model, and based on the estimated distance and confidence, the weighted multilateration is used to realize high-precision, low-delay and end-side scene applicable positioning.

[0041] The above description is only a summary of the technical scheme of the present application. In order to more clearly understand the technical means of the present application, the specific embodiments of the present application can be implemented according to the content of the description, and in order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the following specific embodiments of the present application are described. BRIEF DESCRIPTION OF DRAWINGS

[0042] The drawings incorporated into the specification and forming part of the specification show embodiments consistent with the present application and, together with the specification, serve to explain the principles of the present application.

[0043] In order to more clearly illustrate the technical scheme of the embodiments of the present application or the prior art, the drawings needed in the embodiments or the prior art description will be briefly introduced as follows. Obviously, for those skilled in the art, other drawings can also be obtained without creative labor.

[0044] Figure 1 A multi-base station positioning scene schematic diagram provided by the embodiments of the present application is shown;

[0045] Figure 2 A schematic diagram of a downlink positioning scene provided by the embodiments of the present application is shown;

[0046] Figure 3 A schematic diagram of a multilateration method in a downlink positioning scene provided by the embodiments of the present application is shown;

[0047] Figure 4 A schematic diagram of a positioning method based on machine learning or deep learning provided by the embodiments of the present application is shown;

[0048] Figure 5 A flowchart of a wireless positioning method provided by an embodiment of the application is shown.

[0049] Figure 6 A schematic diagram of wireless positioning provided by an embodiment of the application is shown.

[0050] Figure 7 A screening flowchart of confidence and distance provided by an embodiment of the application is shown.

[0051] Figure 8 Another screening flowchart of confidence and distance provided by an embodiment of the application is shown.

[0052] Figure 9 A flowchart of wireless positioning provided by an embodiment of the application is shown. DETAILED DESCRIPTION

[0053] Embodiments of the application will be described in more detail below with reference to the accompanying drawings. It should be noted that the embodiments and features of the embodiments in the application can be combined with each other without conflict.

[0054] Current wireless positioning technologies are mainly divided into traditional geometric positioning methods and positioning methods based on machine learning or deep learning. The traditional geometric positioning method takes signal features and geometric relationships as the core to realize positioning, and typical schemes include weighted multi-lateral positioning. Weighted multi-lateral positioning can estimate the distance by measuring the signal strength between the user equipment and the base station, combining the free space path loss model, or estimating the distance between the user equipment and the base station based on other data, and using the geometric intersection of at least three base stations to determine the coordinates of the user equipment. Such methods rely on fixed propagation models and static weight allocation strategies, such as equal weight averaging, and have low algorithm complexity and strong compatibility, and are suitable for simple static environments.

[0055] The positioning method based on machine learning or deep learning uses machine learning or deep learning to improve positioning accuracy, and typical technologies include using models such as random forest, support vector regression, or deep models such as convolutional neural networks. Such methods learn the mapping relationship between data (such as communication data between user equipment and each base station) and user equipment location by training the model, and further use the trained model to infer the user equipment location in the inference stage. Such methods have high positioning accuracy and can adapt to complex environments, but require a large number of parameters in the model to support its inference accuracy, resulting in high latency and low applicability in low-power end-side application scenarios.

[0056] The aforementioned wireless positioning technologies all have significant limitations. Traditional geometric positioning methods calculate user location by combining signal strength or time difference with base station coordinates, but they rely on fixed propagation models such as free-space path loss formulas, leading to large errors in complex environments due to multipath effects and obstacle obstruction. For example... Figure 1 The diagram illustrates a multi-base station positioning scenario provided in an embodiment of this application. One base station's propagation path to the user equipment is obstructed by obstacles such as buildings, affecting signal transmission. This significantly impacts the estimation of the Time of Arrival (ToA) and Angle of Arrival (AoA) for this path, leading to a large deviation in the distance estimation between the user equipment and the base station, resulting in low robustness. Traditional geometric positioning methods struggle to assess the reliability (confidence) of the transmission path between a specific base station and the user equipment; therefore, their weights are typically fixed. However, in reality, the transmission path between the base station and the user equipment is dynamically changing, making it difficult to achieve high positioning accuracy.

[0057] Improved solutions based on machine learning or deep learning directly predict the location information of user devices through neural networks or regression models. These methods are highly resistant to interference and easily achieve good positioning results. However, they still have core drawbacks. These methods require integrating data from each base station and user device to form the model's input data, resulting in excessively large model input dimensions and computational load. In this case, if the model is computed on the edge, the required computing resources are substantial. If the model is computed in the cloud, data from each base station needs to be transmitted to the cloud for computation, generating significant data transmission overhead. Furthermore, these methods require a large amount of labeled data to train the model to ensure good positioning results, necessitating significant human and material resources.

[0058] In some embodiments, the wireless positioning method is applied in downlink positioning scenarios, such as... Figure 2 The diagram illustrates a downlink positioning scenario. Downlink positioning refers to the network side (such as a base station) sending signals to the user equipment (UE), which then calculates its location information based on the received signals. In a downlink positioning scheme, the UE needs to receive data sent by the base station, such as downlink positioning reference signals and channel state information. The UE then performs data analysis and location estimation based on the received signals. In a downlink positioning scheme, data processing and positioning are completed on the UE side, requiring minimal data transmission overhead and resulting in low latency.

[0059] like Figure 3As shown in FIG. 1, it is a schematic diagram of a multi-lateral positioning method in a downlink positioning scenario. The traditional multi-lateral positioning method needs the user equipment to calculate the distance between each base station and the user equipment based on the transmission data between each base station and the user equipment, and further position the location through geometric principles. Although this method has small calculation amount and simple implementation, it is easily affected by non-line-of-sight (NLOS), multipath effect and environmental noise, resulting in limited positioning accuracy.

[0060] As shown in FIG. 2, it is a schematic diagram of a positioning method based on machine learning or deep learning in a downlink positioning scenario. The positioning method based on machine learning or deep learning needs the user equipment to perform spatiotemporal feature splicing on multi-dimensional data such as channel state information (CSI) and received signal strength (RSS) between each base station and the user equipment, form a high-dimensional tensor input deep learning module for processing, and then directly output the location coordinates of the user through a regression layer. This method has higher positioning accuracy and stronger robustness to environmental interference, but needs a large amount of labeled data for model training and has high calculation complexity, which poses challenges to the deployment on terminal devices with limited resources. Figure 4

[0061] Therefore, in order to solve the above problems, the embodiment provides a multi-base station cooperative high-precision wireless positioning method. On the user equipment side, the distance between each base station and the user equipment and the confidence of the transmission path between each base station and the user equipment are estimated by a deep learning method. Based on the estimated distance and confidence, a weighted multi-lateral positioning is performed to realize high-precision and fast positioning.

[0062] As shown in FIG. 3, it is a flowchart of a wireless positioning method provided by the embodiment of the application. The method is applied to a terminal device and includes the following steps. Figure 5

[0063] S501: receiving positioning reference data sent by a plurality of base stations.

[0064] The positioning reference data is data used for assisting positioning calculation, including but not limited to downlink positioning reference signals, received signal strength (RSS) and channel state information (CSI).

[0065] S502: inputting the positioning reference data into a parameter estimation model to estimate the distance and confidence between the base stations and the terminal device.

[0066] ​​The confidence level is used to characterize the degree of interference in the transmission path between the base station and the terminal device, and the parameter estimation model is a pre-trained neural network model.

[0067] S503: Assign weights to distances based on confidence level.

[0068] S504: Calculate the current location of the terminal device based on weight and distance.

[0069] It is understood that terminal devices refer to user devices that interact with users. This embodiment does not limit the specific implementation of terminal devices. For example, they can be mobile terminals such as mobile phones, tablets, and smartwatches, or IoT terminals such as smart sensors, smart home appliances, and monitoring equipment.

[0070] like Figure 6 As shown, the above wireless positioning method involves a parameter estimation module and a positioning module. The parameter estimation module uses a lightweight parameter estimation model and estimates the distance d between each base station and the terminal device based on the data between each base station and the terminal device. i Let i = 1, ..., n, and simultaneously estimate the confidence level u of the transmission path between each base station and the terminal device. i The confidence level, i = 1, ..., n, reflects the degree of interference in the transmission path. The higher the confidence level, the lower the degree of interference. The estimated distance and confidence level are transmitted to the positioning module. The positioning module uses a weighted multilateral positioning method to calculate the weight of each side based on the confidence information output by the parameter estimation module. The weighted multilateral positioning is combined with the position estimation to obtain the position information of the terminal device.

[0071] In this embodiment, the advantages of deep learning and multi-way positioning are combined to achieve a balance between positioning accuracy and computational efficiency. Specifically, the parameter estimation module learns complex wireless channel characteristics through end-to-end training, effectively overcoming the ranging bias of traditional methods in non-line-of-sight environments. The confidence information output by the parameter estimation model is more stable and effective, enabling better evaluation of the confidence of each propagation path. The positioning module uses the confidence output by the parameter estimation module to calculate weights and adaptively adjust the contribution of each base station's measurement results, significantly improving the robustness of positioning.

[0072] In some embodiments, the step of training the parameter estimation model includes: acquiring historical actual distance and historical positioning reference data between the base station and the terminal device; constructing a Gaussian distribution, where the mean of the Gaussian distribution represents the distance output by the parameter estimation model, and the standard deviation of the Gaussian distribution is negatively correlated with the confidence level of the parameter estimation model output; and using the negative log-maximum likelihood of the Gaussian distribution as the loss function to train the parameter estimation model based on the historical actual distance and historical positioning reference data.

[0073] In this embodiment, considering the limited computing power of the terminal device, a lightweight distance estimation scheme based on deep learning is designed. Specifically, the designed parameter estimation model does not need to splice the data of each base station to form high-dimensional data. The data pair used in the training of the parameter estimation model is the positioning reference data between the terminal device and the i-th base station and the actual distance between the terminal device and the i-th base station.

[0074] The parameter estimation model is used to estimate the distance d between the terminal device and the i-th base station i and the confidence u i . In this embodiment, a Gaussian modeling method is used to model the distance information between the terminal device and the base station as a Gaussian distribution, and the probability density function is:

[0075]

[0076] The negative log-likelihood of the probability density function is used as the loss function L:

[0077]

[0078] The parameter estimation model trained according to the loss function can estimate the distance d between each base station and the terminal device i and the confidence u of this estimation result i .

[0079] In addition, the same parameter estimation model is deployed for each base station, that is, the same weight data is used. All base stations can share the same parameter estimation model, and there is no need to train a model for each base station, reducing the training overhead. For a group of distance data of terminal devices, n groups of training samples (corresponding to n base stations) can be generated, which can greatly increase the amount of training data and greatly reduce the manpower and material resources required for data collection.

[0080] In this embodiment, distance prediction and confidence estimation are realized through a specific loss function. Compared with traditional deep learning-based positioning methods, the input data dimension is smaller, the calculation amount is lower, and the speed is faster. Distance prediction and confidence estimation between base stations and terminal devices can be realized, and under the same conditions, more training data can be used, and the model effect is better.

[0081] In some embodiments, in order to realize real-time positioning, the positioning reference data sent by the plurality of base stations can be received according to a preset period, the distance and the confidence between the base station and the terminal device are estimated through the parameter estimation model, and the real-time position of the terminal device is calculated based on the distance and the confidence.

[0082] The preset period can be set based on actual needs. For example, for high dynamic scenarios, the preset period can be set to 50ms to improve response speed; for static scenarios, the preset period can be set to 200ms to reduce power consumption.

[0083] In some embodiments, the distance and confidence level of a single period are susceptible to transient interference (such as burst multipath or signal blockage). Therefore, the distance and confidence level corresponding to multiple consecutive periods can be processed by exponential moving average to reduce the impact of transient noise. The distance and confidence level are smoothed by exponential moving average, making the distance and confidence level more stable.

[0084] For example, the distance and confidence level can be processed by an exponential moving average as follows:

[0085]

[0086] Where α is the smoothing coefficient, and its value range is 0-1, d i This represents the distance corresponding to the current cycle. The distance after performing an exponential moving average, u i The confidence level corresponding to the current period. The confidence level after performing an exponential moving average.

[0087] In some embodiments, to ensure the accuracy of positioning calculations, after estimating the distances and confidence levels between each base station and the terminal device using a parameter estimation model, the distances and confidence levels can be filtered to select those with higher confidence levels and their corresponding distances for subsequent positioning calculations. By filtering the estimated data from low-confidence base stations, positioning accuracy can be improved and the computational load reduced.

[0088] like Figure 7 The diagram illustrates a filtering process for confidence level and distance provided in this embodiment. After obtaining the distance and confidence level between each base station and the terminal device, a preset confidence level threshold is obtained (S701). The confidence level is compared with the confidence level threshold (S702). If the confidence level is less than or equal to the confidence level threshold, it indicates that the distance estimation of that base station is greatly affected by interference, and the confidence level and the corresponding distance are removed (S703). If the confidence level is greater than the confidence level threshold, the confidence level and the corresponding distance are retained (S704).

[0089] like Figure 8As shown, another filtering process diagram of confidence and distance provided by the embodiment is shown, after the distance and confidence between each base station and the terminal device are obtained, the confidence threshold is obtained (S801), the number of confidence greater than the confidence threshold is detected (S802), the number of confidence is compared with the target number (S803), in response to the number of confidence being less than or equal to the target number, the confidence is sorted in descending order, and the first target number of confidence and the corresponding distance are retained (S804).

[0090] In some embodiments, the distance is assigned a weight according to the confidence, the weight can be calculated according to the confidence first, and then the corresponding distance is configured with the weight. The weight is the ratio of the confidence to the sum of the confidence, and the confidence is positively correlated with the weight. The weight in the weighted multilateration is obtained by the confidence predicted by the deep learning method, and is a dynamic weight, which can obtain the weight of each base station of the terminal device in different situations and improve the positioning accuracy.

[0091] For example, the confidence after the filtering operation is The weight can be calculated based on the following formula:

[0092]

[0093] Wherein, w j is the weight calculated according to the jth confidence , and n l is the number of confidence after the filtering operation.

[0094] In some embodiments, the position information of the terminal device is calculated by geometric modeling and weighted least squares method in combination with the weighted distance of the multiple base stations. That is, the step of calculating the current position of the terminal device according to the weight and the distance includes: obtaining the base station position coordinates of the base station, constructing a circle equation according to the base station position coordinates and the distance between the base station and the terminal device, converting the circle equation into a linear equation, and constructing a weight matrix according to the weight. The current position is calculated based on the linear equation and the weight matrix using the least squares method.

[0095] The positioning method provided by the present application is described below. Figure 9 The positioning method provided by the present application is described below.

[0096] The initialization time T is 1, the distance is initialized, and the confidence .

[0097] Step 1, obtaining the positioning reference data between each base station and the terminal device at time T, estimating the distance d i and the confidence u i between each base station and the terminal device at time T according to the positioning reference data through the reference estimation model.

[0098] Step 2, smoothing the obtained distance d by exponential moving average operation i and the confidence u i , to obtain the smoothed distance and the confidence

[0099]

[0100] wherein a is a smoothing coefficient, and the value range is 0-1.

[0101] Step 3, detecting the time T and the threshold K, if the time T is greater than the threshold K, performing the elimination of the abnormal distance and the confidence According to the smoothed confidence , detecting the number of the confidence greater than the preset confidence threshold ε, if the number is less than or equal to 3, retaining the top 3 confidences and the corresponding distances

[0102] If the number is greater than 3, retaining the confidence greater than the preset confidence threshold ε and the corresponding distance , and eliminating the confidence less than or equal to the preset confidence threshold ε and the corresponding distance , to reduce the interference of the abnormal estimation data on the positioning result.

[0103] Step 4, assuming that the number of the remaining distance and the confidence is n l , according to the remaining confidence , calculating the weight w of each propagation path j :

[0104]

[0105] wherein w j is the weight calculated according to the jth confidence .

[0106] Step 5, calculating the position information of the terminal device based on the weighted multilateration method. The position coordinate information of each base station is known as P i =(x i ,y i ), i≤n l , and the distance between the terminal device and each base station can be constructed as the following circle equation:

[0107]

[0108] Based on the elimination linearization, the circle equation is converted into a linear equation, the weight of the lowest corresponding circle equation, that is, the mth circle equation, is determined, and the remaining n l -1 circle equation is subtracted from the mth circle equation, respectively, to obtain a linear equation, and converted into a matrix form as shown below:

[0109]

[0110] Then, the weight matrix is constructed according to the weight as follows:

[0111]

[0112] where the diagonal element w i represents the weight of the ith equation.

[0113] Based on the linear equation and the weight matrix, the position of the terminal device is calculated by using the least square method:

[0114] P = (A T WA) -1 A T Wb;

[0115] where P is the current position of the terminal device.

[0116] The above steps 1 to 5 are repeated to perform real-time position calculation of the terminal device.

[0117] In this embodiment, first, distance estimation and confidence estimation are performed by a deep learning method, and then an optimized multilateral weighted positioning method is used for real-time positioning. The optimized multilateral weighted positioning uses the distance and confidence estimated by the deep learning method for real-time position calculation.

[0118] Based on the deep learning model trained using a specific loss function, the distance and confidence between each base station and the terminal device are estimated for subsequent weighted multilateral positioning. Compared with traditional methods based on TOA, DOA, etc., the anti-interference ability is stronger, and the confidence of the transmission path between each base station and the terminal device can be estimated synchronously. The higher the confidence, the lower the interference.

[0119] An optimized weighted multilateral positioning method is used to further calculate the position based on the estimated distance and corresponding confidence between each base station and the terminal device. An optimized process is used to reduce interference and improve positioning accuracy. By smoothing the estimated distance and confidence at multiple times, high-confidence base station data is further selected to reduce interference. The confidence is calculated as a weight, and different weights are used for the transmission path between each base station and the terminal device to improve the accuracy of weighted multilateral positioning. The confidence is dynamically estimated, and the estimated result is different for different terminal device positions, which improves the positioning accuracy.

[0120] The designed scheme combines deep learning and weighted multilateration, reduces model complexity, reduces the calculation amount of the deep learning model, improves positioning accuracy, realizes positioning acceleration, and is suitable for end-side scenarios.

[0121] The scheme provided by the embodiments of the present application can significantly improve positioning accuracy and environmental adaptability, and can be applied to the fields of intelligent retail, industrial Internet of Things, smart city, etc. For example, in a large shopping mall, the method can realize accurate indoor navigation and user behavior analysis, helping merchants to optimize store layout and personalized marketing; in warehouse logistics, the positions of goods and personnel can be tracked in real time, improving operation efficiency; in smart city management, traffic monitoring, emergency response, etc. can be supported, improving public safety and service level. Its characteristics of high precision, low delay and strong anti-interference can provide technical support for the location service market.

[0122] Based on the above method, the embodiment further provides a multi-base-station cooperative high-precision wireless positioning system, comprising a parameter estimation module and a positioning module.

[0123] The parameter estimation module is configured to receive positioning reference data sent by multiple base stations, input the positioning reference data into a parameter estimation model, and estimate the distance and confidence between the base stations and the terminal device.

[0124] The confidence is used to represent the interference degree of the transmission path between the base station and the terminal device, and the parameter estimation model is a pre-trained neural network model.

[0125] The positioning module is configured to assign weights to the distances according to the confidences, and calculate the current position of the terminal device according to the weights and the distances.

[0126] It should be noted that the above system architecture provided by the embodiments does not constitute a limitation on the system, and can include more or fewer component modules, or combine certain modules, or different module arrangements. Other corresponding descriptions of the functions of the wireless positioning system provided by the embodiments can be referred to the corresponding descriptions in the wireless positioning method, which will not be described here.

[0127] Based on the above method, the embodiment further provides a computer-readable storage medium having a computer program stored thereon, and the computer program is executed by a processor to implement the above method.

[0128] Based on the above method, the embodiment further provides a computer program product having a computer program stored thereon, and the computer program product is executed by a processor to implement the above method.

[0129] Based on the understanding, the technical scheme of the present application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a U disk, a mobile hard disk, etc.), and includes a plurality of instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute the method of various implementation scenarios of the present application.

[0130] Based on the above method, the embodiment also provides a chip, including a processor and a memory, the memory stores a computer program, and the processor calls and executes the computer program from the memory to realize the above method.

[0131] It can be understood that the scheme of the present application needs to have real-time computing capability, and at the same time, it also needs to meet the low power consumption requirement. The memory-compute integrated chip can provide an ideal hardware solution for this requirement, mainly in the following three aspects: first, in the distance and confidence estimation link, the memory-compute integrated architecture provides efficient matrix multiplication and addition operation, which can significantly accelerate the forward inference process of the neural network. Second, in the core positioning solution link, the weighted multi-lateral positioning based on the least square method needs to perform frequent matrix operation, and the memory-compute integrated chip can greatly improve the operation efficiency. Finally, in terms of energy efficiency ratio, the memory-compute integrated technology can greatly reduce the overall power consumption of positioning calculation by eliminating data transfer overhead. This hardware-algorithm collaborative optimization provides reliable technical support for high-precision location services.

[0132] Based on the above method, the embodiment of the present application also provides a terminal device, which includes a storage medium and a processor; the storage medium is used to store a computer program; and the processor is used to execute the computer program to realize the above method.

[0133] Optionally, the above-mentioned entity device can also include a user interface, a network interface, a camera, a radio frequency (RF) circuit, a sensor, an audio circuit, a WI-FI module, etc. The user interface can include a display screen (Display), an input unit such as a keyboard (Keyboard), etc. The optional user interface can also include a USB interface, a card reader interface, etc. The network interface can optionally include a standard wired interface, a wireless interface (such as a WI-FI interface), etc.

[0134] Those skilled in the art can understand that the above-mentioned entity device structure provided by the embodiment does not constitute a limitation on the entity device, and can include more or fewer components, or combine certain components, or different component arrangements.

[0135] The storage medium can further include an operating system and a network communication module. The operating system is a program for managing hardware and software resources of the information processing entity, and supports the running of information processing programs and other software and / or programs. The network communication module is used to realize communication between components in the storage medium, and communication with other hardware and software in the information processing entity.

[0136] Through the above description of the embodiments, those skilled in the art can clearly understand that the present application can be implemented by means of software and necessary general hardware platforms, or by hardware. Compared with the prior art, by applying the technical solutions of the embodiments, the deficiencies of wireless positioning technology in precision, efficiency, anti-interference ability, etc. can be solved, and high-precision, low-latency, and end-side scene applicable positioning can be realized.

[0137] It should be noted that, in this document, relational terms such as "first" and "second", and the like, are used solely to distinguish one entity or action from another entity or action, without necessarily requiring or implying any actual such relationship or order between such entities or actions. Moreover, the terms "comprises", "comprising", or any other variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without more limitations, an element preceded by "comprises... " does not, without more constraints, foreclose the existence of additional identical elements in the process, method, article, or apparatus that comprises the recited element.

[0138] The above description is merely one specific implementation of the present application, which enables those skilled in the art to understand or implement the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to these embodiments described herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A wireless positioning method, characterized by, The application is applied to a terminal device, comprising: receiving positioning reference data sent by a plurality of base stations; inputting the positioning reference data into a parameter estimation model to estimate a distance between the base stations and the terminal device and a confidence level; the confidence level is used to represent an interference degree of a transmission path between the base stations and the terminal device, and the parameter estimation model is a pre-trained neural network model; assigning a weight to the distance according to the confidence level; calculating a current position of the terminal device according to the weight and the distance.

2. The method of claim 1, wherein, The training of the parameter estimation model comprises: obtaining historical actual distances between the base stations and the terminal device and historical positioning reference data; constructing a Gaussian distribution; a mean value of the Gaussian distribution represents the distance output by the parameter estimation model, and a standard deviation of the Gaussian distribution is negatively correlated with the confidence level output by the parameter estimation model; using a negative log maximum likelihood of the Gaussian distribution as a loss function to train the parameter estimation model based on the historical actual distances and the historical positioning reference data.

3. The method of claim 1, wherein, The method comprises: receiving positioning reference data sent by a plurality of base stations according to a preset period.

4. The method of claim 3, wherein, The method comprises: performing exponential moving average processing on distances and confidence levels corresponding to a plurality of continuous periods.

5. The method of claim 1, wherein, After the step of inputting the positioning reference data into the parameter estimation model to estimate the distance between the base stations and the terminal device and the confidence level, the method comprises: obtaining a confidence level threshold; in response to the confidence level being less than or equal to the confidence level threshold, eliminating the confidence level and the distance; in response to the confidence level being greater than the confidence level threshold, retaining the confidence level and the distance.

6. The method of claim 1, wherein, After the step of inputting the positioning reference data into the parameter estimation model to estimate the distance between the base stations and the terminal device and the confidence level, the method comprises: obtaining a confidence level threshold; detecting a number of confidence levels greater than the confidence level threshold; in response to the number of confidence levels being less than or equal to a target number, sorting the confidence levels in descending order, and retaining the confidence levels and the distances of the front target number.

7. The method of claim 1, wherein, The step of assigning a weight to the distance according to the confidence level comprises: calculating a weight according to the confidence level, the weight being a ratio of the confidence level to a sum of all confidence levels; configuring the weight for the distance.

8. The method of claim 1, wherein, The step of calculating the current position of the terminal device according to the weight and the distance comprises: obtaining base station position coordinates of the base stations; constructing a circle equation according to the base station position coordinates and the distance; converting the circle equation into a linear equation; constructing a weight matrix according to the weight; calculating the current position based on the linear equation and the weight matrix using a least square method.

9. A wireless location system characterized by, The application is applied to a terminal device, comprising: a parameter estimation module, configured to receive positioning reference data sent by a plurality of base stations; input the positioning reference data into a parameter estimation model to estimate a distance between the base stations and the terminal device and a confidence level; the confidence level is used to represent an interference degree of a transmission path between the base stations and the terminal device, and the parameter estimation model is a pre-trained neural network model; A positioning module is configured to assign a weight to the distance according to the confidence degree, and calculate a current position of the terminal device according to the weight and the distance.

10. A computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program, when executed by a processor, implements the method of any one of claims 1 to 8.

11. A terminal device comprising a storage medium, a processor, and a computer program stored on the storage medium and executable on the processor, characterized in that, The processor, when executing the computer program, implements the method of any one of claims 1 to 8.

12. A chip comprising a processor and a memory for storing a computer program executable on the processor, characterized in that, The processor, when executing the computer program, implements the method of any one of claims 1 to 8. The processor, when executing the computer program, implements the method of any one of claims 1 to 8.

Images