A multi-scenario positioning enhancement method for Beidou and 5G integration

Abstract

The present invention relates to the field of navigation and positioning technology, and specifically to a multi-scenario positioning enhancement method for the integration of Beidou and 5G. 5G positioning data and Beidou positioning data of the area to be positioned are obtained; the 5G positioning data and Beidou positioning data are respectively filtered and processed and input into a multi-scenario positioning mode recognition system; the multi-scenario positioning mode recognition system determines the current positioning mode of the area to be positioned according to the received positioning data of the area to be positioned; the filtered 5G positioning data and Beidou positioning data are fused according to the determined positioning mode to obtain a fused positioning result; the fused positioning result is input into a neural network for training, and an enhanced positioning result is obtained according to the trained neural network. The present invention introduces filtering self-checking and neural network training, which effectively improves the accuracy and stability of the positioning system, thereby realizing positioning enhancement in multiple scenarios, and providing a solid foundation for the further development of indoor and outdoor positioning technology.

Description

Technical Field

[0001] The present invention relates to the field of navigation and positioning technology, and specifically to a multi-scenario positioning enhancement method for the integration of Beidou and 5G. Background Art

[0002] Advances in today's positioning technology have been widely used in fields including navigation, logistics, and security monitoring. Although the Global Positioning System (GPS) and other satellite navigation systems (such as the Beidou system) perform well in outdoor environments, in complex indoor environments, factors such as building structures and materials cause signal reflection and attenuation, which limits positioning accuracy. Therefore, traditional navigation and positioning technology faces challenges in accuracy and stability in complex indoor and outdoor environments, especially in transition areas where signal attenuation is severe. Single Beidou positioning and 5G base station observations are subject to problems such as insufficient signal coverage, environmental occlusion, and positioning mode switching, and are often unable to adapt to positioning in multiple scenarios. Summary of the invention

[0003] In order to solve the problems existing in the prior art, the present invention provides a multi-scenario positioning enhancement method for the integration of Beidou and 5G. The scheme includes: obtaining 5G positioning data and Beidou positioning data of the area to be positioned; filtering the 5G positioning data and Beidou positioning data respectively and inputting them into the multi-scenario positioning mode recognition system; the multi-scenario positioning mode recognition system determines the current positioning mode of the area to be positioned according to the received positioning data of the area to be positioned; according to the determined positioning mode, the filtered 5G positioning data and Beidou positioning data are fused to obtain the fused positioning result; the fused positioning result is input into the neural network for training, and the enhanced positioning result is obtained according to the trained neural network. The present invention introduces filtering self-checking and neural network training, which effectively improves the accuracy and stability of the positioning system, thereby realizing positioning enhancement in multiple scenarios, and providing a solid foundation for the further development of indoor and outdoor positioning technology.

[0004] The present invention adopts the following technical solution, a multi-scenario positioning enhancement method for the integration of Beidou and 5G, including:

[0005] The 5G positioning data of the area to be positioned is obtained through the fingerprint positioning method; the Beidou positioning data of the area to be positioned is obtained through the multi-star positioning method;

[0006] The 5G positioning data and the Beidou positioning data are filtered respectively, and the filtered 5G positioning data and the Beidou positioning data are input into a multi-scenario positioning pattern recognition system;

[0007] The multi-scenario positioning mode recognition system determines the current positioning mode of the area to be positioned according to the received positioning data of the area to be positioned;

[0008] The multi-scenario positioning mode recognition system fuses the filtered 5G positioning data and the Beidou positioning data according to the determined positioning mode to obtain a fused positioning result;

[0009] The fused positioning result is input into a neural network for training, and an enhanced positioning result is obtained according to the trained neural network.

[0010] Furthermore, the 5G positioning data of the area to be positioned is obtained by the fingerprint positioning method, including:

[0011] Collect RSSI data at multiple fixed points in the area to be located as fingerprint data, and establish a fingerprint data set based on the fingerprint data at the multiple fixed points;

[0012] Obtain the fingerprint data of any point in the area to be positioned and compare it with the fingerprint data set, take the fixed point with the closest comparison result as its corresponding position, and obtain the 5G positioning data of the area to be positioned.

[0013] Furthermore, the Beidou positioning data of the area to be positioned is obtained by a multi-star positioning method, including:

[0014] Assume that the coordinates of the three satellites in the Earth coordinate system are , , , the coordinates of the target radiation source in the area to be located are , the positioning equation group corresponding to the target radiation source is expressed as:

[0015] ;

[0016] in, is the distance from the target coordinate to the main star, is the distance from the target coordinates to the two auxiliary stars, is the electromagnetic wave transmission rate, Indicates the distance difference to different observation platforms;

[0017] By solving the positioning equation group of multiple target radiation sources in the area to be positioned, the Beidou positioning data of the area to be positioned is obtained.

[0018] Further, filtering is performed on the 5G positioning data and the Beidou positioning data respectively, including:

[0019] The Kalman filtering technology is used to filter the 5G positioning data and the Beidou positioning data, and the specific steps are as follows:

[0020] set up and They are The state quantity and observation quantity at the moment, the state equation and observation equation of the system are expressed as:

[0021] ;

[0022] in, is from arrive The state transfer matrix at time , is the system noise weighting matrix, yes The system noise at the moment, is the observation matrix, is the observation noise;

[0023] The Kalman filter algorithm is used to calculate the state estimation value and the error covariance, and the filtering processing of the 5G positioning data and the Beidou positioning data is completed by updating and iterating until the set conditions are met.

[0024] Furthermore, the multi-scenario positioning mode recognition system determines the current positioning mode of the area to be positioned according to the received positioning data of the area to be positioned, including:

[0025] The multi-scenario positioning pattern recognition system analyzes the received positioning data to obtain environmental characteristics of the positioning data; the environmental characteristics include: signal strength, multipath effect and noise level;

[0026] The positioning modes include: 5G positioning, Beidou positioning and 5G and Beidou integrated positioning;

[0027] The positioning modes are switched according to the environmental characteristics of the positioning data of the area to be positioned to determine the current positioning mode.

[0028] Furthermore, switching among various positioning modes is performed according to the environmental characteristics of the positioning data of the area to be positioned, including:

[0029] Detecting the number of signals of positioning data in the to-be-positioned area received by the multi-scenario positioning mode recognition system, and when only one signal is received, not switching, and using the positioning mode of the received signal as the current positioning mode;

[0030] When two signals are received at the same time, the positioning accuracy of the positioning mode corresponding to each signal is obtained, and the positioning mode with higher positioning accuracy is used as the current positioning mode;

[0031] When the positioning accuracies of the positioning modes corresponding to the two signals are the same, the joint positioning mode of the two positioning modes is used as the current positioning mode.

[0032] Further, the filtered 5G positioning data and the Beidou positioning data are fused to obtain a fused positioning result, including:

[0033] Assume the coordinates of the target terminal in the area to be located are , The coordinates of the BeiDou satellites are , where the receiver clock error is , The coordinates of the 5G base stations are , and get the joint positioning system model:

[0034] ;

[0035] Combining the observation equation of the multi-scene positioning pattern recognition system, we get:

[0036] ;

[0037] Among them, the parameters to be requested X = [ x , y , z , δ ] T , where the receiver clock error is , Indicates the distance from the first Beidou satellite to the target terminal. Indicates the distance from the Mth Beidou satellite to the target terminal, Indicates the distance from the first 5G base station to the target terminal. Indicates the distance from the Nth 5G base station to the target terminal, is the noise vector, The noise vector representing the distance from the first BeiDou satellite to the target terminal, The noise vector representing the distance from the Mth BeiDou satellite to the target terminal, The noise vector representing the distance from the first 5G base station to the target terminal, The noise vector representing the distance from the Nth 5G base station to the target terminal, function It is a nonlinear function in the joint localization model;

[0038] The nonlinear least squares method based on Gauss-Newton iteration method is used to approximate the system model of joint positioning, and the regression coefficient is corrected by iteration. , until the sum of squared errors converges to the preset threshold, the iteration ends and the fusion positioning result is obtained.

[0039] Furthermore, the fused positioning result is input into a neural network for training. The training process is specifically as follows:

[0040] The BP neural network is used to process the fusion positioning results, and the fault-tolerant processing and nonlinear relationship of the positioning data in complex environments are fitted by back propagation.

[0041] The BP neural network includes an input layer, a hidden layer and an output layer. The initial values ​​of the weights and thresholds of each layer of the neural network are set to random numbers, and the assignment error function is: , calculate the output of hidden layer neurons and the net output of the output layer neurons ;

[0042] According to the error function, the expected output value of the neuron is obtained The net output of the output layer neurons , and calculate the error estimate of the neural network ;

[0043] According to the error estimate and the neuron output of the current hidden layer Get the neuron output of the hidden layer at the next moment ;

[0044] The output of neurons in the hidden layer Reverse the weights of the hidden layer , complete the weight update, and update iteratively until the root mean square error of the neural network output meets the set threshold, and the neural network training is completed.

[0045] The beneficial effects of the present invention are: (1) Improving positioning accuracy and stability: By integrating multi-source positioning data and intelligent switching strategies, the accuracy and sustainability of positioning are significantly improved, especially in complex indoor and outdoor transition areas. The new technology does not rely solely on a single positioning signal, but comprehensively utilizes multiple signal sources from Beidou and 5G networks, thereby effectively overcoming signal obstruction and multipath propagation problems, ensuring that users obtain high-precision positioning services;

[0046] (2) Reduce the switching frequency of the positioning system: effectively reduce the frequent switching of the positioning system caused by signal attenuation or environmental changes, and ensure continuous location tracking and service stability. Through intelligent switching strategies and real-time environmental perception, the system can automatically optimize the signal source and reduce unnecessary switching, thereby improving user experience and system reliability;

[0047] (3) Adaptability to changing environments: Through intelligent algorithms and optimized training, the positioning system's ability to adapt to complex environments is enhanced, ensuring that it can still provide accurate location information under various challenging conditions. Whether in cities with dense high-rise buildings, shopping malls with complex indoor structures, or mountainous areas, the system can work stably and maintain high-precision positioning capabilities;

[0048] (4) Intelligent optimization and continuous improvement: Utilize neural network training to achieve continuous optimization and improvement of positioning algorithms. The system analyzes historical data and real-time feedback, optimizes algorithm parameters, and performs iterative learning to cope with increasingly complex and changing positioning needs, providing users with stable and reliable positioning services. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0050] Figure 1 A schematic diagram of the logical framework of a 5G and Beidou fusion enhanced positioning method for outdoor scenarios provided by an embodiment of the present invention;

[0051] Figure 2 A flow chart of a 5G and Beidou fusion enhanced positioning method for outdoor scenarios provided by an embodiment of the present invention;

[0052] Figure 3 A schematic diagram of a flow chart of an indoor 5G fingerprint positioning method provided by an embodiment of the present invention;

[0053] Figure 4 A schematic diagram of a process flow of an outdoor Beidou positioning technology provided by an embodiment of the present invention;

[0054] Figure 5 A schematic diagram of a filtering self-checking technology process provided by an embodiment of the present invention;

[0055] Figure 6 A schematic diagram illustrating an indoor and outdoor positioning area and a transition area positioning mode provided by an embodiment of the present invention;

[0056] Figure 7 A schematic diagram of a positioning switching algorithm flow based on multi-parameter judgment provided by an embodiment of the present invention;

[0057] Figure 8 A schematic diagram of a BP neural network algorithm flow chart provided in an embodiment of the present invention;

[0058] Fig. 9 A schematic diagram of a test track provided by an embodiment of the present invention;

[0059] Fig.10 A schematic diagram of a test scenario provided by an embodiment of the present invention;

[0060] Fig.11A schematic diagram of error curves in three directions during an outdoor-to-indoor positioning process provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0061] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0062] One of the purposes of this application is to provide a positioning technology solution that intelligently switches indoor and outdoor positioning modes and integrates Beidou satellite positioning and 5G base station observations to solve positioning challenges in complex indoor and outdoor environments, especially in transition areas with severe signal attenuation. To achieve the above purposes, the embodiments of this application provide the following: Figure 1 The positioning enhancement method shown realizes joint positioning and solution of position by comprehensively utilizing the data of Beidou satellite positioning and 5G base station observation.

[0063] Specifically, a flowchart of a multi-scenario positioning enhancement method for Beidou and 5G integration according to an embodiment of the present invention is as follows: Figure 2 As shown, including:

[0064] S1. Obtain raw positioning data from Beidou satellites and 5G base stations.

[0065] The embodiment of the present invention first establishes a fingerprint database for indoor positioning through the collected 5G signal strength fingerprint data; at the same time, the multi-star positioning method of "three-sphere intersection" is used to obtain Beidou positioning data for outdoor positioning. The specific method is as follows:

[0066] Indoor 5G positioning uses fingerprint library positioning to obtain location information through received signal strength indication (RSSI). After a certain number of reference points are evenly distributed indoors, the RSSI matrix of each reference point and the point to be measured is first obtained, and then their RSSI values ​​are compared to select the reference points with the smallest difference; finally, the weight coefficient of each reference point is calculated based on these differences, and weighted to obtain the final coordinates of the point to be measured. Figure 3 As shown in the figure, this method is divided into two stages: offline stage (building fingerprint database) and online stage (matching the data to be tested) to achieve positioning. The operation steps of the offline stage system include:

[0067] Environmental scanning and data collection: Assuming that there is The reference points are evenly spaced and the test points are randomly distributed indoors. The signal strength of each reference point can be obtained by measurement;

[0068] Fingerprint generation: For each reference point, the collected signal strength data forms a vector, which is stored together with the known coordinates of the reference point as the fingerprint of the location. The fingerprint information collected at each point is expressed as: ;in, Indicated in Fingerprint data collected from each point; The fingerprint data indicating this point feature information, taking RSSI value as fingerprint data, then Indicated in The strength value of the first base station signal collected at a fixed point;

[0069] Database construction: The fingerprints of all reference points are stored in a database. This database is a two-dimensional matrix, and each row represents the fingerprint data of a reference point:

[0070] ;

[0071] Among them, The true coordinate value of the point corresponding to the fingerprint data at the point is recorded as , then all coordinate true value data sets can be expressed as:

[0072] Rss = [ x 1 y 1 x 2 y 2 ⋮ ⋮ x i y i ] ;

[0073] Among them, the fingerprint dataset and true value coordinate dataset There is a one-to-one correspondence between them. and A one-to-one corresponding data pair is formed, which together constitute a complete fingerprint database.

[0074] The online positioning phase is used for actual position estimation and its steps include:

[0075] Collect the 5G signal strength at the current location: The mobile device to be located scans the surrounding 5G signals at the current location to form a fingerprint vector of the current location;

[0076] Fingerprint matching: The collected fingerprint vector is uploaded to the server, and the server uses the fingerprint similarity matching algorithm to compare and match the uploaded fingerprint vector with the fingerprint of each reference point in the database;

[0077] The distance between the point to be measured and the reference point is:

[0078] ;

[0079] in, Indicates the real collected data. Indicates The reference point and The Euclidean distance of the points to be located, so The smaller it is, the closer the point to be measured is to the reference point. The larger the value, the farther the distance. Therefore, for each target to be measured, find the closest The nearest reference point, and then according to The distance between the reference point and the point to be measured is used to calculate the weight coefficient of each reference point:

[0080] ;

[0081] in Indicates the nearest neighbor of the i-th test point The weight coefficient of each reference point.

[0082] Position estimation: Finally, the estimated position of the mobile device is determined by the reference point with the highest matching degree. The coordinates of this position are used as the estimated coordinates of the device to be located, so the final positioning coordinates are:

[0083] ;

[0084] in For the The coordinates of the nearest reference point.

[0085] In fingerprint-based indoor positioning technology, feature fingerprint data is unique. The fingerprint data of each location can uniquely reflect the distribution characteristics of the point in time and space, such as signal strength value, signal status information, and signal propagation delay. In an actual indoor environment, due to the complex propagation path of the signal and the ubiquitous existence of multipath effects, a location to be measured will receive signals from multiple base stations after the base station is fixed. Therefore, in order to achieve the accuracy of indoor positioning, the embodiment of the present invention uses RSSI in the 5G signal as the fingerprint data type. With the help of these fingerprint data and based on the distribution of all feature data in the known area to be measured, the fingerprint feature that best matches the point to be measured can be found, thereby achieving accurate indoor positioning.

[0086] For outdoor positioning, the embodiment of the present invention adopts multi-satellite positioning of "three-sphere intersection", and the target position coordinates can be obtained by jointly solving the positioning equations.

[0087] The BeiDou satellite navigation system uses the "three-sphere intersection measurement principle" for positioning. This principle requires the ground center station to use two geosynchronous satellites with known accurate positions as the sphere center and the distance from the satellite to the user as the radius to construct two spheres. At the same time, a third sphere is constructed with the radius of the earth plus the user's elevation as the radius. By calculating the intersection of these three spheres on the earth, the user's position can be determined. For targets on the earth's surface, the earth's surface can be regarded as another positioning surface, so the intersection of the three positioning surfaces is obtained as the target's position.

[0088] For the three observation platforms, since the target signal reaches the receivers of each platform at different distances, the time it takes to reach its receiver is also different. In the process of locating the target through the platform coordinates, the time difference between the target and the receivers of each platform is used as the main detection indicator, and the target position is determined by the arrival time difference.

[0089] In three-dimensional space, comparing the time difference between the same target signal and the receiving end of the platform in two different spaces can be considered as a half-double-leaf rotating hyperboloid centered on the two receivers. The intersection of the two rotating hyperboloids determined by the three platforms constitutes a time difference positioning curve, and each point on the curve represents a possible target position. Assuming the position coordinates of the target source and the position coordinates of the three observation platforms, the distance formula between the three targets and the platform is written using the simultaneous equations of the distance formula between two points. Then, based on the time-distance-speed formula, the time difference is converted into the distance difference to establish a formula. The platform between two of the three platforms is defined as the main platform, and the others are auxiliary platforms. The distance from the main platform and the auxiliary platform to the target position is set as .

[0090] The midpoint of the spherical model is the origin of the coordinate system. Assume that the rectangular coordinates of the three platforms and the rectangular coordinates of the target position are and , then the three-star time difference positioning equation group can be obtained as follows:

[0091] ;

[0092] in, is the distance between the target and the main platform, is the distance between the target and the auxiliary platform, is the speed of light, is the time difference of reaching different observation platforms, for and The distance difference, that is, the distance difference from the target to the main platform and the auxiliary platform, can be solved by solving the time difference positioning equation to obtain the target coordinates. The specific algorithm flow is as follows: Figure 4 As shown, the solution idea is as follows:

[0093] In the embodiment of the present invention, the analytical solution of the above three-star time difference positioning equation group is obtained by solving a high-order equation:

[0094] ;

[0095] In order to simplify the equation calculation, the geometric relationship can be used to set k i = 1 2 [ ( x 0 2 + y 0 2 + z 0 2 ) − ( x 1 2 + y 1 2 + z 1 2 ) + Δ r 1 2 ] , represents a partial solution to the linear system of equations, the earth's surface equation: , and the difference between it and the three-star time difference positioning equations can be obtained:

[0096] ;

[0097] in, , the above combination is expressed in matrix form as follows: ;

[0098] in, A = [ x 1 − x 0 y 1 − y 0 z 1 − z 0 x 2 − x 0 y 2 − y 0 z 2 − z 0 x 0 y 0 z 0 ] , X = [ x , y , z ] T , B = [ r 0 ⋅ Δ r 1 + k 1 r 0 ⋅ Δ r 2 + k 2 − 1 2 r 0 2 + k 3 ] ;

[0099] Let X contain the polynomial:

[0100] X = [ m 1 + p 1 ⋅ r 0 2 + n 1 ⋅ r 0 m 2 + p 2 ⋅ r 0 2 + n 2 ⋅ r 0 m 3 + p 3 ⋅ r 0 2 + n 3 ⋅ r 0 ] ;

[0101] In summary, we can get a The fourth-order equation is They represent the fourth-order equation Polynomial coefficients:

[0102] ;

[0103] The coefficients are:

[0104] ;

[0105] in, represents the coefficient of the quadratic term; represents the coefficient of the first-order term; represents a constant;

[0106] By solving the above quartic equation, we can get The solution of , then the target coordinates can be obtained .

[0107] S2. Filtering processing and effective data extraction: use filtering technology to filter the original data, identify and extract effective positioning data.

[0108] Step S2 of the embodiment of the present invention is specifically:

[0109] In order to effectively process the invalid information in the generated indoor and outdoor observation data, the present invention introduces a filtering self-checking technology. This technology takes into account the potential impact of the non-line-of-sight (NLOS) environment on the positioning results. Therefore, the Kalman filtering technology is used in the embodiment of the present invention to filter the observation data affected by the environment.

[0110] Kalman filtering is a recursive estimation method that calculates the estimated value of the current state through the estimated value of the state at the previous moment and the observed value of the current state. Therefore, there is no need to keep the historical records of observations or estimates. The key elements involved in Kalman filtering technology include system model, observation model, observation quantity and its covariance, and state quantity and its covariance. The algorithm is mainly divided into two core modules: system prediction and observation update to ensure effective processing and reduction of the negative impact of environmental interference on positioning accuracy, such as Figure 5 The figure shows the Kalman filtering process; in the prediction phase, the filter uses the estimate of the previous state to predict the current state; in the update phase, the filter optimizes the predicted value obtained in the prediction phase based on the observed value of the current state to obtain a more accurate new estimate.

[0111] set up and They are The state quantity and observation quantity at the moment, then the state equation and observation equation of the system can be expressed as:

[0112] ;

[0113] in, is from arrive The state transfer matrix at time , is the system noise weighting matrix, yes The system noise at the moment, is the observation matrix, is the observation noise.

[0114] The two states of the Kalman filter are represented by the following two variables:

[0115] Indicates at time The state estimate of , the superscript ^ indicates the estimated value;

[0116] Indicates known past The state of the tense, Prediction of momentary status;

[0117] is the posterior estimation error covariance matrix, which represents the accuracy of the metric estimate.

[0118] The self-check update process is as follows:

[0119] ;

[0120] In the predictive self-check stage, the state at the current moment is predicted based on the state and control quantity at the previous moment. This predicted value is an estimate because it has not yet considered the observed value at the current moment. The error covariance matrix of the predicted value is calculated through the error covariance matrix of the previous moment and the system noise covariance matrix.

[0121] ;

[0122] in, Indicated in Always estimates; Indicates the effect on State transition model on ; Indicates at time -1 is the state estimate; Represents the vector acting on the controller Input-control model on; is the estimated error covariance matrix from time k-1 to time k; For transformation model The transpose of is the estimated error covariance matrix at time k-1; represents the covariance matrix of process noise (mean is 0);

[0123] In the update step, the state estimate at the current moment is calculated based on the observed value and the predicted value at the current moment. This estimate is more accurate because the observed value at the current moment has been taken into account. The error covariance matrix of the state estimate is calculated by the error covariance matrix calculated in the prediction step, the observation noise covariance matrix and the Kalman gain.

[0124] ;

[0125] in, represents the optimal Kalman gain; is the estimated error covariance matrix from time k-1 to time k; is the observation model, which maps the real state space into the observation space; represents the transpose of the observation model; represents the covariance matrix; Indicates at time The updated state estimate of Indicates known past The state of the tense, Prediction of momentary status; represents the updated covariance matrix, is the estimated error covariance matrix updated from time k-1 to time k; represents the identity matrix;

[0126] The 5G positioning data and Beidou positioning data after filtering and self-checking are respectively input into the multi-scenario positioning mode recognition system. The system intelligently identifies the current positioning scene type, such as indoor, outdoor or transition area, according to environmental characteristics and signal conditions, and selects the most appropriate positioning mode for switching accordingly.

[0127] S3. Scene recognition, determining whether the current area to be positioned belongs to the indoor-outdoor transition area.

[0128] The step of scene recognition in the embodiment of the present invention specifically includes:

[0129] Data transmission and reception: After filtering and self-checking, 5G positioning data and Beidou positioning data are transmitted to the multi-scenario positioning pattern recognition system through appropriate communication channels. and the observed quantity Indicates that it contains relevant information such as position, speed, time, etc.

[0130] Analysis of environmental characteristics and signal conditions: The multi-scenario positioning pattern recognition system receives and parses the incoming positioning data. The system analyzes the characteristics of the current environment, such as signal strength, multipath effects, noise levels, etc., as well as the quality and availability of the received 5G and Beidou signals;

[0131] Scene type judgment: According to the analysis of environmental characteristics and signal conditions, the preset model is used to judge the scene type of the current positioning, such as indoor, outdoor or transitional areas (such as urban high-rise dense areas or mountainous areas). The positioning process involves three different positioning modes. Judging the optimal mode is the basis for ensuring smooth switching of overlapping areas; therefore, the indoor and outdoor positioning pattern recognition system based on neural networks aims to integrate key parameter training sets by training neural networks, repeatedly adjusting the learning rate and the number of neurons to obtain the optimal positioning recognition model; when considering the structure of indoor and outdoor positioning systems, the key part is to judge the optimal positioning mode based on the existing parameter information. In the transition from outdoor to indoor, the receiver terminal needs to complete the switching of the three positioning modes. In practical applications, the number of visible Beidou satellites, the number of 5G base stations and the Beidou positioning prior root mean square error can be used as the input signal vector of the training sample.

[0132] S4. Positioning mode switch, select the most suitable positioning mode according to the environmental characteristics.

[0133] The indoor and outdoor positioning proposed in the embodiment of the present invention needs to integrate the positioning functions of two different systems to provide accurate, stable and smooth positioning services. The scene switching usually involves the switching of heterogeneous networks. The positioning system switching has the following characteristics:

[0134] Prior location information: Deployed positioning base stations usually have accurate location calibration. During the positioning process, the receiver location information is collected in real time and logged. The positioning results of the previous period can be used as a reference, but relying solely on location information for handover decisions may not be reliable enough, as environmental influences may lead to positioning errors.

[0135] Reception of multi-source positioning information: The positioning base station can serve multiple users simultaneously during the positioning process, and the receiver can receive positioning information from multiple signal sources, which is helpful for data fusion processing and optimizing switching decisions.

[0136] Signal extension zone: Even in the GNSS system, the receiver can still communicate with the satellite to obtain necessary information when it cannot directly locate. The prior information in the extension zone can estimate the possibility of the next handover and prepare for the handover in advance.

[0137] Heterogeneous positioning system: Indoor and outdoor positioning is usually completed by different systems working together. It is necessary to consider the positioning accuracy and the optimization of weight distribution between systems to meet user needs.

[0138] Switching area: The signal coverage of the positioning system is relatively limited, and switching includes inter-system switching and system edge switching. The former completes the switching between different systems through decision strategies, while the latter involves the judgment and identification of the receiving terminal at the signal coverage boundary.

[0139] In the embodiment of the present invention, the system first detects the positioning area where the receiving terminal is located. If only one positioning system signal is received, there is no need to trigger the handover. The handover is only triggered when two positioning system signals are received at the same time, and the handover decision process is started. The handover decision is the core of the whole process. The system needs to accurately evaluate the quality, stability and handover frequency limit of the current positioning signal to decide which positioning system signal to use or how to merge the information of the two. The system positioning result is , the weight is , The system positioning result is , the weight is , then the positioning result after data fusion is:

[0140] ;

[0141] Then in a certain direction, the positioning error after weighted fusion is:

[0142] ;

[0143] in, and They represent the positioning accuracy of the system working independently, is the joint positioning accuracy, assuming , then:

[0144] ;

[0145] Solving the above formula yields:

[0146] ;

[0147] After finishing, we can get:

[0148] ;

[0149] when When it is close to 1, it means that the accuracy of the two positioning systems is close, and the joint positioning effect is better; conversely, if the accuracy of system A is much higher than that of system B, the joint positioning result will be biased towards system A, and the performance cannot be improved. This conclusion is based on The premise is established, and vice versa; therefore, it can be inferred that the Beidou and 5G joint positioning system provides the best positioning result when the 5G independent positioning accuracy is similar to the Beidou independent positioning accuracy; when the receiving terminal is about to leave the signal receiving area of ​​a certain system, it should consider switching to indoor or outdoor independent positioning; therefore, the best time to switch is: when the signal quality of one system deteriorates rapidly, while the signal quality of the other system continues to improve.

[0150] It is worth mentioning that in the area where the positioning system signals overlap, the system may frequently switch the positioning mode due to fluctuations in signal quality, resulting in waste of resources and reduced positioning accuracy. In order to eliminate the ping-pong effect, the system needs to set appropriate switching logic and parameter thresholds to ensure a stable and effective positioning process. The parameter threshold can be set according to actual conditions, and the present invention does not impose any restrictions. At the same time, when the switch is triggered, the system can obtain prior information transmitted by the new positioning source, such as signal quality, error range, etc. This information is crucial for deciding the optimal positioning system or fusing positioning information, ensuring the accuracy and smoothness of the switching process.

[0151] When moving from outdoor to indoor, the receiver terminal needs to switch between three positioning modes; Figure 6 As shown in the figure, the Beidou and 5G-based systems involve three different positioning modes; the key to ensuring smooth switching is to accurately determine the current optimal positioning mode. Therefore, the goal of the neural network-based positioning mode recognition system is to use the training set, integrate key parameters, and train the optimal positioning recognition model by adjusting the learning rate and the number of neurons.

[0152] Key parameters include:

[0153] The number of Beidou visible satellites (SatNum) is the number of Beidou satellite signals that the receiver can receive at a certain moment in the positioning process; the number of 5G base stations (BaseNum) is the number of 5G base station signals that the receiver can receive at a certain moment in the positioning process; the Beidou positioning prior root mean square error (PreRMS) is the average value of the RMS of the Beidou single-point positioning data recorded at the previous 10 moments as the prior error at the next moment; the minimum TOA detection amount (minTOA) is the TOA observation amount of the 5G base station closest to the receiver.

[0154] The update formulas for the above parameters are derived as follows. training samples :

[0155] ;

[0156] in, Indicates the number of visible Beidou satellites. Indicates the number of 5G base stations, represents the prior root mean square error of Beidou positioning, Indicates the minimum TOA detection amount;

[0157] Assume the neural network output According to the chain rule, it is easy to deduce the update formula of the three-layer neural network. Taking the cross entropy as the cost function and taking the natural base, the error function is:

[0158] ;

[0159] For The output components are:

[0160] ;

[0161] in, Indicates that 3 input As an index The power of each power is divided by the sum of the three powers. The value of

[0162] In summary, we get the parameter update formula in the neural network model for indoor and outdoor positioning pattern recognition. The specific algorithm flow is as follows: Figure 7 As shown; according to the results of pattern recognition, the positioning mode can be flexibly switched: in indoor environments, 5G positioning technology is preferred; in outdoor environments, Beidou positioning technology is used. In transition areas, a joint positioning strategy is used to ensure the continuity and accuracy of positioning.

[0163] S5. Multi-source data fusion: Fusion of processed Beidou satellite positioning and 5G base station observation data.

[0164] When the joint positioning mode is used in the indoor and outdoor transition area, the Beidou satellite signal is severely attenuated due to building obstructions in the area where the Beidou signal and the 5G signal overlap, and the receiver may not be able to see enough Beidou satellites (less than 4) in its field of view, resulting in failure to position normally. Under normal circumstances, the receiver can receive the positioning signal of the 5G base station. Therefore, integrating the measurement data of the 5G base station into the positioning solution can make up for the insufficient number of Beidou satellites, significantly improve the positioning accuracy in the signal overlapping area, and ensure the overall positioning performance.

[0165] Assume that the target terminal coordinates are , The coordinates of the BeiDou satellites are , The coordinates of the 5G base stations are , we can get the joint positioning system model:

[0166] ;

[0167] Rewriting the observation equation in vector form yields:

[0168] ;

[0169] in, ;

[0170] Among them, the parameters to be requested X = [ x , y , z , δ ] T , where the receiver clock error is , Indicates the distance from the first Beidou satellite to the target terminal. Indicates the distance from the Mth Beidou satellite to the target terminal, Indicates the distance from the first 5G base station to the target terminal. Indicates the distance from the Nth 5G base station to the target terminal, is the noise vector, The noise vector representing the distance from the first BeiDou satellite to the target terminal, The noise vector representing the distance from the Mth BeiDou satellite to the target terminal, The noise vector representing the distance from the first 5G base station to the target terminal, The noise vector representing the distance from the Nth 5G base station to the target terminal, function It is a nonlinear function in the joint localization model;

[0171] The embodiment of the present invention adopts a nonlinear least square method based on the Gauss-Newton iteration method to approximate the nonlinear model of joint positioning, and iterates the regression coefficients. Correction is performed until the sum of squares of convergence errors meets the iteration threshold. The specific steps are as follows:

[0172] Perform total differentiation with respect to X, that is ,in, For function The Jacobian matrix of ; convergence error: , then according to the convergence error, the cost function can be defined :

[0173] ;

[0174] right About parameters Find the partial derivative:

[0175] ;

[0176] In nonlinear systems, Describes the function of variables and parameters, has no closed solution, so give Set the initial value and use the iteration method to approximate the closed solution. The update formula is as follows:

[0177] ;

[0178] Indicates the current iteration number, is the updated value of the parameter for each iteration, exist Use the Taylor series expansion formula to expand and retain the first-order term:

[0179] ;

[0180] in, For the Iteration Matrix, then:

[0181] ;

[0182] like , so the convergence error is:

[0183] ;

[0184] Comprehensive considerations yield:

[0185] ;

[0186] The above is the final iterative formula. When the convergence error is less than the preset value, the iteration ends and the result is used as the final joint positioning result. Therefore, the data after filtering, pattern recognition and switching processing is input into the data processing channel for information fusion. At this time, the system will consider the preset information allocation factors, such as the number of available satellites and positioning accuracy requirements, to generate preliminary 5G and Beidou fusion positioning results, further improving the overall positioning accuracy and reliability.

[0187] S6. Neural network optimization training: Use the neural network model to further optimize and train the initially integrated positioning data.

[0188] In order to optimize the positioning algorithm to adapt to different environments, the present invention will also input the neural network model for training after obtaining the fusion positioning result. The training process is as follows: Figure 8 As shown, this process is designed to improve the performance and stability of the positioning system and ensure that the system can provide accurate and stable positioning services in various application scenarios.

[0189] In the embodiment of the present invention, BP neural network is used for data processing, and fault-tolerant processing of positioning data in complex environments and fitting of nonlinear relationships are achieved through back propagation. BP neural network is usually composed of input layer, hidden layer and output layer, and the network structure is constructed through dense connection, so as to achieve modeling of the relationship between positioning results and position coordinates in different scenarios. The initial values ​​of weights and thresholds of each layer of the neural network are set to small random numbers, and the assignment error function is: , calculate the output of hidden layer neurons and the net output of the output layer neurons ;

[0190] set up , according to the error function, we get the expected output value of the neuron The net output of the output layer neurons , and calculate the error estimate of the neural network ;

[0191] According to the error estimate and the neuron output of the current hidden layer Get the neuron output of the hidden layer at the next moment ;

[0192] The output of neurons in the hidden layer Reverse the weights of the hidden layer , complete the weight update;

[0193] The hidden layer weights are iteratively updated in the above steps until the root mean square error of the neural network output meets the set threshold. The neural network training is completed and the optimized position prediction model is obtained. Then, the hidden layer neuron output calculated in the above step S6 can be It is expressed as:

[0194] ;

[0195] in, The input layer neurons to the hidden layer The connection weights of neurons; represents the hidden layer Neuron threshold; Represents the output of hidden layer neurons; Represents the input state quantity, A network representing the outputs of the hidden layer neurons;

[0196] Calculate the net output of the output layer neurons It can be expressed as:

[0197] ;

[0198] in, Represents the output layer weight matrix; Represents the current hidden layer Output of neurons at each moment; Represents the neuron output of the hidden layer at the next moment, Represents the hidden layer The threshold of the neuron at that moment, is the error estimate of the neural network.

[0199] S7. Positioning accuracy and stability evaluation: Comprehensively evaluate the optimized positioning results.

[0200] The embodiments of the present invention have been comprehensively evaluated, including the consideration of positioning accuracy, stability and response time. The positioning system can meet the needs of various practical application scenarios. In the evaluation process, the root mean square error (RMSE) is one of the main indicators for measuring positioning accuracy. It reflects the accuracy of positioning by calculating the average square root error between the predicted position and the actual position. A smaller RMSE value usually means that the system can provide more accurate positioning results, which is suitable for scenarios with higher requirements for positioning accuracy, such as navigation and precise positioning services. In the multi-scenario positioning for Beidou and 5G integration, it is assumed that the actual position of the positioning point is , the estimated position after initial training , the root mean square error definition expression is:

[0201] RMSE = 1 n ∑ i = 1 n [( x − x i ) 2 + ( y − y i ) 2 ] ;

[0202] On the other hand, the cumulative probability distribution function (CDF) provides a comprehensive perspective on the distribution of positioning errors. By drawing the cumulative distribution curve of positioning errors, it reflects the distribution of errors in different ranges. The ideal CDF curve should rise at a faster rate, indicating that most positioning errors are concentrated in a smaller range, which is crucial to ensuring the stability and reliability of the positioning system in a variety of scenarios.

[0203] Assume that the positioning error is , the threshold value is , the total number is , the formula used to calculate the probability value is as follows:

[0204] ;

[0205] in, Function used for statistical error Less than threshold value The number of

[0206] S8. Model adjustment and learning: Adjust neural network parameters based on real-time feedback and historical data to improve positioning adaptability in complex environments and transition areas.

[0207] In the embodiment of the present invention, by performing real-time feedback and historical data analysis on the neural network model parameters, the system's adaptability to complex environments and transitional areas is further optimized. This process continuously enhances the effect of the positioning system and ensures that users can obtain high-precision and reliable positioning services in various situations. Specifically, the following steps are included:

[0208] Real-time feedback collection: The positioning system collects data from multiple positioning devices in real time, including information from the Beidou satellite system and 5G networks, while taking into account environmental conditions such as terrain and buildings;

[0209] Historical data analysis: Analyze the large amount of positioning data collected previously, including the distribution, frequency, and influencing factors of positioning errors. This step helps identify common positioning problems and patterns, providing a basis for the next step of model adjustment;

[0210] Neural network parameter adjustment: According to the results of real-time feedback and historical data analysis, the parameters of the neural network model are adjusted. The goal of the adjustment is to optimize the model's fitting ability and ensure that it can accurately capture the positioning features of complex environments and transition areas.

[0211] Model validation and testing: The adjusted neural network model is tested and validated, and its positioning accuracy and stability in complex environments are evaluated using independent datasets to ensure that the new model performs well in a variety of practical usage scenarios.

[0212] Final enhancement result: After multiple adjustments and verifications, the final multi-scenario positioning enhancement method for the integration of Beidou and 5G was obtained, which verified that the present invention can provide accurate and stable positioning services in complex environments and transition areas, and meet users' high-quality positioning needs.

[0213] In another specific embodiment of the present invention, the present invention designs a comprehensive hardware system, including Beidou satellite positioning receivers and 5G base station observation equipment, and ensures that these devices are compatible with each other and can effectively transmit data, while selecting high-performance antennas and receivers to ensure good signal reception and processing capabilities in complex environments; further develop a software platform that supports multiple positioning algorithms, including positioning algorithms based on satellite signals and 5G signals. These algorithms include but are not limited to time difference measurement (TDOA), Doppler measurement, and signal strength fingerprint positioning technology, thereby implementing an efficient data fusion algorithm to integrate data from different signal sources and improve positioning accuracy and robustness.

[0214] like Fig. 9 As shown in the figure, the four indoor 5G base stations constitute the test environment. The dotted line indicates the moving route from outdoor to indoor. The mobile terminal starts from outdoor and is equipped with Beidou and 5G positioning signal receivers and a joint positioning system to ensure that it can receive good Beidou satellite signals. The mobile terminal moves slowly along the route shown in the figure until it enters the indoor area. The entire positioning process lasts about 3 minutes, and one test point is calibrated per second.

[0215] After data preprocessing, a total of 150 valid points are obtained, among which the black x represents the reference trajectory, and the five-pointed star represents the trajectory map output by the indoor and outdoor positioning systems. Although there are certain deviations in local areas, the overall trend is consistent with the reference trajectory; in the indoor part, the change of the trajectory is due to the switch to the indoor positioning system, and the effect of indoor positioning is usually better than outdoor positioning and combined positioning. Therefore, the feasibility of the present invention is verified through actual measurements, and the algorithm and parameter settings of the system are optimized according to the test results to ensure that the system can stably and reliably provide high-precision positioning services.

[0216] Through the above detailed implementation steps, such as Fig. 9 and Fig.10 As shown, the test trajectory schematic diagram and scenario schematic diagram shown in the present invention. The multi-scenario positioning enhancement method for Beidou and 5G integration proposed in the present invention can effectively cope with positioning challenges in different environments, and provide technical support and solutions for the development of indoor and outdoor positioning technologies.

[0217] Fig.11The error curves in three directions during the outdoor-to-indoor positioning process are shown. The positioning error and system switching process can be seen in the figure. The initial positioning mode is Beidou single-point positioning, which switches to the joint positioning mode from around the 95th sampling point. After the 120th sampling point, the error curve stabilizes, indicating that it has switched to the 5G indoor positioning mode.

[0218] The embodiment of the present invention is oriented to the field of navigation and positioning technology, and proposes a positioning enhancement method for the integration of Beidou and 5G. Based on predictive filtering, it combines scene recognition and positioning mode switching, and adopts a neural network algorithm to predict the position of indoor and outdoor and transition areas. The method proposed by the present invention aims to cope with the challenges of positioning accuracy and stability in complex indoor and outdoor environments, especially in transition areas with severe signal attenuation, by intelligently switching indoor and outdoor positioning modes and integrating Beidou positioning and 5G base station observations. The positioning data is processed using filtering self-checking technology, invalid information is eliminated, environmental occlusion factors are considered, and the best positioning strategy is selected according to the specific environment. Subsequently, the positioning accuracy is further improved through neural network optimization training to ensure the acquisition of high-precision location information. At the same time, the method proposed by the present invention also introduces pattern recognition conditions to effectively avoid positioning jumps and repeated invalid positioning switching problems, and effectively improves the positioning accuracy in multiple scenarios.

[0219] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A multi-scenario positioning enhancement method for Beidou and 5G integration, characterized in that: include: Obtain 5G positioning data of the area to be positioned through the fingerprint positioning method; Obtain Beidou positioning data of the area to be positioned through a multi-star positioning method; The 5G positioning data and the Beidou positioning data are filtered respectively, and the filtered 5G positioning data and the Beidou positioning data are input into a multi-scenario positioning pattern recognition system; The multi-scenario positioning mode recognition system determines the current positioning mode of the area to be positioned according to the received positioning data of the area to be positioned; The multi-scenario positioning mode recognition system fuses the filtered 5G positioning data and the Beidou positioning data according to the determined positioning mode to obtain a fused positioning result; Specifically include: Assume the coordinates of the target terminal in the area to be located are , The coordinates of the BeiDou satellites are , The coordinates of the 5G base stations are , and get the joint positioning system model: ; Combining the observation equation of the multi-scene positioning pattern recognition system, we get: ； Among them, the parameters to be requested , where the receiver clock error is , Indicates the distance from the first Beidou satellite to the target terminal. Indicates the distance from the Mth Beidou satellite to the target terminal, Indicates the distance from the first 5G base station to the target terminal. Indicates the distance from the Nth 5G base station to the target terminal, is the noise vector, The noise vector representing the distance from the first BeiDou satellite to the target terminal, The noise vector representing the distance from the Mth BeiDou satellite to the target terminal, The noise vector representing the distance from the first 5G base station to the target terminal, The noise vector representing the distance from the Nth 5G base station to the target terminal, function It is a nonlinear function in the joint localization model; The nonlinear least squares method based on Gauss-Newton iteration method is used to approximate the system model of joint positioning, and the regression coefficient is corrected by iteration. , until the sum of squared errors converges to the preset threshold, the iteration ends and the fusion positioning result is obtained; The fused positioning result is input into a neural network for training, and an enhanced positioning result is obtained according to the trained neural network.

2. According to claim 1, a multi-scenario positioning enhancement method for Beidou and 5G integration is characterized in that: The 5G positioning data of the area to be positioned is obtained through the fingerprint positioning method, including: Collect RSSI data at multiple fixed points in the area to be located as fingerprint data, and establish a fingerprint data set based on the fingerprint data at the multiple fixed points; Obtain the fingerprint data of any point in the area to be positioned and compare it with the fingerprint data set, take the fixed point with the closest comparison result as its corresponding position, and obtain the 5G positioning data of the area to be positioned.

3. According to claim 1, a multi-scenario positioning enhancement method for Beidou and 5G integration is characterized in that: The Beidou positioning data of the area to be positioned is obtained through the multi-star positioning method, including: Assume that the coordinates of the three satellites in the Earth coordinate system are , , , the coordinates of the target radiation source in the area to be located are , the positioning equation group corresponding to the target radiation source is expressed as: ； in, is the distance from the target coordinate to the main star, is the distance from the target coordinates to the two auxiliary stars, is the electromagnetic wave transmission rate, is the time difference of reaching different observation platforms, Indicates the distance difference to different observation platforms; By solving the positioning equation group of multiple target radiation sources in the area to be positioned, the Beidou positioning data of the area to be positioned is obtained.

4. According to claim 1, a multi-scenario positioning enhancement method for Beidou and 5G integration is characterized in that: The 5G positioning data and the Beidou positioning data are filtered respectively, including: The Kalman filtering technology is used to filter the 5G positioning data and the Beidou positioning data, and the specific steps are as follows: set up and They are The state quantity and observation quantity at the moment, the state equation and observation equation of the system are expressed as: ； in, is from arrive The state transfer matrix at time , is the system noise weighting matrix, yes The system noise at the moment, is the observation matrix, is the observation noise; The Kalman filter algorithm is used to calculate the state estimation value and the error covariance, and the filtering processing of the 5G positioning data and the Beidou positioning data is completed by updating and iterating until the set conditions are met.

5. The multi-scenario positioning enhancement method for Beidou and 5G integration according to claim 1 is characterized in that: The multi-scenario positioning mode recognition system determines the current positioning mode of the area to be positioned according to the received positioning data of the area to be positioned, including: The multi-scenario positioning pattern recognition system analyzes the received positioning data to obtain environmental characteristics of the positioning data; the environmental characteristics include: signal strength, multipath effect and noise level; The positioning modes include: 5G positioning, Beidou positioning and 5G and Beidou integrated positioning; Switch among various positioning modes according to the environmental characteristics of the positioning data of the area to be positioned to determine the current positioning mode.

6. The multi-scenario positioning enhancement method for Beidou and 5G integration according to claim 5 is characterized in that: Switch between various positioning modes according to the environmental characteristics of the positioning data of the area to be positioned, including: Detecting the number of signals of positioning data in the to-be-positioned area received by the multi-scenario positioning mode recognition system, and when only one signal is received, not switching, and using the positioning mode of the received signal as the current positioning mode; When two signals are received at the same time, the positioning accuracy of the positioning mode corresponding to each signal is obtained, and the positioning mode with higher positioning accuracy is used as the current positioning mode; When the positioning accuracies of the positioning modes corresponding to the two signals are the same, the joint positioning mode of the two positioning modes is used as the current positioning mode.

7. The multi-scenario positioning enhancement method for Beidou and 5G integration according to claim 1 is characterized in that: The fused positioning result is input into the neural network for training. The training process is specifically as follows: The BP neural network is used to process the fusion positioning results, and the fault-tolerant processing and nonlinear relationship of the positioning data in complex environments are fitted by back propagation. The BP neural network includes an input layer, a hidden layer and an output layer. The initial values ​​of the weights and thresholds of each layer of the neural network are set to random numbers, and the assignment error function is: , calculate the output of hidden layer neurons and the net output of the output layer neurons ; According to the error function, the expected output value of the neuron is obtained The net output of the output layer neurons , and calculate the error estimate of the neural network ; According to the error estimate and the neuron output of the current hidden layer Get the neuron output of the hidden layer at the next moment ; The output of neurons in the hidden layer Reverse the weights of the hidden layer , complete the weight update, and update iteratively until the root mean square error of the neural network output meets the set threshold, and the neural network training is completed.

Images