Urban rail transit high-precision positioning method based on joint confidence selection

Abstract

This invention relates to the field of intelligent positioning technology, specifically to a joint confidence selection method for high-precision positioning in urban rail transit, comprising the following steps: constructing a wireless signal filtering model; estimating distances based on the received signal strength values ​​of base station nodes and satellite nodes to filter signal sources; calculating a geometric accuracy factor based on the estimated distance and received signal strength values ​​from the previous step; evaluating the positioning accuracy of the signal sources using the geometric accuracy factor; dynamically adjusting the weights of each node to generate preliminary positioning results; utilizing sensor data and the filtered wireless signals, optimizing device position by fusing multi-source information constraints through a factor graph framework; and iteratively optimizing the device attitude trajectory within the factor graph framework to minimize positioning errors. This joint confidence selection method for high-precision positioning in urban rail transit, by integrating multi-source information and a factor graph optimization algorithm, ensures high-precision positioning while reducing computational complexity, thus meeting the real-time positioning requirements of rail transit systems.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent positioning, and particularly relates to a high-precision positioning method for urban rail transit based on joint confidence selection. BACKGROUND

[0002] With the rapid development of urban rail transit systems and the increasing demand for intelligentization, precise positioning technology has become one of the core elements to ensure the efficient and safe operation of rail transit systems. However, in complex urban environments, a single sensor system often cannot meet the demand for high-precision positioning. Specifically, the existing technology faces the following challenges:

[0003] Limitations of visual SLAM (Simultaneous Localization and Mapping) technology: SLAM technology is susceptible to external factors such as weather and lighting, resulting in a significant decrease in positioning accuracy in harsh environments or poor lighting conditions, or even failure to work normally.

[0004] GNSS (Global Navigation Satellite System) signal blocking problem: GNSS signals are easily disabled in blocked environments such as tunnels and densely populated areas, making it impossible to provide stable and reliable positioning information, severely affecting the safety and operational efficiency of rail transit systems.

[0005] IMU (Inertial Measurement Unit) error accumulation: Although IMU can provide high-frequency pose information, its measurement error will accumulate over time, resulting in inaccurate long-term positioning results, especially in long-distance or long-time running scenarios.

[0006] LiDAR (Light Detection and Ranging) precision decline in dynamic environments: LiDAR's positioning accuracy will significantly decrease in dynamic environments or areas with sparse features, making it difficult to meet the demand for high-precision positioning.

[0007] To solve the above problems, multi-sensor data fusion technology becomes the key to improve positioning accuracy and system robustness. Currently, multi-source data fusion methods are mainly divided into two categories: filtering methods and optimization algorithms. Filtering methods, such as Kalman Filter (KF), Extended Kalman Filter (EKF), and Particle Filter (PF), achieve positioning by updating system state in real time, which is suitable for systems with high real-time requirements. However, in complex and variable rail transit environments, the performance of these methods is often limited. For example, EKF may fail to position in high dynamic environments, while PF is difficult to meet real-time requirements due to high computational complexity.

[0008] In contrast, Factor Graph Optimization (FGO) performs state estimation by minimizing the cost function, which shows higher accuracy and good flexibility and scalability when dealing with nonlinear systems. However, factor graph optimization has a large amount of computation, especially in dynamic rail transit environments, which may not meet real-time performance requirements.

[0009] With the gradual application of 5G and future 6G technology, the communication capability between terminals and infrastructure in rail transit systems has been significantly improved, providing rich data support for high-precision positioning. However, how to effectively fuse emerging communication technology with traditional sensor data is still a problem to be solved. Although 5G signals provide dense positioning references, they are limited by base station density and signal transmission conditions; Low Earth Orbit (LEO) signals are easily blocked in urban or tunnel environments. Therefore, it is necessary to dynamically evaluate signal quality and adaptively adjust signal source weights to improve the positioning accuracy and stability of the system.

[0010] Current signal selection methods mostly rely on Received Signal Strength Indicator (RSSI) or Signal-to-Noise Ratio (SNR), but these methods cannot fully reflect the positioning accuracy of signals. Geometric Dilution of Precision (GDOP) as an indicator to measure the influence of signal source geometric distribution on positioning accuracy has an advantage in selecting high-confidence signal sources. However, existing research has less combined GDOP with dynamic weight allocation, resulting in the inability to flexibly adjust signal source weights in different environments. In addition, traditional multi-source fusion frameworks ignore the timing synchronization problem of sensors. If the high-frequency data of IMU and the low-frequency data of GNSS or LiDAR are not accurately aligned, it may introduce large errors.

[0011] In summary, to solve the problems in the prior art, the present application proposes a high-precision positioning method for urban rail transit combined with confidence selection, aiming to dynamically select high-confidence signal sources by combining RSSI and GDOP dual indicators, and to realize efficient fusion of multi-source information by using a factor graph optimization framework, so as to improve positioning accuracy and reliability and meet the real-time positioning needs of urban rail transit systems. SUMMARY

[0012] To solve the above problems, the present application provides a high-precision positioning method for urban rail transit combined with confidence selection, which integrates multi-source information and factor graph optimization algorithms to ensure high-precision positioning while reducing computational complexity, thus meeting the real-time positioning needs of rail transit systems. This method can improve positioning accuracy under the synergistic action of multi-source data and achieve high robustness and real-time performance through factor graph optimization, making it suitable for high-precision positioning in urban rail transit systems, especially for precise positioning and navigation of trains, terminals and other equipment in rail transit scenarios, and effectively ensuring the efficient operation and safety of the transportation system.

[0013] The technical solutions of the present application are as follows:

[0014] The high-precision positioning method for urban rail transit combined with confidence selection comprises the following steps:

[0015] Step 1: Construct a wireless signal screening model to estimate the distance based on the received signal strength value of the base station node and the satellite node, and screen the signal source;

[0016] Step 2: Calculate the geometric dilution of precision based on the estimated distance and received signal strength value of step 1, evaluate the positioning accuracy of the signal source through the geometric dilution of precision, and dynamically adjust the weight of each node to generate a preliminary positioning result;

[0017] Step 3: Use sensor data and screened wireless signals to fuse multi-source information constraints through a factor graph framework to optimize the device position;

[0018] Step 4: Iteratively optimize the device attitude trajectory in the factor graph framework to minimize the positioning error.

[0019] Further, in step 1, the wireless signal screening model is:

[0020] ;

[0021] In the formula: is the received signal strength at a distance of d from the transmitter, dBm;

[0022] is the received signal strength at a distance of d from the transmitter, dBm; ​Reference signal strength at the location of the terminal device, dBm;

[0023] n is the path loss exponent, dimensionless; The degree of attenuation is determined by the path loss exponent n, which is generally a constant between 2-4: n=2 in free space; n is usually taken as 2.5-3.5 in open areas (such as suburbs, highways); n is generally 3.5-4 in dense urban areas or indoor environments;

[0024] is the current distance between the terminal device and the signal source, m;

[0025] is the reference distance, m;

[0026] is the random noise subject to normal distribution, dimensionless;

[0027] is the standard deviation of the noise, dB; represents the fluctuation amplitude of shadow fading, which is usually taken as dB in the V2X scenario, the more complex the environment is, the larger .

[0028] Further, in step 1, the signal source screening step directly excludes the signal source with a signal strength value lower than a preset threshold value.

[0029] Further, in step 2, the geometric dilution of precision includes the following steps:

[0030] Step 2.1: Establish a joint positioning model of the satellite and the base station:

[0031] ;

[0032] In the formula: is the observation vector, dimensionless;

[0033] is the parameter to be solved, dimensionless;

[0034] is the noise vector, is the satellite noise component, is the base station noise component, dimensionless;

[0035] is a nonlinear function in the joint positioning model, dimensionless.

[0036] The geometric dilution of precision GDOP is iteratively calculated by the following formula:

[0037] ;

[0038] In the formula: is the geometric dilution of precision, dimensionless;

[0039] Jacobian matrix of function , dimensionless

[0040] When the GDOP value is less than a preset threshold, the iteration is ended; the threshold typically ranges from 1.5 to 3.0, and needs to be determined comprehensively in combination with the accuracy requirement, signal source quality and iteration efficiency.

[0041] Step 2.2: The positioning result is determined by the following formula:

[0042]

[0043] In the formula: is a weighted positioning result, the unit of which is consistent with the coordinate system;

[0044] is a satellite signal positioning weight, dimensionless;

[0045] is a base station signal positioning weight, dimensionless;

[0046] is a positioning coordinate of the terminal using satellite, the unit of which is consistent with the coordinate system;

[0047] is a positioning coordinate of the terminal using 5G base station signal, the unit of which is consistent with the coordinate system.

[0048] Further, the satellite signal positioning weight and the base station signal positioning weight are calculated by the following formula:

[0049]

[0050]

[0051] In the formula: is a satellite geometric dilution of precision, dimensionless;

[0052] is a base station geometric dilution of precision, dimensionless.

[0053] Further, in step 3, the factor graph framework includes:

[0054] IMU factor node : constructed based on IMU measurement data, used for correcting the terminal attitude;

[0055] LiDAR factor node : constructed based on the relative pose change between LiDAR frames;

[0056] ​​​Dynamic Incremental Information Factor Node : Constructed based on the filtered satellite signals and base station signals.

[0057] Furthermore, IMU factor nodes Defined as:

[0058] ;

[0059] In the formula: For the first The IMU factor node function at time step 1 has dimensions that depend on the input variables. ;

[0060] For the first The dimensions of the system state variables at any given time need to be determined based on the specific composition of the state variables.

[0061] For the first The state estimate obtained solely from the IMU at a given time must have its dimensions determined based on the specific composition of the state variables.

[0062] The cost function has dimensions that depend on the input variables. .

[0063] Furthermore, lidar factor nodes Defined as:

[0064] ;

[0065] In the formula: Right now , for the first The time-based inter-frame matching factor node function of the lidar, the first row represents the position error between consecutive lidar frames, the unit is consistent with the coordinate system, the second row represents the rotation error between consecutive lidar frames, dimensionless;

[0066] and Represent and Pose estimation at time LO;

[0067] and Represent and Global pose estimation at time step, where The rotational component is dimensionless. This represents the position component, with units consistent with the coordinate system.

[0068] Furthermore, dynamic incremental information factor nodes Defined as:

[0069] ;

[0070] wherein: is the difference between the position observed at time and the model prediction information, with the same unit as the coordinate system;

[0071] is the position observation value provided by the dynamic incremental signal source after confidence selection at time ;

[0072] is the model prediction position function based on the system state at time ;

[0073] Further, in step 4, the terminal pose is optimized by maximizing the posterior probability density, and the objective function is:

[0074] ;

[0075] wherein: is the observation model, with the same unit as ;

[0076] is the measurement value, with different dimensions for different signal source types, but all related to position or distance.

[0077] The beneficial effects of the present application are:

[0078] 1. The urban rail transit high-precision positioning method of joint confidence selection disclosed in the present application dynamically selects high-confidence signal sources by combining the RSSI values of wireless signals (such as base station nodes and satellite nodes) and the geometric dilution of precision (GDOP) double indicators, effectively eliminates signal sources with poor signal quality, and thus improves the positioning accuracy. In specific implementation, a log-normal distribution model is used for RSSI ranging, and the positioning accuracy of various signal sources is evaluated through GDOP, and then the weights of each node are adjusted to obtain a more accurate preliminary positioning result.

[0079] 2. The joint confidence selection urban rail transit high-precision positioning method disclosed in the application, through the introduction of a factor graph optimization framework, fuses multi-source information data (including IMU, LiDAR and filtered wireless signals) as constraint factors, effectively improves the robustness of the system, and in a complex and variable urban rail transit environment, even if part of the sensor data is disturbed or fails, the system can still maintain high positioning performance through other reliable data sources.

[0080] 3. The joint confidence selection urban rail transit high-precision positioning method disclosed in the application, through multiple iterations of optimizing the device attitude trajectory, reduces the positioning error while ensuring the real-time performance of the algorithm.

[0081] 4. The joint confidence selection urban rail transit high-precision positioning method disclosed in the application, in view of the differences in signal quality in different environments (such as tunnels, high building sheltered areas, open spaces, etc.), can dynamically evaluate the signal quality and adaptively adjust the weight of the signal source. This environmental adaptability ensures the stable operation of the system in various complex scenarios, improves the reliability and availability of the positioning service.

[0082] 5. The joint confidence selection urban rail transit high-precision positioning method disclosed in the application, the multi-source data fusion method not only integrates wireless signals and sensor data, but also solves the problem of sensor time synchronization ignored in the traditional fusion framework. By accurately aligning the high-frequency data of IMU with the low-frequency data of GNSS or LiDAR, the error introduced by the time synchronization is reduced, and the efficiency and accuracy of data fusion are improved.

[0083] 6. The joint confidence selection urban rail transit high-precision positioning method disclosed in the application, compared with the traditional factor graph optimization method, through optimization algorithm design and implementation strategy, the computational complexity is effectively reduced, so that high-precision real-time positioning on resource-limited embedded devices is possible, further promoting the wide application of the technology in urban rail transit systems.

[0084] 7、The disclosed high-precision positioning method for urban rail transit based on joint confidence selection measures the performance of the optimized positioning system by evaluating the absolute trajectory error (ATE), and comprehensively evaluates the positioning accuracy using the root mean square error (RMSE), which ensures the accuracy and reliability of the positioning results and provides strong support for the safe and efficient operation of the rail transit system. BRIEF DESCRIPTION OF DRAWINGS

[0085] Fig. 1 An algorithm flowchart of an embodiment of the present application;

[0086] Fig. 2 A factor graph framework diagram of an embodiment of the present application;

[0087] Fig. 3 A positioning system implementation diagram of an embodiment of the present application. DETAILED DESCRIPTION

[0088] The present application will be further described in detail below with reference to the accompanying drawings. In different embodiments, similar elements are designated by similar reference numerals. In the following embodiments, many details are described in order to make the present application better understood. However, those skilled in the art can easily recognize that some features can be omitted or replaced by other elements, materials or methods in different cases. In some cases, some operations related to the present application are not shown or described in the specification in order to avoid the core part of the present application being overwhelmed by too much description, and it is not necessary to describe these related operations in detail for those skilled in the art based on the description in the specification and general technical knowledge in the art.

[0089] In addition, the features, operations or characteristics described in the specification can be combined in any appropriate manner to form various embodiments. At the same time, the steps or actions in the method description can also be sequentially adjusted or adjusted in a manner that is obvious to those skilled in the art. Therefore, the various sequences in the specification and drawings are only for the purpose of clearly describing a certain embodiment, and do not mean that the sequence is necessary, unless otherwise stated that a certain sequence must be followed.

[0090] REFERENCE Figs. 1-3 The high-precision positioning method for urban rail transit based on joint confidence selection mainly includes the following steps:

[0091] Step 1) Construct a wireless signal screening model, estimate the distance according to the RSSI value of the base station and satellite node, and screen out the signal source with larger signal strength, and exclude the source with weak signal:

[0092] (1) Wireless signal propagation model:

[0093] The terminal is not fixed in a mobile state, and the terminal is equipped with V2X (Vehicle-to-Everything, Chinese name: vehicle wireless communication technology or Internet of Vehicles communication technology).

[0094] Only consider the signal transmission path loss process, adopt the log-normal distribution model as the theoretical model of RSSI ranging, that is:

[0095] ;

[0096] Where, is the received signal strength at a distance of d from the transmitter, dBm;

[0097] is the reference signal strength received at a distance of d0, dBm; n is the path loss exponent, dimensionless;

[0098] The degree of RSSI attenuation is determined by the path loss exponent n.

[0099]

[0100] Generally, n takes a constant between 2-4; in free space, n=2;

[0101] In open areas (such as suburbs, highways), n is usually taken as 2.5-3.5;

[0102] In dense urban areas or indoor environments, n is generally 3.5-4.

[0103] is the current distance between the terminal device and the signal source, m;

[0104] is the reference distance, m;

[0105] is a random noise subject to normal distribution, satisfying , dimensionless;

[0106] σ is the standard deviation of the noise, dB; represents the fluctuation amplitude of shadow fading;

[0107] In the V2X scenario, usually dB, the more complex the environment, the larger σ.

[0108] ​​(2) Construct a wireless signal screening model:

[0109] In the RSSI screening process, sources with higher signal strength are preferentially selected, and a threshold value is set to avoid selecting sources with poor communication quality that may lead to high interruption probability; The minimum acceptable signal strength is -85dBm~ -80dBm in the V2X scenario.

[0110] When the received RSSI is lower than the threshold, the system determines that the signal quality is insufficient, which in turn leads to a high interruption probability.

[0111] The interruption probability can be calculated by the following formula:

[0112] ;

[0113] Where, is the interruption probability, dimensionless;

[0114] is the probability symbol, representing the probability of an event, dimensionless.

[0115] The Gaussian Q function is used to describe the signal interruption probability, which accurately models the impact of noise on positioning accuracy.

[0116] The calculation formula of the interruption probability can be expressed as:

[0117] ;

[0118] Where is the Gaussian Q function, defined as:

[0119] ;

[0120] Step 2) Calculate GDOP based on the distance and RSSI value estimated in step 1, evaluate the positioning accuracy of each type of signal source through GDOP, and adjust the weight of each node to obtain the preliminary positioning result:

[0121] (1) Calculate the GDOP value obtained by different signal sources for target terminal positioning:

[0122] Establish a satellite and base station joint positioning system, the model is as follows:

[0123] ;

[0124] Rewriting the above formula in vector form can be obtained:

[0125] ;

[0126] Where, is the observation vector, dimensionless;

[0127] is the parameter to be solved, dimensionless;

[0128] is the noise vector, dimensionless;

[0129] is the satellite noise component;

[0130] is the base station noise component;

[0131] is a nonlinear function in the joint positioning model.

[0132] A nonlinear least square method based on Gauss-Newton iterative method is used to approximate the nonlinear model of joint positioning, and the correction of is carried out through iteration.

[0133] After full differentiation, we can get:

[0134] ;

[0135] where is the Jacobian matrix of the function :

[0136] ;

[0137] For the Jacobian matrix calculated in each iteration, GDOP can be calculated as follows:

[0138] ;

[0139] where is the geometric dilution of precision, dimensionless;

[0140] is the transpose matrix of the Jacobian matrix calculated in the th iteration, dimensionless;

[0141] is the Jacobian matrix, dimensionless;

[0142] When the GDOP value is less than a preset threshold, the iteration is ended; the threshold typically ranges from 1.5 to 3.0, and needs to be determined comprehensively in combination with the accuracy requirement, signal source quality and iteration efficiency.

[0143] (2) Dynamic weight distribution:

[0144] The weights of satellite signal positioning and base station signal positioning are set as and GDOP value of each positioning result and is proportional to the reciprocal of GDOP value of each positioning result, i.e.:

[0145] ;

[0146] In order to further ensure that the weight distribution matches the contribution of the signal source, normalization processing is performed, and then the following can be obtained:

[0147] ;

[0148] The final weighted positioning result is a weighted average based on the weights of satellite signals and base station signals:

[0149] ;

[0150] In the formula, is the positioning coordinate of the terminal using satellite, unit: consistent with the coordinate system (such as degree (°, decimal latitude and longitude) under the global coordinate system WGS84, meter (m, east / north / skyward coordinate) under the local plane coordinate system (UTM / ENU));

[0151] is the positioning coordinate of the terminal using 5G base station signals, unit: consistent with the coordinate system;

[0152] is the weighted positioning result, unit: consistent with the coordinate system;

[0153] is the satellite signal positioning weight, dimensionless;

[0154] is the base station signal positioning weight, dimensionless.

[0155] Step 3) Optimize the device position through sensor data and filtered wireless signals, and use the factor graph framework to fuse multi-source information data as constraint factors:

[0156] Construct three different types of factor nodes, specifically including:

[0157] IMU factor node: constructed based on IMU measurement data, used to correct the terminal attitude;

[0158] Laser radar factor node: constructed based on the relative pose change between LiDAR frames;

[0159] Dynamic incremental information (DII) factor node: constructed based on filtered satellite signals and base station signals.

[0160] (1) IMU factor node:

[0161] ;

[0162] wherein: is the IMU factor node function at time t, dimension depends on input variables ;

[0163] is the system state variable at time t, dimension needs to be determined according to the specific composition of state variables; ;

[0164] is the state estimate only obtained by IMU at time t, dimension needs to be determined according to the specific composition of state variables; ;

[0165] The IMU factor node is the difference between and the current estimate , wherein the cost function is represented by , and the terminal pose estimate obtained by IMU is represented by .

[0166] (2) LiDAR factor node:

[0167] ;

[0168] wherein: is the laser radar inter-frame matching factor node function at time t, the first row represents the position error between consecutive radar frames, unit: consistent with the coordinate system, the second row represents the rotation error between consecutive radar frames, dimensionless; ;

[0169] and represent the pose estimates of LO at and ;

[0170] and represent the global pose estimates at and , wherein is the rotation component, dimensionless; is the position component, unit consistent with the coordinate system;

[0171] represents the measurement residual of the laser radar odometry, and represents the relative pose observation value obtained by LiDAR point cloud matching from time t-1 to time t;

[0172] represents the global constraint based on the global coordinate system, which is derived from the global pose at time t-1 derives the theoretical pose at time t;

[0173] represents the deviation between local observations and global predictions, which is minimized to optimize the global pose and achieve the fusion of local odometry and global constraints.

[0174] (3) DII factor node:

[0175] ;

[0176] where, represents the difference between the observed position at time t and the model prediction information after confidence selection, which quantifies the increment of the current state of the system, thus reflecting the size of the positioning error;

[0177] is the position observation value provided by the dynamic incremental signal source after confidence selection at time t, which is consistent with the coordinate system; is the model prediction position function based on the system state at time t, which is consistent with the coordinate system.

[0178] Step 4) Multiple iterations are performed in the factor graph framework to optimize the device pose trajectory and reduce the positioning error: The performance of the optimized positioning system is measured by evaluating the absolute trajectory error (ATE), and the root mean square error (RMSE) is used to comprehensively evaluate the positioning accuracy.

[0179] The terminal pose is optimized by maximizing the posterior probability density, and the objective function is:

[0180] ;

[0181] where,

[0182] is the observation model, which has the same dimension as ;

[0183] is the measurement value, which has different dimensions depending on the type of signal source, but is related to position or distance.

[0184] is the measurement value, which has different dimensions depending on the type of signal source, but is related to position or distance.

[0185] ​​​In conclusion, the present application provides a kind of urban rail transit high-precision positioning method of combined confidence selection, to overcome the defects in the prior art, improve positioning accuracy and reliability, by combining wireless signal, sensor data, factor graph framework and optimization algorithm, the present application realizes the efficient fusion of multi-source data, and optimizes the pose and trajectory of equipment;It is suitable for the joint optimization of wireless signals transmitted by base station nodes, satellite nodes and other sensor data such as inertial measurement unit (IMU) and laser radar (LiDAR).

[0186] In the description of the present application, the description of the terms "one embodiment", "some embodiments", "example", "specific example" or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In the present application, the illustrative description of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any suitable manner in any one or more embodiments or examples. Furthermore, the skilled in the art can combine and combine the different embodiments or examples described in the present application and the features of different embodiments or examples without contradiction.

[0187] The above application of specific examples is used to illustrate the present application, which is only used to help understand the present application, and does not limit the present application. For the skilled in the art, according to the idea of the present application, a number of simple deductions, deformations or substitutions can be made.

Claims

1. A high-precision positioning method for urban rail transit based on joint confidence selection, characterized in that, Comprising the following steps: Step 1: Construct a wireless signal screening model, estimate the distance according to the received signal strength value of the base station node and the satellite node, and screen the signal source; Step 2: Calculate the geometric dilution of precision based on the estimated distance and the received signal strength value in step 1, evaluate the positioning accuracy of the signal source through the geometric dilution of precision, dynamically adjust the weight of each node to generate a preliminary positioning result, including: (1) Calculate the GDOP value obtained by different signal sources for target terminal positioning: Establish a satellite and base station joint positioning system, the model is as follows: ； Rewrite the above formula in vector form: ； wherein is the observation vector, dimensionless; for the parameter to be found, dimensionless; Noise vector, dimensionless; satellite noise component; is a base station noise component; is a nonlinear function in the joint localization model; The non-linear least square method based on Gauss-Newton iteration is used to approximate and fit the non-linear model of joint positioning, and the iteration is used to correct ; After full differentiation, we get: ； wherein is the Jacobian matrix of the function ​ ； The Jacobian matrix computed for each iteration GDOP is computed by ； wherein, G is a geometric dilution of precision factor, dimensionless; For the first The transpose of the Jacobian matrix used in the next iteration is dimensionless. Jacobian matrix, dimensionless; When the GDOP value is less than the preset threshold, the iteration is ended; The threshold value is typically in the range of 1.5~3.0, and the specific value needs to be determined comprehensively according to the accuracy requirement, signal source quality and iteration efficiency; (2) Dynamic weight distribution: Setting weights for satellite signal positioning and base station signal positioning and GDOP values of the respective positioning results and are inversely proportional to the reciprocal of the GDOP values of the respective positioning results, i.e.: ； In order to further ensure that the weight distribution matches the contribution of the signal source, normalization processing is performed, and then the following formula is obtained: ； The final weighted positioning result is a weighted average based on the weights of the satellite signals and base station signals: ； In the formula, is the positioning coordinate of the terminal by the satellite, unit: consistent with the coordinate system; is the positioning coordinate of the terminal using the 5G base station signal, unit: consistent with the coordinate system; is the weighted positioning result, unit: consistent with the coordinate system; is a satellite signal positioning weight, dimensionless; is the base station signal positioning weight, dimensionless; Step 3: Use sensor data and screened wireless signals to fuse multi-source information constraints to optimize device location through the factor graph framework; Step 4: Iteratively optimize the device attitude trajectory in the factor graph framework to minimize the positioning error.

2. The urban rail transit high-precision positioning method of federated confidence selection according to claim 1, wherein, In step 1, the wireless signal screening model is: ； wherein: is the received signal strength at a distance of the transmission source in dBm; To receive the reference signal strength at a distance of 1 m in dBm; n is the path loss exponent, dimensionless; m is the current distance of the terminal device from the signal source; for reference distance, m; For random noise obeying a normal distribution, dimensionless; σ is the standard deviation of noise, dB.

3. The urban rail transit high-precision positioning method of federated confidence selection according to claim 2, characterized in that, In step 1, the signal source screening step directly excludes the signal source whose signal strength value is lower than the preset threshold value.

4. The urban rail transit high-precision positioning method of federated confidence selection according to claim 1, wherein, In step 3, the factor graph framework includes: IMU factor node : based on IMU measurement data construction, for correcting terminal attitude; Lidar factor node : construction based on lidar inter-frame relative pose changes; Dynamic delta information factor node : constructed based on the filtered satellite signals and base station signals.

5. The urban rail transit high-precision positioning method of federated confidence selection according to claim 4, characterized in that, IMU factor node is defined as: ； In the formula, is the first IMU factor node function at the time instant for the first instantaneous system state variable; For the first state estimate from the IMU only; is the cost function.

6. The urban rail transit high-precision positioning method of federated confidence selection according to claim 5, characterized in that, Lidar factor node is defined as: ； wherein: i.e. is the first is the laser radar inter-frame matching factor node function at time t, the first row represents the position error between consecutive radar frames, the second row represents the rotation error between consecutive radar frames, dimensionless. and represent respectively and pose estimate at time LO; and represent and instant global pose estimate.

7. The urban rail transit high-precision positioning method of federated confidence selection according to claim 4, characterized in that, Dynamic delta information factor node is defined as: ； In the formula: After the confidence selection signal source is placed The difference between the position observed at the moment and the model prediction information, the unit is consistent with the coordinate system; For the first The position observation value provided by the dynamic incremental signal source after confidence selection at the moment is consistent with the coordinate system unit. For the first instant system state of the model prediction position function, the unit is consistent with the coordinate system.

8. The urban rail transit high-precision positioning method of federated confidence selection according to claim 1, characterized in that, In step 4, the terminal pose is optimized by maximizing the posterior probability density, and the objective function is: ； wherein: is the observation model, and the units are the same as For measured values, the units are consistent with the signal source.

Images