Map matching method based on GNSS covariance information

Abstract

A map matching method based on GNSS covariance information comprises the steps that after RINEX, GGA and GST message data are collected, the protection level (PL) is calculated based on an RAIM algorithm, a candidate matching point set is generated with the protection level (PL) as a constraint, then a covariance matrix in a GST message is used for establishing a two-dimensional Gaussian error model to calculate the emission probability (EP), and finally a map matching result is obtained. Calculating a transition probability (TP) according to a path distance between adjacent epochs and a GNSS spherical distance; and searching a path with the maximum product of the EP and the TP through a dynamic programming algorithm, and backtracking to obtain a global optimal track. The method can effectively suppress multipath and non-line-of-sight errors in a complex urban environment, realizes synchronous improvement of matching precision and credibility, has the advantages of high robustness, quantifiable credibility, real-time calculation efficiency and the like, and is suitable for scenes such as automatic driving, intelligent traffic and high-precision navigation.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of navigation positioning, and particularly relates to a map matching method based on GNSS covariance information. BACKGROUND

[0002] Existing positioning methods generally depend on absolute position information provided by GNSS, but in complex environments such as urban canyons, shaded roads or viaducts, GNSS signals are easily affected by shielding, multipath effects and non-line-of-sight (NLOS) interference, resulting in a decrease in positioning accuracy or even a serious deviation, thereby adversely affecting the path planning and control decision of an autonomous vehicle. SUMMARY

[0003] The application proposes a map matching method based on GNSS covariance information, which uses the positioning error covariance matrix and integrity protection level information output by the GNSS receiver to realize adaptive constraint of the search range of the candidate matching point and probability optimization of the matching credibility, so as to improve the positioning accuracy and robustness in complex environments.

[0004] The application is implemented through the following technical scheme:

[0005] The application relates to a map matching method based on GNSS covariance information, which comprises the following steps:

[0006] Step 1, collect multiple types of data files output by the GNSS receiver, specifically including: ① a satellite original observation file (RINEX format) recording original observation values such as pseudorange, carrier phase and Doppler; ② a GGA message providing positioning results (longitude, latitude, height, etc.) of each epoch; and ③ a GST message providing a positioning error covariance matrix, wherein the diagonal elements respectively represent the standard deviation squares of the positioning results in the east-west, south-north and vertical directions, and the non-diagonal elements reflect the correlation between the errors in the directions.

[0007] Step 2, calculate the protection level as the boundary of the credible position, which is used to constrain the search range of the candidate matching point: based on the receiver autonomous integrity monitoring (RAIM) algorithm, perform consistency test and error detection on the positioning results of each epoch, and calculate the horizontal protection level (HPL) and the vertical protection level (VPL).

[0008] Step 3, candidate matching point set is generated with the collected positioning result as the center and the protection level as the boundary, and the emission probability (EP) of each candidate matching point in the candidate matching point set is calculated according to the collected positioning error covariance matrix to reflect the matching confidence of each candidate point under the current epoch observation condition.

[0009] Step 4, based on the spherical distance of the adjacent epoch GNSS positioning result and the path distance between the candidate points, the transfer probability (TP) of the candidate matching point in the previous epoch to the candidate matching point in the current epoch is calculated from the second epoch collected in step 1.

[0010] Step 5, the dynamic programming idea is adopted to perform forward recursion on all candidate point sequences. For each current epoch candidate point, the path probability transferred from all candidate points in the previous epoch is calculated.

[0011] Step 6, the candidate point with the maximum cumulative probability is selected at the last epoch, the optimal candidate matching point sequence of the vehicle at all epochs is reconstructed through backward tracing from the rear to the front through the forward transfer pointer, as the accurate historical trajectory of the vehicle in the topological map, and the optimal matching point of the last epoch is taken as the final positioning output of the vehicle.

[0012] Technical effects

[0013] The protection level (PL) output by the GNSS integrity monitoring is introduced into the map matching probability model together with the positioning error covariance information, the PL value calculated by the RAIM algorithm in real time is used as the search boundary through the candidate point search mechanism constrained by the protection level, the search range is dynamically associated with the GNSS integrity, the matching space is adaptively adjusted, the observation error weighted model based on the covariance matrix directly embeds the GNSS error covariance information into the candidate matching point emission probability calculation, the uncertainty of the error in different directions is fully utilized, the directionality of the observation probability is weighted, instead of the existing isotropic Gaussian assumption, the dynamic programming solving mechanism of the emission probability and the transfer probability is combined, the joint maximization criterion of the emission probability (EP) and the transfer probability (TP) is introduced in the path search stage, the global optimal path estimation in the time sequence is formed, and the local optimal problem of the existing independent matching of each epoch is avoided. Compared with the prior art, the matching accuracy and robustness are significantly improved, and the dynamic adaptive search and calculation optimization and the quantifiable positioning result of the credibility are realized. BRIEF DESCRIPTION OF DRAWINGS

[0014] Figure 1 The flowchart of the present application;

[0015] Figure 2 The schematic diagram of the embodiment system;

[0016] Figure 3 The schematic diagram of the embodiment result;

[0017] In the figure: from left to right, the original satellite navigation location, the result of the existing map matching algorithm, and the matching result of the example. Detailed Implementation

[0018] like Figure 2 As shown, this embodiment illustrates a map matching system based on GNSS covariance information, comprising: a data acquisition unit for collecting RINEX files, GGA messages, and GST messages output by a GNSS receiver; an integrity monitoring unit for calculating the positioning protection level based on the RAIM method; a map processing unit for generating a set of candidate matching points under protection level constraints; a data processing unit for calculating the transmission probability (EP) and transition probability (TP) of candidate matching points; a calculation unit for searching a path that jointly maximizes the transmission probability and transition probability based on a dynamic programming algorithm and calculating the final matching result; and an output unit for outputting the optimal path and the final matching result as the precise location of the vehicle in the topology map.

[0019] like Figure 1 The diagram shows the map matching method based on the above system in this embodiment, which includes the following steps:

[0020] Step 1: GNSS Data Acquisition. Acquire various data files output from the GNSS receiver, including: raw satellite observation files in RINEX format, GGA messages containing positioning results, and GST messages containing the positioning error covariance matrix. The RINEX files record multi-constellation pseudorange, carrier phase, and Doppler observations; the GGA messages provide positioning results for each epoch. The GST message provides the positioning error covariance matrix. Its diagonal elements represent the squared standard deviations of the positioning results in the east-west, north-south, and vertical directions, respectively, while the off-diagonal elements reflect the correlation between errors in each direction.

[0021] Step 2: Protection Level Calculation. Based on the GNSS Receiver Autonomous Integrity Monitoring (RAIM) algorithm, consistency checks and error detection are performed on the positioning results at each epoch, and the Horizontal Protection Level (HPL) and Vertical Protection Level (VPL) are calculated.

[0022] In this embodiment, the horizontal protection level is... As the first The credible location boundary of the epoch is used to constrain the search range of candidate matching points.

[0023] The autonomous integrity monitoring mentioned above is achieved using methods such as residual testing, multiple hypothesis solution separation, or advanced receiver autonomous integrity monitoring (ARAIM).

[0024] Step 3: Generate candidate matching points and calculate emission probabilities: based on epochs Location results Centered on, protection level Use the radius to retrieve all distances from an electronic map (such as a shapefile-format topology map). No more than The road segments and intersections form the candidate matching point set for the current epoch. The GNSS positioning error follows a mean of 0 and a covariance of 0. Under the premise of a two-dimensional Gaussian distribution, the emission probability ,in: This is the epoch (time) index, with a value range of... ,in: The total number of epochs of the trajectory; For the first The GNSS positioning observation results of the epoch are two-dimensional coordinate vectors. The unit is meters (m); For the first The protection level of an epoch is used to limit the search range of candidate points, and the unit is meters (m). For the calendar The set of candidate matching points, where For the first Era 1 There are candidate matching points with coordinates of . The unit is meters (m). The number of candidate matching points. Covariance matrix. For the calendar The GNSS positioning error covariance matrix is ​​as follows: A symmetric positive definite matrix, Its determinant reflects the area of ​​the error ellipse. is the inverse of the covariance matrix. This is the transpose of the coordinate difference vector between the observation point and the candidate point. Here, represents the emission probability, and represents the observation point. Belongs to candidate matching point The probability density is used to characterize the spatial consistency between GNSS observations and map candidate points.

[0025] Step 4: Calculate the transition probability: starting from the second epoch ( ), spherical distance based on GNSS positioning results of adjacent epochs Path distance between candidate matching points Calculate the difference between the two. The difference follows the parameter. transition probability under the exponential distribution assumption wherein: is the epoch index, the present step is applicable to the case of ; and are the candidate matching point coordinates of adjacent epochs and ; is the spherical distance or plane-approximation distance between the GNSS observations of epochs and , in meters (m); is the path distance calculated along the road network between the candidate matching points and , in meters (m); the difference between the two is used to reflect the consistency between the GNSS motion observations and the road constraints. The parameter is the scale parameter of the exponential distribution, which is a positive real number, in meters (m), used to control the decay rate of the distance difference on the transition probability. The transition probability is the likelihood of transitioning from the candidate matching point of the previous epoch to the candidate matching point of the current epoch, with a value range of , used to describe the dynamic continuity of the trajectory under spatial and topological constraints.

[0026] Step 5, path probability accumulation and dynamic programming search: the Viterbi idea is adopted to perform forward recursion on the candidate matching points of all epochs. For the candidate matching point of the current epoch , the cumulative probability is the simultaneous record of the optimal path joint probability from the initial epoch to this point, which produces the maximum probability of the previous epoch index, specifically: , wherein: δk(j) is the cumulative probability (i.e., the maximum value of the joint probability of the optimal path) of the candidate matching point of epoch , used for forward recursion calculation in the dynamic programming algorithm (Viterbi algorithm); is the back pointer, used to record the previous epoch candidate point index that makes the cumulative probability maximum, achieving path backtracking reconstruction; is the transition probability from the previous epoch candidate point to the current candidate point; The emission probability of the current epoch candidate point reflects the spatial matching degree between the observation point and the candidate point; "max" is a maximum value operation, and "argmax" is a previous epoch index that makes the maximum value true. All probability terms, including 、 、 , all have a value range of . Through the recursive calculation, the optimal path probability accumulation and dynamic update of the trajectory can be realized, and the real-time and computational efficiency of the algorithm under the condition of only retaining the current and previous epoch data are guaranteed.

[0027] The index is called a forward transition pointer and is used for subsequent path backtracking. Through the above iterative process, real-time dynamic optimal path estimation can be realized under the condition of only retaining the current and previous epoch data, and the online feasibility and computational efficiency of the algorithm are guaranteed.

[0028] Step 6, optimal path backtracking and positioning result output. At the last epoch , the candidate matching point with the maximum cumulative probability is selected , and the optimal candidate matching point sequence at all epochs is reconstructed by backtracking from the back to the front according to the forward pointer . The sequence is the most credible historical trajectory of the vehicle in the topological map, and the optimal matching point of the last epoch is output as the final positioning result, which is used for path planning and decision execution in application scenarios such as automatic driving and intelligent transportation.

[0029] After actual application scene experiments, the performance indicators of the present application are verified in complex urban environments including urban canyons, under elevated bridges, and tree-shaded roads. The same GNSS raw data is processed by using the existing HMM map matching algorithm and the present application. The results show that the average matching deviation of the present application is reduced by about 45% under the same observation conditions, the matching accuracy is improved by about 12% in severe multipath interference scenarios, and the average processing delay is controlled within 20 milliseconds. The experimental data fully prove the high robustness and real-time performance of the present application in complex environments, which can effectively improve the credibility and accuracy of map matching.

[0030] As Figure 3The experimental results of the embodiment in a typical urban road scene are shown. The experiment was conducted on a vehicle-mounted embedded computing platform, with hardware configuration including an ARM Cortex-A72 quad-core processor (clock frequency 2.0 GHz), 2 GB of memory, a Linux operating system, and a C++ language implementation of the core algorithm. The average single epoch processing delay of the system is less than 20 ms, meeting the real-time requirements of autonomous vehicles. The experimental vehicle is equipped with a high-precision GNSS receiver and an inertial measurement unit, and completes field testing in a typical urban road environment in Shanghai. The experimental duration is 60 minutes, the sampling frequency is 1 Hz, and the test scenarios cover open roads, urban canyons, underpasses, and shaded roads in various complex environments.

[0031] During the experiment, the system collects GNSS raw observation data (RINEX file), GGA message, and GST message in real time, executes the algorithm described in the embodiment, including RAIM integrity detection, protection level (PL) calculation, candidate matching point generation, covariance matrix-based transmission probability (EP) and transition probability (TP) calculation, and searches for the global optimal path through a dynamic programming algorithm, and finally outputs the accurate matching result of the vehicle in the topological map. To verify the effectiveness of the embodiment, the existing static map matching algorithm is used as a control group. This algorithm only matches based on the shortest distance between the GNSS observation point and the road geometry, without using GNSS error covariance information and integrity constraints.

[0032] The experiment uses four core indicators for comparison: ① Matching accuracy - the proportion of the algorithm output path consistent with the manually annotated true path; ② Calculation efficiency - the average processing time per epoch; ③ Robustness - the proportion of consecutive two-epoch positioning results with a deviation of more than 10 m; ④ Continuity - the proportion of valid positioning results output under GNSS signal obstruction. The experimental results are shown in Table 1.

[0033] Table 1

[0034] The comparison results show that the embodiment has significant advantages in complex road environments. The existing method is prone to trajectory drift and mismatch in ramp or densely populated road areas, while the embodiment combines GNSS integrity protection level (PL) and error covariance matrix information, establishes an error envelope model, and introduces a dynamic search boundary, significantly improving the stability and reliability of the matching. As shown in Figure 3 the existing algorithm's matching results (middle graph) have deviations and incorrect projections, while the embodiment (right graph) can accurately output a continuous trajectory along the centerline of the lane, accurately reflecting the vehicle's movement path.

[0035] The experimental data in Table 1 further verifies the effect: in a complex environment (urban canyon, viaduct, shaded road), after using RAIM to eliminate abnormal measurements combined with covariance information weighting, the positioning mutation probability can be reduced by about 40%, and the matching accuracy can be improved by 10-15%; by dynamically adjusting the number of candidate points through the protection level constraint, the average calculation overhead is reduced by more than 30%, the average processing time delay of the system is less than 20ms, meeting the real-time requirements of vehicle-mounted; the optimal path derived from the joint probability of EP and TP has a statistical confidence index, and the output trajectory can be directly used for scenarios that require traceable and verifiable positioning accuracy, such as trusted navigation, positioning as payment, positioning as law enforcement, etc.

[0036] In summary, the present application maintains stable output and high reliability in various complex environments, solves the technical bottlenecks of high mismatch rate, large calculation overhead, and insufficient reliability of existing static matching algorithms in dynamic environments, and realizes the output trajectory with quantifiable confidence index by fusing GNSS integrity protection level and positioning error covariance matrix information, and the overall reliability reaches 10 -5 / h order of magnitude, which is significantly better than the existing map matching algorithm based on fixed threshold and isotropic error assumption.

[0037] The above specific implementation can be adjusted in different ways by those skilled in the art without departing from the principles and purposes of the present application, the protection scope of the present application is subject to the claims and is not limited by the above specific implementation, and each implementation scheme within the scope is subject to the constraints of the present application.

Claims

1. A map matching method based on GNSS covariance information, characterized in that, The method comprises the following steps: Step 1, collecting multi-type data files output by a GNSS receiver; Step 2, calculating a protection level as a boundary of a reliable position for restricting a search range of a candidate matching point: based on a RAIM algorithm, performing consistency inspection and error detection on positioning results of each epoch, and calculating a horizontal protection level (HPL) and a vertical protection level (VPL); Step 3, generating a candidate matching point set with the collected positioning results as a center and the protection level as a boundary, and calculating a transmission probability (EP) of each candidate matching point in the candidate matching point set according to a covariance matrix of the collected positioning errors to reflect a matching confidence of each candidate point under current observation conditions; Step 4, based on a spherical distance between adjacent GNSS positioning results and a path distance between candidate points, calculating a transition probability (TP) of a candidate matching point in a previous epoch to a candidate matching point in a current epoch; Step 5, using a dynamic programming idea to perform forward recursion on all candidate point sequences; for each candidate point in a current epoch, calculating a path probability of transition from all candidate points in a previous epoch; Step 6, selecting a candidate point with a maximum cumulative probability at the last epoch, backtracking from the last epoch to the first epoch through a forward transition pointer to reconstruct an optimal candidate matching point sequence of the vehicle at all epochs as an accurate historical trajectory of the vehicle in a topological map, and outputting the optimal matching point in the last epoch as a final positioning result of the vehicle.

2. The GNSS covariance information based map matching method according to claim 1, characterized in that, The multi-type data files specifically comprise: ① a satellite original observation file containing recorded pseudoranges, carrier phases and Dopplers; ② a GGA message containing longitude, latitude and height of each epoch; and ③ a GST message containing a positioning error covariance matrix. Diagonal elements of the positioning error covariance matrix respectively represent squares of standard deviations of positioning results in east-west, north-south and vertical directions, and non-diagonal elements reflect correlations between errors in different directions.

3. The GNSS covariance information based map matching method according to claim 1, characterized in that, Step 3 specifically involves: using epochs Location results Centered on, protection level Use the radius to retrieve all distances from the electronic map. No more than The road segments and intersections form the candidate matching point set for the current epoch. The GNSS positioning error follows a mean of 0 and a covariance of 0. Under the premise of a two-dimensional Gaussian distribution, the emission probability Among them: epoch (time) index , The total epochs of the trajectory. For the first The GNSS positioning observation results of the epoch are two-dimensional coordinate vectors. , For the first The level of protection of the era, For the calendar The set of candidate matching points For the set of There are candidate matching points with coordinates of , For the calendar The number of candidate matching points, For the calendar The GNSS positioning error covariance matrix, Let be the determinant of the covariance matrix. It is the inverse of the covariance matrix. This is the transpose of the coordinate difference vector between the observation point and the candidate point. To reflect GNSS observation points With map candidate points The emission probability of spatial consistency. .

4. The GNSS covariance information based map matching method according to claim 3, characterized in that, The step 4 is specifically: calculating the spherical distance between the adjacent GNSS positioning results based on the second epoch The path distance between the candidate matching points , and calculating the difference between the two ; When the difference obeys an exponential distribution with a parameter , the transition probability , where is the epoch (time) index, and are the candidate matching points of adjacent epochs and , respectively, is the spherical distance or the planar approximate distance in a local range between the GNSS positioning results of epochs and , is the shortest path distance on the road network topology from the candidate point to the candidate point along the feasible road path, is the absolute difference between the path distance and the GNSS distance, is the scale parameter of the exponential distribution, is the transition probability from the candidate point of the previous epoch to the candidate point of the current epoch, is the natural exponential function.

5. The GNSS covariance information based map matching method according to claim 4, characterized in that, Step 5, for the current epoch Candidate matching points Its cumulative probability To determine the joint probability of the optimal path from the initial epoch to this point, we simultaneously record the index of the previous epoch that generates the highest probability, i.e., all candidate indices in the previous epoch. The operation of taking the maximum value is as follows: , Where: when using Viterbi dynamic programming for forward recursion, For the current epoch index, For the current epoch Candidate matching point index, For the previous epoch Candidate matching point index, For the calendar Time Candidate Point The cumulative probability, i.e., the maximum value of the probability of the optimal joint path, As a forward transfer pointer, the record makes Index of candidate points in the previous epoch to obtain the maximum value From the epoch candidate points Transfer to epoch candidate points The transition probability, For the calendar The emission probability, i.e. , For all candidate indices in the previous epoch The operation of taking the maximum value from above. To find the index that makes the maximum value true. ; Recurrence equation With For online updating the optimal path probability and the corresponding predecessor index, the subsequent path backtracking reconstruction is realized.

6. The GNSS covariance information based map matching method according to claim 5, characterized in that, The step 6 is specifically: at the last epoch , a candidate matching point with the maximum cumulative probability is selected , and the forward pointer is traced back from the rear to the front, so as to reconstruct the optimal candidate matching point sequence at all epochs , that is, the most credible historical trajectory of the vehicle in the topological map, and the optimal matching point of the last epoch is output as the final positioning result, which is used for path planning and decision execution in application scenarios such as automatic driving and intelligent transportation.

7. A GNSS covariance information based map matching system implementing the method of any of claims 1-6, characterized by The method comprises the following steps: A data collection unit for collecting RINEX files, GGA messages and GST messages output by a GNSS receiver; An integrity monitoring unit for calculating a positioning protection level based on a RAIM method; a map processing unit for generating a candidate matching point set under the protection level; a data processing unit for calculating transmission probabilities and transition probabilities of candidate matching points; A calculation unit for searching a joint maximum probability path of the transmission probabilities and the transition probabilities based on a dynamic programming algorithm and calculating a final matching result, and an output unit for outputting the optimal path and the final matching result as an accurate position of the vehicle in a topological map.

Images