An improved sar-based complex multipath underground vehicle automatic positioning method

Abstract

The application discloses a kind of complex multipath underground vehicle automatic positioning method based on improved SAR, it is related to radio frequency identification target technical field, through the relative motion between label and antenna, the offset of the physical center and phase center of antenna is analyzed, and the phase center of antenna is calibrated.Then system executes, first, the multipath detection is carried out to phase measurement value, and the influence of multipath effect on phase is eliminated;Then, the lowest point of unwrapping phase is normalized, the phase cycle ambiguity problem is solved, and the fixed offset is eliminated;Finally, a target function is constructed to realize positioning.Compared with grid matching method, the present application directly optimizes the target function, rather than matching the priori phase sequence of each grid point point by point.Meanwhile, the factor of environmental noise is also considered, which further improves the positioning accuracy, so as to realize high-precision positioning of the target vehicle in the case of no GPS and complex multipath environment in underground garage.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of radio frequency target identification, and in particular to a complex multi-path underground vehicle automatic positioning method based on improved SAR. BACKGROUND

[0002] With the rapid development of industrial Internet of Things, positioning algorithms in the industrial field need to meet the following core requirements: first, high precision, which must provide centimeter-level or even millimeter-level positioning accuracy to support fine tasks such as small object grabbing; second, adaptability to complex environments, the algorithm should have strong anti-interference ability to cope with challenges such as noise and multipath effect in real environments; and finally, high efficiency, since the algorithm will be run in real time on edge devices with limited computing resources, the positioning model should be as simple and efficient as possible to ensure fast response.

[0003] RFID (Radio Frequency Identification) technology has been widely used in item positioning and tracking due to its high resolution and low cost. In terms of intelligent applications, RFID technology has been successfully applied to underground garage intelligent car searching systems. By installing RFID tags on parking spaces, the specific location of vehicles can be obtained in real time, and the vehicle of the car owner can be quickly located through the system. This system not only improves the efficiency of finding vehicles, but also effectively reduces the time waste of car owners in the parking lot, optimizes garage management, and improves the utilization rate of parking resources. In addition, the low cost and high efficiency of RFID technology make it widely used and promoted in such applications.

[0004] In recent years, phase-based RFID positioning technology has gradually attracted widespread attention due to its high resolution. Traditional RFID positioning has two major fatal flaws. First, phase measurement is usually affected by unknown physical biases, including inconsistencies in the center of the antenna and additional displacement. Existing research usually regards the physical center of the antenna as the phase center, i.e., the point of signal transmission and reception. However, due to the inherent characteristics of hardware, the points of transmitting and receiving signals often do not completely coincide, resulting in unavoidable phase offset, which makes it difficult to achieve high-precision target positioning. Therefore, it is crucial to determine the offset between the phase center of the antenna and its physical center and calibrate the phase center of the antenna based on this. Second, in actual environments, there are inevitably noises, such as fixed obstacles, pedestrians, or other moving objects, which may cause more significant phase offset and affect positioning accuracy. Especially in indoor environments with complex noise, the dynamic and unpredictable nature of interference sources poses greater challenges to the stability and accuracy of the positioning system. Therefore, the existence of noise must be fully considered to ensure high-precision positioning performance under various environmental conditions.

[0005] SAR (Synthetic Aperture Radar) was originally applied in the military field for target positioning and geographic imaging. Due to its high-precision positioning capability in specific scenarios, SAR-based RFID positioning technology has developed rapidly in recent years. This technology generates a series of virtual antenna / tag arrays by utilizing the relative motion between the tag and the reader antenna, thereby improving the sampling rate of the positioning system and breaking through the size limit of the antenna or tag. When the antenna moves while the tag remains stationary, it is commonly referred to as SAR; if the tag also participates in the motion, it is referred to as ISAR (Inverse Synthetic Aperture Radar). Nevertheless, SAR and ISAR are both classified as part of the SAR technology. Existing SAR research mainly focuses on its combination with holographic imaging technology. The hologram method generally includes two steps: first, calculate the phase sequence of each grid point through prior information such as motion direction and speed; second, locate the tag by comparing the measured phase sequence with the calculated phase sequence. However, as the number of tags in the environment increases, the complexity of the system and the demand for computing power also significantly increase.

[0006] However, the current radio frequency identification positioning target tag field still has the following limitations: (1) In the case of rich multipath environment, the positioning is high, and the environment has a greater impact on radio frequency identification; (2) The circuit of the reader, antenna, and tag hardware devices affects, such as the phase center and the physical center of the antenna are not consistent. (3) RFID reader phase reading problem, due to the hardware limitations of the measurement device, the measured phase value is between 0 and 2π, therefore, the accurate distance cannot be directly obtained from the phase information, and there is a phase period ambiguity problem. SUMMARY

[0007] In order to solve the above technical problems, the present application provides a complex multipath underground vehicle automatic positioning method based on improved SAR, comprising the following steps:

[0008] S1, using the relative motion of the tag and the antenna, analyzing and calibrating the offset between the antenna phase center and the physical center, and eliminating the fixed phase deviation caused by the hardware;

[0009] S2, multi-path detection of phase measurement value: since the phase difference of adjacent frequency bands is the same, calculate the variance of adjacent channel phase difference, and judge the measurement point with phase difference variance greater than the empirical threshold as being strongly interfered by multi-path, and eliminate these measurement points; weak multi-path repair is performed on the retained data: define a weight function according to the dispersion of each frequency band phase, and perform weighted linear fitting to make the phase close to an ideal straight line;

[0010] S3. Unwrap the wrapped phase, remove the 2π jump to obtain a continuous phase; normalize with the lowest point to eliminate the fixed phase offset, compress the uncertainty of the initial phase value into a very small distance error, and form a residual motion / residual distance model and sequence accordingly.

[0011] S4. Construct a cost / objective function from the residual distance sequence and perform nonlinear optimization; use Newton's method iteratively to obtain the tag's two-dimensional position. If the antenna trajectory is not a straight line, take the trajectory slope at the lowest point to rotate the coordinate system, solve in the rotated coordinate system first, and then use the inverse rotation matrix to restore it to the original coordinate system.

[0012] The beneficial effects of this invention are:

[0013] (1) In this invention, an adaptive multipath processing method based on multi-band phase consistency test is proposed to address the problem of data instability caused by multipath interference in the positioning system under complex environments. Traditional algorithms usually discard the phase anomalies directly. Although this is simple to implement, it often leads to a decrease in positioning accuracy due to the reduction of effective observations. To solve this problem, this invention utilizes the physical characteristic that the phase of multi-band should change linearly under ideal conditions. It judges the strength of multipath interference by calculating the variance of the phase difference between adjacent frequency bands, thereby achieving accurate identification and elimination of strong multipath signals. For weak multipath interference that is difficult to completely avoid, this invention further introduces a weighted least squares phase fitting method. It adaptively allocates weights according to the dispersion of phase readings in different frequency bands, so that slightly contaminated observation data can still be used to reconstruct a phase change curve that is closer to the truth after reasonable correction. With the help of this processing flow, the data that was originally considered invalid can play a role again. The stability of the system in complex electromagnetic environments is significantly improved, laying a solid foundation for achieving centimeter-level high-precision positioning.

[0014] (2) In this invention, by constructing an optimization framework consisting of fixed phase offset elimination and residual motion model, the long-standing phase period ambiguity problem in radio frequency identification ranging is effectively solved. Due to the periodic superposition characteristics of the phase itself, coupled with the unknown initial phase of the device, traditional ranging often produces multiple seemingly reasonable solutions, which leads to jumps in positioning results and makes it difficult to guarantee stability. To this end, this invention designs a continuous signal processing flow: first, the wrapped phase is unwrapped to restore its true shape that changes continuously with distance; then, by identifying the lowest point of the phase curve and normalizing it, the fixed phase offset that originally affected the whole is transformed into a local distance error with a small amplitude; based on this processing result, the vertical distance from the tag to the antenna path and this error are then jointly embedded into the residual motion model for joint solution. This modeling method greatly shrinks the solution space mathematically, enabling the optimization process to converge quickly and stably to the unique correct global solution, fundamentally avoiding positioning jumps caused by periodic ambiguity. With the help of this innovative idea, the system can improve the positioning accuracy from decimeter level to centimeter level without additional hardware modification, achieving a performance leap entirely dependent on algorithm breakthroughs.

[0015] (3) In this invention, by designing a dynamic rotation method for coordinate systems oriented towards non-ideal motion trajectories, the dependence of high-precision positioning on straight-line trajectories is overcome. Traditional synthetic aperture radar positioning technology strictly requires the reader to move along an ideal straight line, which is extremely difficult to achieve in actual complex scenarios and seriously restricts the application of the technology. This invention innovatively determines the optimal coordinate system rotation angle by intelligently detecting the lowest point of the unwrapped phase curve and simultaneously obtaining the instantaneous tangent direction of the antenna motion trajectory at that moment. Then, the entire system is mapped to a new coordinate system through a rotation matrix. Under this coordinate system, the antenna motion is equivalently straightened, thereby transforming the complex non-linear trajectory problem into a standard straight-line positioning problem. After the solution is completed, the result is mapped back to the original coordinate system through an inverse transformation. This design, in principle, endows the positioning system with the ability to cope with any complex motion path, significantly improving the adaptability and flexibility of the technology in real unstructured environments, and clearing the key technical obstacles for high-precision positioning to move from the laboratory to industrial applications. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of a specific scenario in an embodiment of the present invention;

[0017] Figure 2 This is a schematic diagram of the overall process of the present invention;

[0018] Figure 3 This is a schematic diagram of the deviation test experiment in an embodiment of the present invention;

[0019] Figure 4 This is a diagram showing the phase measurement results during tag movement in an embodiment of the present invention;

[0020] Figure 5 This is a phase distribution diagram of the multi-band received signal in an embodiment of the present invention;

[0021] Figure 6 This is a schematic diagram of phase repair in an embodiment of the present invention;

[0022] Figure 7 This is a schematic diagram of phase unwrapping in an embodiment of the present invention;

[0023] Figure 8 This is a schematic diagram of the reader motion model in an embodiment of the present invention;

[0024] Figure 9 This is a schematic diagram of coordinate system transformation in an embodiment of the present invention;

[0025] Figure 10 This is a schematic diagram comparing the positioning results of different methods in an embodiment of the present invention;

[0026] Figure 11 This is a schematic diagram illustrating the relationship between positioning error and the number of frequencies in an embodiment of the present invention;

[0027] Figure 12 This is a schematic diagram illustrating the impact of noise variance on positioning error in an embodiment of the present invention;

[0028] Figure 13 This is a schematic diagram of the antenna's motion trajectory in an embodiment of the present invention;

[0029] Figure 14 This is a CDF diagram showing the positioning error in an embodiment of the present invention;

[0030] Figure 15 This is a schematic diagram illustrating the influence of relative height on positioning error in an embodiment of the present invention;

[0031] Figure 16 This is a graph showing the relationship between positioning error and sampling interval in an embodiment of the present invention. Detailed Implementation

[0032] This embodiment provides an improved SAR-based localization method (ISLM) for complex multipath underground vehicles. Considering the impact of antenna phase center calibration and multipath interference on the positioning results, this embodiment's method is used to solve the problem of high-precision vehicle positioning in enclosed indoor spaces. Specific scenarios include... Figure 1 As shown, the specific method is as follows: Figure 2 As shown, it includes the following steps:

[0033] S1. By utilizing the relative motion between the tag and the antenna, the offset between the antenna phase center and the physical center is analyzed and calibrated to eliminate the fixed phase deviation caused by the hardware.

[0034] The phase center of an antenna often deviates from its physical center. To verify the existence of this deviation, the following was first performed: Figure 3 The experiment shown assumes the origin is located at the physical center of the antenna. Tags spaced 1m apart are placed in front of the antenna, and the tags are moved horizontally and vertically (i.e., along the x-axis and y-axis), collecting phase data every 2.5cm. Figure 4 The measured data shown indicates that, regardless of whether the label is moved horizontally or vertically, the minimum distance measured is not at the origin position, with a deviation of about 2-5 cm from the origin position.

[0035] Deviations in the estimation of the antenna phase center position can significantly affect the accuracy of data processing. Therefore, to improve measurement accuracy, the antenna phase center must be precisely calibrated. Specifically, the lowest point of the phase curves obtained from the above experiments in different directions is taken as the phase center point in that direction. The distance between the actual measured observation point with the lowest phase and the antenna's physical center is used to compensate for the antenna's physical center, reducing the impact of phase center deviation on subsequent data processing.

[0036] S2. Perform multipath detection on phase measurements: Since the phase difference between adjacent frequency bands is the same, calculate the variance of the phase difference between adjacent channels. The measurement points whose phase difference variance is greater than the empirical threshold are judged to be subject to strong multipath interference and these measurement points are removed. Perform weak multipath repair on the retained data: Define a weight function based on the dispersion of the phase of each frequency band and perform weighted linear fitting to make the phase close to the ideal straight line.

[0037] The COST RFID device provides K (K=16) frequency bands located between 920.625 and 924.375 MHz. The gap between any two adjacent frequency bands is the same, 0.25 MHz. Therefore, theoretically, the measured phase difference between two adjacent frequency bands is... Same, that is , , and Let represent the phase measured in the m-th and m-1-th frequency bands, respectively. This represents the phase vector measured by the antenna at a certain measurement point from different channels. This represents the phase measurement value of the k-th frequency band. , This represents the clean, interference-free phase corresponding to that frequency band. Represented as:

[0038] (1);

[0039] in, This represents the phase deviation in the k-th frequency band caused by multipath propagation. This indicates the clean phase corresponding to the first frequency band; for different frequency bands... The values ​​differ because the frequency and propagation path variations in different frequency bands can lead to varying degrees of multipath effects. The following experiment was conducted, assuming a fixed distance between the tag and antenna, performing phase measurements in both open areas and multipath environments, and recording the phase data for each frequency band. .

[0040] In open regions, due to the limited number of reflection paths, the multipath effect on the signal is negligible, and the detected phase does indeed follow the pattern shown. Figure 5 The points in the circle shape represent a linear relationship. At this point, the phase measurements across different frequency bands exhibit a relatively consistent linear relationship. For the k-th frequency band,

[0041] (2);

[0042] Here, y is called an ideal straight line because it is a straight line only when there is no multipath effect. On the other hand, in environments with severe multipath effects, multipath variables... This cannot be ignored. Therefore, from frequency band 1 to K, the phase information fluctuates sharply. These fluctuations reflect the impact of multipath effects on the signal, and the magnitude of the fluctuations is closely related to the multipath variables between channels. For example... Figure 5 The points indicated by the triangle shape represent the phase fluctuations received by the K channels under multipath conditions, and the amplitude of these fluctuations is relatively large, proving that the multipath effect interferes with the phase measurement values ​​in this environment.

[0043] This feature can be used for multipath detection. Let the phase difference between adjacent channels be... ,in , To perform multipath detection, the variance of these phase differences can be calculated. When multipath interference is small, Approaching 0; if the variance is greater than the set empirical threshold. (This value is usually determined experimentally), that is This indicates the presence of strong multipath interference. Therefore, it is believed that the phase data of this monitoring point is affected by multipath interference and cannot be used for accurate positioning. In this case, the phase measurement value should be discarded to avoid affecting the subsequent positioning accuracy.

[0044] To achieve the goal of repairing weak multipath phase signals, all measured phase values... Fit a straight line, define the fitted line as Equation (2), and define... as follows:

[0045] (3);

[0046] in, This represents the fitted value corresponding to frequency band k. This represents the weight corresponding to frequency band k. Intuitively, the impact of multipath effects on each frequency band is not the same. For some frequency bands, they may be disrupted by unpredictable movement factors, or they may simply be more sensitive than other channels. Therefore, it is necessary to evaluate and take into account the data quality of each frequency band. To achieve this goal, a weighting function is defined to reduce the severe impact of outliers. The principle of determining the weighting function is simple: the more severe the dynamic multipath effect, the more discrete the phase report. The data quality can be estimated by taking into account the dispersion of the phase information of the received signal. Intuitively, the more discrete the phase of the received signal, the greater the impact of uncontrollable dynamic multipath effects.

[0047] The sample mean difference of the phase received in frequency band k To represent the dispersion:

[0048] (4);

[0049] in, This represents the phase samples reported in frequency band k. This represents the average of all m samples; it is considered true when at least 80% of the phase samples within the observation window have a mean absolute deviation below the upper limit of the static baseline. The expected value should be 0.

[0050] definition for:

[0051] (5);

[0052] The weighting function is further defined as follows:

[0053] (6);

[0054] in, The weighting function will help find a more suitable fitting line, ultimately leading to accurate results.

[0055] To determine the slope of the fitted line And the intercept d value, need to minimize the value in equation (3) To achieve this goal, the variables are calculated separately. and d Partial derivatives:

[0056] (7);

[0057] make Solving for the problem, we get:

[0058] (8);

[0059] in,

[0060] ;

[0061] ;

[0062] From the above formula, we can see that the slope of the fitted line y is... There are errors e1 and e2 between the intercept d and the ideal straight line. If the values ​​of these errors can be determined, the expected value of the phase can be estimated. Based on the expected phase value The straight line obtained by performing phase repair is as follows: Figure 6 As shown in the fitted straight line, the repaired phase value exhibits a linear relationship and roughly matches the theoretical phase.

[0063] S3. Unwrap the wrapped phase, remove the 2π jump to obtain a continuous phase; normalize with the lowest point to eliminate the fixed phase offset, compress the uncertainty of the initial phase value into a very small distance error, and form a residual motion / residual distance model and sequence accordingly.

[0064] For ease of analysis, the phase data of the 920.625MHz band after removing multipath propagation are used below. This will be illustrated using an example. Due to hardware limitations of the measuring equipment, the measured phase values... Therefore, precise distance cannot be directly obtained from phase information. However, phase changes can still be used for tag localization. To address this issue, the phase change can be viewed as a phase curve, such as... Figure 7 As shown by the solid line in the middle.

[0065] The first step in data processing is to unwrap the wrapped phase. The purpose of unwrapping is to remove the ambiguity caused by periodicity, thereby recovering the true phase change curve. The specific method for phase unwrapping can be represented as follows:

[0066] (9);

[0067] in, and These represent the phase data before and after untangling, respectively. , [·] represents the phase after untangling; [·] represents the integer value function; sign(·) represents the sign function.

[0068] Equation (17) below shows that after a phase jump is detected, the subsequent phases are added to or subtracted by 2π; finally, after the phase unwrapping operation, the signal phase no longer has a jump and becomes a continuous phase sequence. Figure 7 The unwrapped phase, shown by the dashed line, can accurately calculate the distance between the reader antenna and the tag, thereby determining the tag's location information.

[0069] Although there is a linear relationship between the phase measurement and the distance, the initial value of the phase remains unknown. Therefore, the unwrapping process still involves an uncertain number of cycles, i.e., the number of periods. Consequently, after unwrapping, the phase still needs to be calibrated to account for errors caused by the uncertainty of the initial phase value.

[0070] phase relative distance The conversion between them can be expressed as:

[0071] (10);

[0072] in, Indicates wavelength. This represents the minimum untangling phase; the relative distance model is updated according to the following formula:

[0073] (11);

[0074] in, This represents the distance corresponding to the initial phase.

[0075] To reduce the influence of the initial phase, the method proposed in this embodiment performs an operation called fixed phase offset cancellation. The relative distance between the tag and the antenna... Includes a fixed vertical distance r and a residual distance Obviously, when the label is located at a vertical point... hour, The phase shift caused by r can be eliminated by the following formula:

[0076] (12);

[0077] in, Indicates the residual phase. Indicates unwrapped phase The minimum value is obtained. Now, the phase curve with the lowest point of 0 is obtained. The phase period ambiguity information is removed, and the shape of the phase curve is maintained. However, there are certain limitations to the interrogation frequency of RFID, making it difficult to guarantee that the tag can be interrogated when the tag and antenna are closest. On the other hand, phase distortion may also change the position of the minimum value. Therefore, although this step cannot completely eliminate the influence of the initial phase, it can help control the unknown fixed phase shift to a very small range.

[0078] Since there is a linear relationship between phase and distance, the phase error is converted into a distance error. Based on this, the motion model shown in equation (11) can be modified into the following residual motion model:

[0079] (13);

[0080] in, This indicates the distance error caused by the elimination of fixed phase offset.

[0081] S4. Construct a cost / objective function from the residual distance sequence and perform nonlinear optimization; use Newton's method iteratively to obtain the tag's two-dimensional position. If the antenna trajectory is not a straight line, take the trajectory slope at the lowest point to rotate the coordinate system, solve in the rotated coordinate system first, and then use the inverse rotation matrix to restore it to the original coordinate system.

[0082] like Figure 8 As shown, in a SAR scenario, the tag's position is fixed and unknown, while the reader antenna's velocity is known and moves in a fixed direction. Let the antenna's initial position (i.e., the position to be located) be... x0 and y0 represent the initial horizontal and vertical coordinates of the antenna; the reader performs the interrogation process at N different times, and let the antenna position at the nth interrogation be y0. x n and y n This represents the horizontal and vertical coordinates of the antenna at the nth monitoring point; and when n=1, The position of the antenna can be obtained by a position sensor.

[0083] When the nth query is executed, the relative distance between the tag and the antenna is:

[0084] (14);

[0085] in, Indicates the position of the label, x t and y t Represents the x and y coordinates of the label; in Figure 8 In the middle, set The x-coordinate represents the position coordinates when the antenna is closest to the tag. min and y minLet x and y represent the horizontal and vertical coordinates when the antenna is closest to the tag, and let r represent the distance between the tag and the antenna's trajectory.

[0086] (15);

[0087] residual distance This represents the actual distance between the antenna position and the tag position when the nth query is executed. The difference between r and r, and

[0088] (16).

[0089] The motion model is as follows:

[0090] (17);

[0091] in, Let n represent the distance of the antenna from the starting point to the nth point. Then we have:

[0092] (18);

[0093] in, Substituting into equation (18), we get:

[0094] (19);

[0095] get:

[0096] (20).

[0097] Through the above data processing, the phase sequence is obtained. and time series , Then the residual distance sequence is obtained. :

[0098] (twenty one);

[0099] Construct the cost function:

[0100] (twenty two);

[0101] in Represented as:

[0102] (twenty three);

[0103] Then, the problem shown in equation (23) is solved by nonlinear optimization to obtain the label's location information:

[0104] (twenty four).

[0105] This paper uses Newton's method to solve the above optimization problem. However, it should be noted that the fixed phase offset elimination process provides a relatively accurate iteration range. and .For example, If the lowest phase value in the unwrapped phase sequence is used, the x-coordinate of the tag to be located can be inferred. ,in express The iteration range. As for... This represents an error, which will be initialized to 0 and iterated over a very small range. Therefore, the optimization process mainly involves adjusting the parameter y. t Perform iterations.

[0106] In some environments, such as corridors and libraries, movement is mostly required along straight lines. However, in other situations, robots may need to avoid obstacles along non-linear paths, such as navigation in warehouses or underground parking garages. Non-linear paths involve complex geometry, which necessitates modifications to the system to enable the solution process as described above.

[0107] In non-linear trajectories, calculating the phase minimum becomes infeasible because the direction of motion no longer remains stable. To address this issue, we assume that the timestamp of the minimum obtained through phase unwrapping is... The trajectory of the antenna movement is in The slope at time is Then perform coordinate system transformation; such as Figure 9 As shown, this embodiment uses coordinate rotation to align the trajectory to a new coordinate system so that the aforementioned method can be applied to solve the problem: first, the rotation angle is determined by the slope of the fitted straight line. Let the center of rotation be the origin and counterclockwise be positive. Then, the position vector in the original coordinate system is calculated according to the formula. Transformed into a rotating coordinate system In this coordinate system, Substituting the residual sequence from equation (22) into the positioning algorithm yields the result, and finally... The result is transformed back to the original coordinate system using the inverse rotation matrix, thus completing the entire calculation process from the original coordinate system to the rotated coordinate system and back to the original coordinate system.

[0108] First, the position of the antenna Rotate, rotation angle Depend on By applying this rotation matrix, the antenna's trajectory is transformed into a new coordinate system. as follows:

[0109] (25);

[0110] The rotated antenna coordinates are ,Will and the residual distance sequence shown in equation (22) Substituting this into the localization algorithm, the coordinates of the target label are calculated as follows: (That is, the position of the label in the rotated coordinate system).

[0111] Next, we use the inverse rotation matrix to transform the rotated label coordinates back to the original coordinate system. The inverse rotation matrix is ​​the transpose of the rotation matrix, and is represented as:

[0112] (26);

[0113] The position of the label in the original coordinate system is calculated. To obtain, that is The above operations transform the final position of the label from the rotated coordinate system back to the original coordinate system, thus completing the entire positioning calculation.

[0114] To verify the performance of the algorithm proposed in this embodiment, the performance of the ISLM algorithm and the MP-R (Multipath Removal) method, which removes strong multipaths without repairing weak multipaths, was simulated and measured based on actual data obtained from MATLAB simulation software and hardware devices.

[0115] To verify the effectiveness of the algorithm in this embodiment, 20 reference tags were randomly generated in a rectangular simulation area of ​​x∈[0,200]cm and y∈[100,200]cm. It is assumed that the antenna moves at a constant speed along a straight trajectory from the starting point (0cm,0cm) to (200cm,0cm), and is sampled every 0.1s during the movement. The following is a detailed simulation evaluation.

[0116] First, simulation comparisons and analyses are performed on the ISLM method, MP-R method, and Moloc method proposed in this embodiment. The experimental results are presented using CDF (Cumulative Distribution Function) curves, as shown below. Figure 10As shown in the figure, compared to the Moloc and MP-R methods, the method in this embodiment significantly improves positioning accuracy, with a positioning error below 8 cm occurring in over 95% of cases, while the probabilities of positioning errors below 8 cm for the Moloc and MP-R methods are 40% and 75%, respectively. This embodiment demonstrates a clear advantage in multipath environments, primarily due to its multipath processing, enabling the system to maintain high-precision positioning in complex environments. In contrast, the Moloc method suffers from lower positioning accuracy due to the lack of multipath optimization, while the MP-R method, although capable of removing some multipath signals, fails to effectively handle weak multipath signals, leaving room for further improvement in positioning accuracy. Therefore, a comprehensive multipath signal processing strategy is crucial for improving positioning accuracy.

[0117] Table 1 shows the x-axis, y-axis, and average 2D positioning errors for the three different positioning methods. As can be seen from Table 1, the ISLM method proposed in this embodiment outperforms the other two methods in terms of both x-axis, y-axis, and 2D positioning error. For the Moloc and MP-R methods, the average 2D positioning errors are 10.12 cm and 5.8890 cm, respectively, while the average positioning error of the ISLM method proposed in this embodiment is only 4.64 cm, demonstrating a significant improvement in positioning accuracy.

[0118] Table 1. Average positioning error (cm) for different methods

[0119]

[0120] Next, phase detection was performed using different numbers of channel frequencies to evaluate the impact of the number of channels on positioning accuracy. The experimental results are as follows: Figure 11 As shown in the figure, the positioning accuracy improves with the increase in the number of channels. When using only one channel, the average positioning error is 6.98 cm; when using four channels, the average positioning error decreases to 6.24 cm; and when eight channels are selected, the tag positioning error reaches 6.03 cm. However, although 16 channels provide the smallest positioning error, the computational load also increases significantly. In contrast, using eight channels achieves near-optimal positioning accuracy while requiring relatively less computation. Therefore, to strike a trade-off between positioning accuracy and computational efficiency, eight channels were selected for all subsequent experiments.

[0121] Different noise variances can also affect positioning accuracy, such as Figure 12 As shown, four cases with noise variances of 0.25, 0.5, 1, and 1.5 were considered, and the impact of the noise variance magnitude on the average positioning error was evaluated. Figure 13It can be seen that the average positioning error increases with the increase of phase noise variance. Meanwhile, compared to the original method, the positioning accuracy of the method in this embodiment shows a significant improvement. Specifically, when the noise variance is 1, the average positioning errors for the Moloc, MP-R, and ISLM methods are 7.09 cm, 4.90 cm, and 4.51 cm, respectively. However, when the noise variance increases to 1.5, the method in this embodiment shows a more significant advantage, with an average positioning error of 5.51 cm, while the average positioning error for the Moloc method is 11.49 cm. Compared to the Moloc method, the positioning error of the method in this embodiment is reduced by 52.04%.

[0122] like Figure 13 The scenario shown depicts 20 tags randomly distributed within a designated area. To verify the impact of the reader's motion trajectory on the tag positioning error, this embodiment considers... Figure 13 The three different motion trajectories are shown. When the reader antenna moves along these three trajectories respectively, the randomized tags are located, and the average positioning error of the tags when the reader moves along these three different motion trajectories is obtained. The results are shown in Table 2.

[0123] Table 2. Tag positioning error under different motion trajectories (cm)

[0124]

[0125] As shown in Table 2, the simulation results indicate that the tag positioning error is lowest when the reader antenna moves along trajectory two, followed by trajectory three, with trajectory one corresponding to the largest average positioning error. When the reader moves along these three trajectories respectively, the average tag positioning errors are 7.13 cm, 5.36 cm, and 6.14 cm. This is because, among all three trajectories, trajectory two is the closest to the tag to be positioned, thus resulting in the lowest positioning error.

[0126] To verify the performance of the algorithm proposed in this embodiment in a real-world scenario, a practical test was conducted. The hardware and software parameter settings used in the test are as follows:

[0127] Hardware: The reader used in the test was an Impinj Speedway R420 with a transmit power of 32.5dB. The antenna was a HAUHF2599 (902-928MHz), a circularly polarized directional antenna with a gain of 9dBic and dimensions of 260mm × 260mm × 20mm, fixed on a mobile bracket. The tags used in the test were Avery-Dennison passive UHF-RFID Dogbone inlay tags using the Impinj Monza R6 chip, conforming to the ISO 18000-6C EPC Class1 Gen2 standard protocol. In the test, 10 tags were affixed to the vehicle's windshield. The host computer was a laptop equipped with an Intel i7-12700H CPU and 16GB RAM, communicating with the reader via Ethernet cable. In this embodiment, the actual position of the antenna was obtained by measuring with a laser rangefinder.

[0128] Software: The RFID reader is controlled by the Java-based Impinj Octane SDK. Test data is processed using PyCharm, and finally, MATLAB is used to calculate the position.

[0129] The actual test was conducted in the underground parking garage of Area A of Linjiang Building at Nanjing University of Information Science and Technology. In the experimental scenario, 10 SL11045 tags were affixed to the vehicle. The antenna was always kept facing the vehicle. The robot car moved in a straight line. The speed of the car could be adjusted. In the experiment, it was set to move at a constant speed of 2 cm / s.

[0130] like Figure 14 As shown, the CDF plot of the positioning error in the actual test scenario is presented. This plot demonstrates that, in the underground parking garage scenario where multipath effects are significant, the method in this embodiment exhibits significantly better positioning performance than the Moloc method. After removing multipath interference, the probability of a positioning error less than 10 cm increases from 0% to 50%. This result indicates that eliminating multipath effects plays a crucial role in improving positioning accuracy. Subsequently, by repairing the multipath problem, the probability of a positioning error less than 10 cm reaches 80%, further improving accuracy. Overall, removing and repairing multipath effects significantly improves positioning accuracy, and gradual optimization makes the positioning system more precise and reliable.

[0131] In reality, different car models can make it difficult to ensure that the tag is at the same height as the antenna. To address this, this study investigated the impact of the relative height between the reader antenna and the tag on tag positioning error. This embodiment, keeping all other factors constant, measures the tag detection error at different relative heights from 0 to 50 cm in 10 cm increments. Figure 15The results show that the positioning error increases with increasing relative height, with the minimum error occurring at a height where the relative error is 0 cm. Overall, the method in this embodiment can achieve relatively accurate positioning even in scenarios with relative height errors.

[0132] In practical applications, the sampling interval of the device is constrained by hardware costs, and factors such as tag movement speed, number of antennas, number of tags in the readable area, and the positioning environment also have a certain impact. Therefore, the influence of the sampling interval on the positioning error was further verified, and the results are as follows: Figure 16 As shown in the figure, the sampling time intervals are 1s, 2s, 3s, 4s, and 5s. The figure shows that the positioning error increases with the increase of the sampling interval. When the sampling interval is 1s, the average positioning error is 6.96cm. As the sampling interval gradually increases, the average positioning errors are 8.93cm, 10.53cm, 12.23cm, and 14.55cm, respectively.

[0133] For the tags affixed to the target vehicle in the experimental scenario, the reader antenna was moved along three different trajectories shown in step S4 in the area directly in front of the vehicle. Table 3 shows the average positioning error results of the tags under different trajectories. The measured data shows that when the antenna moves along trajectory one, trajectory two, and trajectory three respectively, the tag can achieve ideal positioning accuracy, with two-dimensional positioning errors all below 10cm, specifically 9.58cm, 8.36cm, and 8.49cm. Therefore, regardless of the trajectory along which the reader antenna moves, the method of this embodiment can achieve high tag positioning accuracy.

[0134] Table 3. Variation of positioning error with motion trajectory (cm)

[0135]

[0136] Before executing the positioning process, the method in this embodiment analyzes the offset between the physical center and phase center of the antenna by measuring the relative motion between the tag and the antenna, and calibrates the phase center of the antenna. The positioning process of the system includes the following three steps: First, multipath detection is performed on the phase measurement values ​​to eliminate the influence of multipath effects on the phase; then, the lowest point of the unwrapped phase is normalized to solve the phase period ambiguity problem and eliminate fixed offset; finally, an objective function is constructed to achieve positioning.

[0137] Compared to grid matching methods, the proposed method in this embodiment directly optimizes the objective function, rather than matching the prior phase sequence of each grid point point by point. Furthermore, it considers environmental noise, thereby further improving positioning accuracy. This is crucial for scenarios with complex environments requiring high-precision positioning, enabling high-precision positioning of target vehicles in complex multipath environments such as underground parking garages without GPS. Simulation experiments and field test results show that in the complex multipath environment of underground parking garages, when the reader antenna moves along a linear trajectory in front of the vehicle, high-precision positioning can be achieved without relying on external sensors, with positioning accuracies of 6.14 cm and 8.49 cm, respectively.

[0138] In addition to the embodiments described above, the present invention may have other implementations. All technical solutions formed by equivalent substitution or equivalent transformation fall within the protection scope claimed by the present invention.

Claims

1. An automatic positioning method for complex multi-path underground vehicles based on improved SAR, characterized in that: Includes the following steps: S1. Utilize the relative motion between the tag and the antenna to analyze and calibrate the offset between the antenna phase center and the physical center, eliminating the fixed phase deviation caused by the hardware. S2. Multipath detection of phase measurements: Since the phase difference between adjacent frequency bands is the same, the variance of the phase difference between adjacent channels is calculated. Measurement points with a phase difference variance greater than the empirical threshold are judged to be subject to strong multipath interference and these measurement points are removed. Weak multipath repair is performed on the retained data: a weighting function is defined based on the phase dispersion of each frequency band, and a weighted linear fit is performed to make the phase close to the ideal straight line; S3. Unwrap the wrapped phase, remove the 2π jump to obtain a continuous phase; normalize with the lowest point to eliminate the fixed phase offset, compress the uncertainty of the initial phase value into a very small distance error, and form a residual motion / residual distance model and sequence accordingly. S4. Construct a cost / objective function from the residual distance sequence and perform nonlinear optimization; The two-dimensional position of the tag is obtained iteratively using Newton's method. If the antenna trajectory is not a straight line, the trajectory slope is taken at the lowest point to rotate the coordinate system. The solution is first obtained in the rotated coordinate system, and then the inverse rotation matrix is ​​used to restore it to the original coordinate system. In step S3, the phase change is considered as a phase curve, and the wrapped phase is unwrapped, as shown in the following formula: (1)； in, and These represent the phase data before and after untangling, respectively. , The unwrapped phase is represented by i=1,2,...,N; [·] represents an integer value function; sign(·) represents a sign function; after a phase jump is detected, the subsequent phase is added to or subtracted by 2π; finally, the signal phase after the phase unwrapping operation no longer has any jumps and becomes a continuous phase sequence; the distance between the reader antenna and the tag is calculated based on the unwrapped phase to determine the location information of the tag; In step S4, the initial position of the antenna, i.e., the position to be located, is set as follows: x0 and y0 represent the initial horizontal and vertical coordinates of the antenna; the reader performs the interrogation process at N different times, and let the antenna position at the nth interrogation be y0. x n and y n This represents the horizontal and vertical coordinates of the antenna at the nth monitoring point; and when n=1, When the nth query is executed, the relative distance between the tag and the antenna is: (2)； in, Indicates the position of the label, x t and y t Represent the x and y coordinates of the label; let The x-coordinate represents the position coordinates when the antenna is closest to the tag. min and y min Let x and y represent the horizontal and vertical coordinates when the antenna is closest to the tag, and let r represent the distance between the tag and the antenna's trajectory. (3)； residual distance This represents the actual distance between the antenna position and the tag position when the nth query is executed. The difference between r and r, and (4)； The motion model is as follows: (5)； in, Let n represent the distance of the antenna from the starting point to the nth point. Then we have: (6)； in, Substituting into equation (6), we get: (7)； get: (8)； Through the above data processing, the phase sequence is obtained. and time series , Then the residual distance sequence is obtained. : (9)； Construct the cost function: (10)； in Represented as: (11)； Then, the problem shown in equation (11) is solved by nonlinear optimization to obtain the label's location information: (12)； The above optimization problem is solved using Newton's method.

2. The method for automatic positioning of complex multi-path underground vehicles based on improved SAR according to claim 1, characterized in that: In step S1, assuming the origin is located at the physical center of the antenna, a tag with a spacing of 1m is placed in front of the antenna, and the tag is moved in the horizontal and vertical directions respectively. Phase data is collected every 2.5cm. The lowest point of the phase curve obtained in different directions is taken as the phase center point in that direction, and the physical center of the antenna is compensated by the distance between the observation point with the lowest actual phase and the physical center of the antenna.

3. The method for automatic positioning of complex multi-path underground vehicles based on improved SAR according to claim 1, characterized in that: In step S2, K frequency bands located between 920.625 and 924.375 MHz are provided by the COST RFID device, and K=16; the gap between two adjacent frequency bands is the same, which is 0.25 MHz. Theoretically, the phase difference between two adjacent frequency bands is measured. Same, that is , , and Represent the phase measured in the m-th and m-1-th frequency bands, respectively; using This represents the phase vector measured by the antenna at a certain measurement point from different channels. This represents the phase measurement value of the k-th frequency band. , This represents the clean, interference-free phase corresponding to that frequency band. Represented as: (13)； in, This represents the phase deviation in the k-th frequency band caused by multipath propagation. This indicates the clean phase corresponding to the first frequency band; for different frequency bands... The values ​​are different; Let the phase difference between adjacent channels be ,in , Let the variance of the phase difference be . The empirical threshold is When a certain measuring point If the time is right, remove that measurement point.

4. The method for automatic positioning of complex multi-path underground vehicles based on improved SAR according to claim 3, characterized in that: In step S2, weak multipath repair is performed on the retained data, and all measured phase values ​​are restored. Fit a straight line to the data, and define the fitted line as follows: (14)； Where y represents the ideal straight line, ; and define as follows: (15)； in, This represents the fitted value corresponding to frequency band k. This represents the weight corresponding to frequency band k; The sample mean difference of the phase received in frequency band k To represent the dispersion: (16)； in, This represents the phase samples reported in frequency band k. This represents the average of all m samples; it is considered true when at least 80% of the phase samples within the observation window have a mean absolute deviation below the upper limit of the static baseline. The expected value should be 0; definition for: (17)； The weighting function is further defined as follows: (18)； in, .

5. The method for automatic positioning of complex multi-path underground vehicles based on improved SAR according to claim 4, characterized in that: In step S2, by minimizing equation (15) Determine the slope of the fitted line And the intercept d value, calculate the variables respectively and d Partial derivatives: (19)； make Solving for the problem, we get: (20)； in, ； ； From the above formula, we can derive the slope of the fitted line y. There is an error between the intercept d and the ideal straight line. By determining the value of the error, the expected value of the phase can be obtained. Make an estimate.

6. The method for automatic positioning of complex multi-path underground vehicles based on improved SAR according to claim 1, characterized in that: In step S3, the unwrapping process of the phase still involves an uncertain number of cycles, i.e., the number of periods. Therefore, after unwrapping the phase, it is calibrated. relative distance The conversion between them is expressed as follows: (21)； in, Indicates wavelength. This represents the minimum untangling phase; the relative distance model is updated according to the following formula: (22)； in, This represents the distance corresponding to the initial phase.

7. The method for automatic positioning of complex multi-path underground vehicles based on improved SAR according to claim 6, characterized in that: In step S3, the fixed phase offset is eliminated by minimum point normalization, and the relative distance between the tag and the antenna is... Includes a fixed vertical distance r and a residual distance When the label is located at a vertical point hour, The phase shift caused by r is eliminated by the following formula: (23)； in, Indicates the residual phase. Indicates unwrapped phase The minimum value is obtained; thus, the phase curve with the lowest point of 0 is obtained, the phase period ambiguity information is removed, and the shape of the phase curve is maintained; since there is a linear relationship between phase and distance, the phase error is converted into a distance error; the motion model shown in equation (22) is modified into the following residual motion model: (24)； in, This indicates the distance error caused by the elimination of fixed phase offset.

8. The method for automatic positioning of complex multi-path underground vehicles based on improved SAR according to claim 1, characterized in that: In step S4, it is assumed that the timestamp of the lowest point obtained through phase unwrapping is... The trajectory of the antenna movement is in The slope at time is Then perform coordinate system transformation; First, the position of the antenna Rotate, rotation angle Depend on By applying this rotation matrix, the antenna's trajectory is transformed into a new coordinate system. as follows: (25)； The rotated antenna coordinates are ,Will and the residual distance sequence shown in equation (10) Substituting this into the localization algorithm, the coordinates of the target label are calculated as follows: That is, the position of the label in the rotated coordinate system; Next, we use the inverse rotation matrix to transform the rotated label coordinates back to the original coordinate system. The inverse rotation matrix is ​​the transpose of the rotation matrix, and is represented as: (26)； The position of the label in the original coordinate system is calculated. To obtain, that is The above operations transform the final position of the label from the rotated coordinate system back to the original coordinate system, thus completing the entire positioning calculation.

Images