A weighted fusion 5G NR joint positioning method and system

Abstract

This invention relates to the field of wireless positioning and mobile communication technology, and discloses a weighted fusion 5G NR joint positioning method and system. The method first acquires TDOA observations, Multi-RTT observations, and dual-frequency difference PDOA observations between multiple base stations with known locations and the target to be positioned, and selects one base station as a reference base station. Then, it constructs TDOA, Multi-RTT, and dual-frequency difference PDOA observation models and corresponding residual vectors, respectively. Subsequently, it obtains an initial position estimate of the target to be positioned based on the TDOA observation information, and obtains a local reference position estimate based on TDOA through Gauss-Newton iteration. Finally, it constructs a particle swarm search neighborhood and outputs the optimal swarm position as the final positioning result. This invention can alleviate the local minima problem caused by PDOA phase wrapping, thereby improving the positioning accuracy, stability, and engineering applicability in 5G NR positioning scenarios.

Description

Technical Field

[0001] This invention belongs to the field of wireless positioning and mobile communication technology, and particularly relates to a weighted fusion 5G NR joint positioning method and system. Background Technology

[0002] With the rapid development of fifth-generation mobile communication systems, the functions of communication networks have gradually expanded from traditional data transmission to integrated communication, sensing, and positioning. High-precision location information has significant application value in scenarios such as the Industrial Internet, intelligent transportation, unmanned systems, virtual reality, emergency rescue, and indoor asset management.

[0003] 5G NR systems, with their large bandwidth, multiple antennas, high frequency bands, and dense networking capabilities, provide a new physical foundation for improving wireless positioning accuracy. Existing 5G NR positioning methods include Multi-RTT, PDOA, Downlink Time Difference of Arrival (DL-TDOA), Uplink Time Difference of Arrival (UL-TDOA), Downlink Angle-of-Departure (DL-AoD), and Uplink Angle-of-Arrival (UL-AoA), and allow for the combined use of multiple positioning methods. Therefore, multi-source observation fusion has become an important direction for improving the positioning accuracy and stability of 5G NR.

[0004] In 5G NR positioning scenarios, TDOA, Multi-RTT, and PDOA observations can be obtained using base stations (BS), user equipment (UE), and their radio frequency transceiver modules, clock synchronization modules, and baseband processing modules. The BS includes 5G NR base stations, transmission reception points (TRPs), and other wireless nodes with known locations. All of the above observations can be obtained based on Orthogonal Frequency Division Multiplexing (OFDM) signals: TDOA observations are obtained by measuring the arrival time of downlink reference signals transmitted by multiple BS to the same UE and subtracting the values ​​from a reference BS; Multi-RTT observations are obtained through bidirectional reference signal interaction between the BS and the UE, and the round-trip time is calculated based on the transmission and reception times; PDOA observations are obtained by measuring the carrier phase of uplink reference signals transmitted by the UE to multiple BS and constructing a phase difference observation based on a reference BS.

[0005] However, in real-world 5G NR positioning environments, single-observation models are susceptible to factors such as base station geometry, measurement noise, synchronization errors, and non-line-of-sight propagation, resulting in limited positioning accuracy and stability. While multi-source observation fusion can improve positioning performance, TDOA, Multi-RTT, and PDOA correspond to time difference, round-trip time, and phase difference observations, respectively, and their dimensions, noise variance, and error correlations differ. Directly performing equal-weighted fusion can easily lead to high-noise observations negatively impacting positioning results. Furthermore, PDOA phase observations have inherent periodicity, and the actual phase difference is usually confined to a preset principal value range, easily leading to phase wrapping and phase ambiguity issues. When PDOA residuals are directly introduced into the positioning objective function, the objective function typically exhibits multiple local minima, making traditional local iterative methods prone to converging into erroneous phase ambiguity regions. Although some existing TDOA-PDOA joint positioning methods employ search methods such as particle swarm optimization to alleviate local extremum problems, they often fail to fully utilize the Multi-RTT bidirectional ranging information in 5G NR systems, and the search area often relies on fixed boundaries or empirical parameters, making it difficult to balance search reliability and computational complexity.

[0006] Therefore, there is an urgent need in this field for a 5G NR joint positioning method that can comprehensively utilize multi-source heterogeneous observation information from TDOA, Multi-RTT, and PDOA, and reasonably weight it according to the noise statistical characteristics of different observation types, while combining local reference position estimation to determine the search neighborhood, so as to improve the positioning accuracy, stability, and robustness in complex environments. Summary of the Invention

[0007] To overcome the shortcomings of existing technologies, the present invention aims to provide a weighted fusion 5G NR joint positioning method and system, which improves the positioning accuracy and stability under multi-source observation conditions, enhances the rationality of heterogeneous observation fusion and the search reliability under dual-frequency PDOA phase wrapping conditions, reduces the dependence of subsequent particle swarm search on the initial search area, takes into account local positioning efficiency and non-convex objective function search capabilities, and has good engineering deployment value.

[0008] This invention constructs corresponding observation models and residual vectors based on TDOA, Multi-RTT, and dual-frequency PDOA observation information between multiple known-location base stations (BSs) and the UE to be located. Weight matrices are constructed according to the covariance of various observation noise types to form a joint weighted nonlinear least squares objective function. Dual-frequency PDOA (Dual-Frequency Phase Difference of Arrival, DF-PDOA) refers to the observation constructed by using the phase difference observations obtained from the same non-reference base station and reference base station at two different carrier frequencies, and performing differential operations on the phase differences at different carrier frequencies. This is used to reduce the ambiguity caused by the periodicity in single-frequency phase observations, improving the uniqueness of phase observations and positioning stability. First, an initial position estimate of the UE to be located is obtained based on TDOA observation information. Then, using this initial position estimate as the starting point, a Gauss-Newton iteration based on the TDOA observation model is used to locally correct the UE position, obtaining a local reference position estimate based on TDOA. Finally, a particle swarm neighborhood search is performed centered on this local reference position estimate to alleviate the local extremum problem caused by PDOA phase encapsulation, and the final positioning result is output.

[0009] To overcome the limitations of existing single-observation positioning methods, the susceptibility of direct equal-weight fusion of different observations to high-noise observation terms, and the multiple local minima of the objective function caused by PDOA phase wrapping, this invention adopts the following technical solution: a weighted fusion 5G NR joint positioning method, comprising the following steps: Step 1: Obtain 5G NR wireless observation information by observing the radio frequency transceiver and baseband processing between multiple BSs with known locations and the UE to be located; The wireless observation information includes TDOA observations obtained based on downlink reference signal arrival time measurements, Multi-RTT observations obtained based on bidirectional signal interaction between the BS and the UE, and PDOA raw phase observations at two carrier frequencies, with one of the BSs selected as the reference BS. Step 2: Construct dual-frequency difference PDOA observations based on the original PDOA phase observations at two carrier frequencies; Based on the location of the BS, the location of the reference BS, and the location variables of the UE to be located, construct the TDOA observation model, the Multi-RTT observation model, and the dual-frequency difference PDOA observation model respectively, and construct the corresponding residual vectors according to the TDOA observations, Multi-RTT observations, and dual-frequency difference PDOA observations respectively. Step 3: Construct the TDOA weight matrix, Multi-RTT weight matrix, and dual-frequency difference PDOA weight matrix according to the noise statistical characteristics of various observations, and construct a joint weighted nonlinear least squares objective function from the residual vector and the weight matrix; Step 4: Obtain the initial position estimate of the UE to be located based on the TDOA observation information, and use the initial position estimate as the starting point of the iteration. Use the Gauss-Newton iterative method based on the TDOA observation model to locally correct the UE position and obtain the local reference position estimate based on TDOA. Step 5: Construct a particle swarm search neighborhood centered on the local reference position estimate based on TDOA, and use the particle swarm optimization algorithm within the search neighborhood, with the joint weighted nonlinear least squares objective function as the fitness function to evaluate candidate positions and obtain the swarm optimal position. Step 6: Output the optimal location of the group as the final positioning result of the UE to be located.

[0010] Furthermore, in step 2, the TDOA observation model is used to characterize the signal arrival time difference between the non-reference BS and the reference BS, the Multi-RTT observation model is used to characterize the bidirectional signal round-trip propagation time between the UE to be located and each BS, and the dual-frequency difference PDOA observation model is used to characterize the difference in carrier phase difference observations between the same non-reference BS and the reference BS at two carrier frequencies.

[0011] Furthermore, the dual-frequency difference PDOA observation model constructs dual-frequency difference phase observations using the original PDOA phase observations at two carrier frequencies, in order to reduce the impact of phase periodicity at a single carrier frequency on the positioning results.

[0012] Furthermore, the dual-frequency difference PDOA residual vector is constructed using a phase wrapping function, which maps the dual-frequency difference PDOA phase residual to a preset principal value interval, preferably... .

[0013] Furthermore, in step 3, the TDOA weight matrix, Multi-RTT weight matrix, and dual-frequency difference PDOA weight matrix are determined by the inverse of the corresponding observation noise covariance matrix; when the observation noises of TDOA, Multi-RTT, and dual-frequency difference PDOA are independent of each other, the joint weight matrix is ​​constructed in a block diagonal form.

[0014] Furthermore, TDOA observation noise includes link time measurement noise and base station synchronization error; since different TDOA differential observations share the link time measurement noise of the reference BS, the TDOA observation noise covariance matrix includes non-zero off-diagonal elements.

[0015] Furthermore, in step 4, the initial position estimate is obtained by pseudo-linearizing the TDOA observation model and solving it using a two-step weighted least squares method.

[0016] Further, in step 4, the Gauss-Newton iterative method includes: calculating the TDOA residual vector and the TDOA Jacobian matrix at the current UE position estimate; calculating the position increment based on the TDOA residual vector, the TDOA Jacobian matrix, and the TDOA weight matrix; and updating the UE position estimate based on the position increment. When the difference between two consecutive UE position estimates is less than a preset convergence threshold, or when the preset maximum number of iterations is reached, the TDOA-based local reference position estimate is output.

[0017] Furthermore, in step 5, the perturbation scale of the particle swarm search neighborhood is determined based on the Jacobian matrix and the TDOA weight matrix in the TDOA local reference position estimation process; and, in the particle swarm initialization process, the TDOA-based local reference position estimate is set as the position of one of the initial particles.

[0018] Furthermore, the particle swarm optimization algorithm performs neighborhood search centered on the local reference position estimate based on TDOA, instead of performing a global random search directly over the entire localization region, thereby enhancing the search capability for non-convex objective functions while reducing computational complexity.

[0019] A weighted fusion 5G NR joint positioning system for performing a 5G NR joint positioning method, comprising: The observation information acquisition module is used to acquire TDOA observation values, Multi-RTT observation values, and PDOA raw phase observation values ​​at two carrier frequencies between multiple BSs with known locations and the UE through the RF transceiver module, clock synchronization module, and baseband processing module of the BS and the UE, and select one of the BSs as the reference BS. The observation model and residual construction module is used to construct dual-frequency difference PDOA observation values ​​based on the original phase observation values ​​of PDOA under the two carrier frequencies, and to construct TDOA observation model, Multi-RTT observation model and dual-frequency difference PDOA observation model based on the position of the BS, the position of the reference BS and the position variable of the UE to be located, and construct the corresponding residual vectors. The weight matrix and joint objective function construction module is used to construct the TDOA weight matrix, Multi-RTT weight matrix and dual-frequency difference PDOA weight matrix according to the noise statistical characteristics of various observations, and construct the joint weighted nonlinear least squares objective function from the residual vector and the weight matrix. The initial estimation module is used to obtain the initial position estimate of the UE to be located based on TDOA observation information; The local optimization module is used to locally correct the UE position using the initial position estimate as the starting point of the iteration and the Gauss-Newton iterative method based on the TDOA observation model, so as to obtain a local reference position estimate based on TDOA. The particle swarm search module is used to construct a particle swarm search neighborhood centered on the estimated local reference position based on TDOA, and to evaluate candidate positions within the search neighborhood using a particle swarm optimization algorithm with the joint weighted nonlinear least squares objective function as the fitness function, thereby obtaining the swarm's optimal position; and The positioning result output module is used to output the optimal position of the group as the final positioning result of the UE to be located.

[0020] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. This invention comprehensively utilizes three types of observation information—TDOA, Multi-RTT, and dual-frequency PDOA—within a unified positioning framework, enabling complementary constraints between different observation methods. This compensates for the insufficient accuracy and reliability of a single observation method in complex environments, improves the availability of the 5G NR positioning system, and enhances the positioning accuracy and stability under multi-source observation conditions.

[0021] 2. This invention constructs a weight matrix based on the noise covariance of different observation types, enabling heterogeneous observations with different dimensions and noise levels to be reasonably integrated in a unified objective function, reducing the adverse effects of high-noise observation terms on the positioning results, and improving the rationality of heterogeneous observation fusion.

[0022] 3. This invention introduces phase wrapping processing into the PDOA residual and enhances the search capability of the non-convex objective function through particle swarm neighborhood search, which can alleviate the local minima problem caused by PDOA phase ambiguity, improve the reliability of the final positioning result, and improve the search reliability under dual-frequency PDOA phase wrapping conditions.

[0023] 4. The present invention first obtains the local reference position estimate of the UE to be located based on TDOA observation information, and uses the local reference position estimate as the center of the subsequent particle swarm search, thereby reducing the uncertainty of the global random search, improving the stability of the search process, and reducing the dependence of the subsequent particle swarm search on the initial search area.

[0024] 5. The particle swarm search neighborhood is limited by the local reference position estimate based on TDOA, avoiding large-scale global random search in the entire localization area, thereby reducing computational complexity. At the same time, during the particle swarm search stage, the candidate position is evaluated by a joint weighted nonlinear least squares objective function of TDOA, Multi-RTT and dual-frequency PDOA, which retains the search capability for non-convex objective functions and takes into account both local localization efficiency and non-convex objective function search capability.

[0025] 6. This invention is applicable to 5G NR indoor or outdoor positioning scenarios. It can obtain observation data based on downlink positioning reference signals, uplink detection reference signals, or bidirectional signal interaction between the BS and UE. It does not require changes to the basic hardware architecture of the positioning system, is easy to implement in engineering, and has good engineering deployment value. Attached Figure Description

[0026] Figure 1 This is a flowchart of the 5G NR joint positioning method based on the fusion of TDOA, Multi-RTT and dual-frequency PDOA information of the present invention.

[0027] Figure 2 This is a distribution diagram of BS and UE in this invention.

[0028] Figure 3 The RMSE measurement error over time in this invention The change curve.

[0029] Figure 4 In this invention, RMSE varies with the number of particle swarms. N The change curve.

[0030] Figure 5 The RMSE measurement error in this invention is related to the phase difference observation. Change curve graph.

[0031] Figure 6 This is a block diagram of the joint positioning system of the present invention. Detailed Implementation

[0032] The following specific examples illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. Those skilled in the art can make equivalent adjustments to the observed signal source, number of base stations (BS), number of carrier frequencies, iteration threshold, and search neighborhood range without departing from the concept of the present invention. It should be noted that, unless otherwise specified, the following embodiments and features described therein can be combined with each other.

[0033] The purpose of this invention is to address the shortcomings of existing technologies by providing a weighted fusion 5G NR joint positioning method and system.

[0034] Example 1 The preferred embodiment of the present invention considers a two-dimensional positioning scenario, and has a total of The location of the BS is known. This BS can be a 5G NR base station, a Transmitter-Receiver Point (TRP), or other known wireless nodes. The UE to be located can be a mobile terminal, an unmanned device, or other wireless terminal requiring location tracking. The position of each BS is represented as follows: The location of the UE to be located is represented as follows: The UE to be located is at the... The geometric distance between BSs is represented as: .

[0035] Based on the above spatial model, such as Figure 1 As shown, a preferred embodiment of the present invention discloses a weighted fusion 5G NR joint positioning method, the steps of which are as follows: Step 1: Acquire 5G NR radio observation information. Multiple BSs with known locations and the UE to be located transmit and receive 5G NR reference signals through the radio frequency transceiver module. The baseband processing module performs synchronization, demodulation, correlation detection, time estimation, and phase estimation to obtain TDOA, Multi-RTT, and dual-frequency PDOA observation information. The BSs mentioned above include 5G NR base stations, TRPs, and other radio nodes with known locations.

[0036] In this embodiment, the TDOA observation is obtained by estimating the arrival time of downlink reference signals transmitted by multiple BSs to the same UE and subtracting them from a reference BS; the Multi-RTT observation is obtained by bidirectional reference signal interaction between the BS and the UE, and the round-trip propagation time is calculated based on the transmission and reception times; the PDOA raw phase observation is obtained by the UE performing channel estimation on uplink reference signals transmitted by multiple BSs. Specifically, the UE estimates the complex channel response corresponding to each BS based on the known uplink reference signal, and takes the phase angle of the complex channel response to obtain the carrier phase of each BS signal arriving at the UE; then, the carrier phase of the non-reference BS and the reference BS are subtracted to obtain the PDOA raw phase observation. The system acquires the PDOA observation corresponding to the same BS at two carrier frequencies respectively, and subtracts the PDOA observations at these two carrier frequencies to construct dual-frequency PDOA observations.

[0037] Step 2: Construct TDOA, Multi-RTT, and Dual-Frequency PDOA Observation Models and Joint Residual Models. Based on the location of the BS, the location of the reference base station, and the location variables of the UE to be located, construct TDOA observation models, Multi-RTT observation models, and Dual-Frequency PDOA observation models respectively.

[0038] In TDOA positioning, assume the UE signal arrives at the first... The measurement times for the individual BS and the reference BS are respectively and Then the first The TDOA observation model of a BS relative to a reference BS is expressed as: , in, For signal propagation speed, For the first The TDOA time difference measurement noise of a non-reference BS relative to a reference BS.

[0039] To characterize the error sources in TDOA differential observation, this embodiment decomposes them into two parts: link time measurement error and base station synchronization error. Let the first... The link time measurement noise of the individual BS and the reference BS are respectively and And set the first The equivalent base station synchronization error corresponding to each differential observation is: Then there is Assume that the time measurement noise of each link is independent and satisfies the following conditions: , Meanwhile, it is assumed that the equivalent base station synchronization errors corresponding to different differential observations are independent of each other, and satisfy the following conditions: Therefore, for any non-reference base station, its TDOA time difference measurement noise satisfies The TDOA observation vector is represented as: , in, , , .

[0040] For the multi-BS collaborative multi-RTT observation model, UE and the first The round-trip propagation time between two BSs is obtained through bidirectional signal exchange. Let the first BS be... The Multi-RTT observations corresponding to each BS are: The Multi-RTT observation model can then be expressed as: , in, For RTT time measurement noise, preferably, the RTT time measurement noise is composed of the superposition of uplink and downlink single-link noise, and the different single-link noises are independent of each other and obey the rules. Then there is The Multi-RTT observation vector is represented as: , in, , , , .

[0041] In PDOA positioning, the phase difference between the non-reference BS signal and the reference BS signal arriving at the UE to be positioned is used for position estimation. To reduce the impact of phase periodicity under a single carrier frequency on the positioning results, this embodiment preferably adopts a dual-frequency PDOA observation model. This model acquires the PDOA observation values ​​of the same non-reference BS relative to the reference BS at two different carrier frequencies, and then subtracts the observation values ​​at the two carrier frequencies to construct dual-frequency difference PDOA observation values. Assume the system uses two carrier frequencies. and The first BS is the reference BS. For the... Each non-reference BS, at the carrier frequency and The phase difference observations relative to the reference BS are constructed below, denoted as . and Based on this, the PDOA observations at two carrier frequencies are subtracted from the same non-reference BS and reference BS to obtain the dual-frequency PDOA observations: . Therefore, the first The dual-frequency PDOA observation model corresponding to a non-reference BS can be expressed as: , in, This represents the phase wrapping function, used to constrain the phase within... Within the range; This indicates the phase measurement noise of the dual-frequency PDOA, and . and Let be the PDOA phase noise at different carrier frequencies. Assume that the phase noise at different carrier frequencies is independent, with a mean of 0 and a variance of . The dual-frequency PDOA noise then satisfies: The PDOA observation vector is represented as: , in, , , , .

[0042] Therefore, the residual vectors of TDOA, Multi-RTT, and PDOA are respectively expressed as: , , .

[0043] The joint residual vector is represented as: .

[0044] Step 3: Construct the joint weighted objective function. Since the residuals of TDOA, Multi-RTT, and PDOA correspond to time difference, round-trip time, and phase difference observations, respectively, their dimensions and noise statistics differ. Therefore, this example uses a weighted sum of squared residuals to construct the joint localization objective function.

[0045] The joint localization objective function can be expressed as: The weight matrix is ​​determined by the covariance matrix of the corresponding observation noise, and is expressed as: .

[0046] The joint weight matrix is ​​expressed as: .

[0047] Since each TDOA differential observation includes reference BS measurement noise, there is a correlation between the noise from different TDOA observations. The off-diagonal elements of the covariance matrix represent the variance of the reference BS measurement noise. Therefore, the TDOA noise covariance matrix can be expressed as: , in, .

[0048] The RTT noise covariance matrix is ​​expressed as: .

[0049] The noise covariance matrix of PDOA is expressed as: .

[0050] Step 4: Obtain the local reference position estimate based on TDOA. To reduce the dependence of subsequent particle swarm neighborhood search on the initial search area, this embodiment first uses TDOA observation information to obtain the local reference position estimate of the UE to be located. It should be noted that the local reference position estimate is not used as the final positioning result, but rather as the search center for subsequent particle swarm neighborhood search.

[0051] In a preferred embodiment, an initial position estimate is first obtained based on the TDOA observation model. The initial position estimate is obtained in this example using the TDOA two-step weighted least squares method.

[0052] Subsequently, using the initial position estimate As the starting point of the iteration, the TDOA observation model is locally iteratively corrected. Let the... The position estimate at the next iteration is: The corresponding TDOA residual vector is represented as: ,in, For TDOA observation vectors, Current location The TDOA observation vector is shown below. The estimated value at the current location is... Nearby, a first-order approximation is made to the TDOA observation function, and the corresponding Jacobian matrix is ​​constructed: .

[0053] For the For each non-reference BS, its corresponding Jacobian vector can be expressed as: .

[0054] Based on the TDOA residual vector, TDOA weight matrix, and TDOA Jacobian matrix, calculate the... Position increment in the next iteration: .

[0055] Then update the UE location estimate as follows: , in, This is the step size factor, used to control the position update magnitude in each iteration.

[0056] The difference between the estimates in the current two iterations is less than a preset threshold. At that time, that is The iteration is terminated. After the iteration terminates, the local reference position estimate based on TDOA is obtained: .

[0057] Local reference position estimate based on TDOA Used to determine the search center for subsequent particle swarm neighborhood searches.

[0058] Step 5: Particle Swarm Optimization (PSO) Neighborhood Search. To alleviate the multiple local minima problem caused by PDOA phase wrapping, this embodiment introduces a particle swarm optimization algorithm near the local reference position estimate based on TDOA. Unlike performing a global random search directly over the entire positioning area, this embodiment uses the local reference position estimate based on TDOA... Using the search center as the search area, a joint weighted objective function of TDOA, Multi-RTT, and PDOA is searched within its neighborhood.

[0059] The particle swarm optimization algorithm uses the aforementioned joint weighted objective function as the fitness function, expressed as: .

[0060] In a preferred embodiment, the particle swarm search neighborhood is determined based on the approximate covariance obtained during the TDOA local reference position estimation process. Let the Jacobian matrix at the convergence position of the TDOA Gaussian-Newton iteration be... Then the approximate covariance of the local reference position estimate can be expressed as: .

[0061] Based on this approximate covariance, the search neighborhood radius can be expressed as: , in, This is the search radius scaling factor. Centered on, with a radius of Initialize the particle swarm within the neighborhood of . Let the . The initial positions of the particles are Then it can be expressed as: ,in, This is the identity matrix. To ensure that the local reference position itself can participate in subsequent searches, in a preferred embodiment, the initial position of one of the particles is set to: .

[0062] Let the first The particle in the first The position and velocity at the next iteration are respectively and Its speed and position update rules are as follows:

[0063] in, The velocity contraction factor, and These are individual learning factors and group learning factors, respectively. and To get the value of an element Random vectors within an interval Represents element-wise product; Indicates the first The historical best position of each particle This represents the historical best position of the entire particle swarm.

[0064] In each iteration, according to the cost function Calculate the fitness value corresponding to the current position of each particle, and update the historical best position of each particle and the swarm's optimal position accordingly. The particle swarm search terminates when the change in the swarm's optimal fitness value between two iterations is less than a preset threshold or when the maximum number of searches is reached. After the particle swarm search terminates, the swarm's optimal position is taken as the final joint localization result for the UE to be located, expressed as:

[0065] in, This represents the actual number of termination iterations in the particle swarm search.

[0066] This embodiment limits the particle swarm search neighborhood by using the local reference position estimate based on TDOA, which avoids performing a large-scale global random search directly over the entire positioning area, thereby reducing computational complexity. At the same time, during the particle swarm search stage, the candidate positions are evaluated using a joint weighted objective function of TDOA, Multi-RTT, and PDOA, so that the final positioning result can integrate multi-source observation information and correct the local minima problem caused by PDOA phase wrapping.

[0067] like Figure 6 As shown, a weighted fusion 5G NR joint positioning system of the present invention is used to execute a 5G NR joint positioning method, including: The observation information acquisition module is used to acquire TDOA observation values, Multi-RTT observation values, and PDOA raw phase observation values ​​at two carrier frequencies between multiple BSs with known locations and the UE through the RF transceiver module, clock synchronization module, and baseband processing module of the BS and the UE, and select one of the BSs as the reference BS. The observation model and residual construction module is used to construct dual-frequency difference PDOA observation values ​​based on the original phase observation values ​​of PDOA under two carrier frequencies, and to construct TDOA observation model, Multi-RTT observation model and dual-frequency difference PDOA observation model based on the position of BS, the position of reference BS and the position variables of the UE to be located, and construct the corresponding residual vectors. The weight matrix and joint objective function construction module is used to construct the TDOA weight matrix, Multi-RTT weight matrix and dual-frequency difference PDOA weight matrix according to the noise statistical characteristics of various observations, and construct the joint weighted nonlinear least squares objective function from the residual vector and the weight matrix. The initial estimation module is used to obtain the initial position estimate of the UE to be located based on TDOA observation information; The local optimization module is used to locally correct the UE position using the initial position estimate as the starting point of the iteration and the Gauss-Newton iterative method based on the TDOA observation model, so as to obtain the local reference position estimate based on TDOA. The particle swarm optimization (PSO) module constructs a PSO search neighborhood centered on the TDOA-based local reference position estimate. Within this neighborhood, it employs a PSO algorithm, using a joint weighted nonlinear least squares objective function as the fitness function to evaluate candidate positions and obtain the swarm's optimal position. The positioning result output module is used to output the optimal location of the group as the final positioning result of the UE to be located.

[0068] Example 2 Based on Embodiment 1, the following application examples are provided to verify the technical advantages of this invention. It should be noted that the following application examples are only used to illustrate the positioning performance of the method under typical parameter conditions and are not intended to limit the scope of protection of this invention. Without departing from the concept of this invention, those skilled in the art can make equivalent adjustments to the relevant parameters based on the size of the positioning area, the geometric distribution of base stations, the observation noise level, the number of carrier frequencies, and computational resource constraints.

[0069] Scenario: See Figure 2 The distribution map of BS and UE is a two-dimensional 5G NR positioning scenario.

[0070] BS: Set up six known BS ​​locations with coordinates as follows: (500,0), (250,433), (-250,433), (-500,0), (-250,-433), (250,-433). Select the first BS as the reference BS.

[0071] UE: Set the actual location of the UE to be located to (100, 50).

[0072] The signal propagation speed is set to: .

[0073] The system uses two carrier frequencies, namely: .

[0074] Zero-mean Gaussian noise was added to the TDOA, Multi-RTT, and PDOA observations to simulate measurement errors in actual 5G NR positioning. 1000 Monte Carlo simulations were performed under each set of parameters, and the root mean square error (RMSE) was used as the positioning performance evaluation metric.

[0075] The maximum number of Gauss-Newton iterations is set to: The convergence threshold is set as follows: Step size factor is set as follows: .

[0076] Common parameter settings for particle swarm search are as follows: Particle swarm size Maximum number of particle swarm searches Particle initialization perturbation scale scaling factor velocity contraction factor Individual learning factor Group learning factor The parameters described above are only a preferred setting in this embodiment. In practical applications, they can be adjusted according to the size of the positioning area, the geometric distribution of base stations, the observation noise level, and computing resource constraints.

[0077] To verify the positioning performance of the method of the present invention under the influence of different factors, the following three sets of simulation experiments were set up in this embodiment.

[0078] The first set of experiments was used to verify the impact of time measurement errors on positioning performance. In this experiment, the BS location, UE location, carrier frequency, particle swarm parameters, and PDOA phase measurement errors were kept constant, while the standard deviation of the single-link time measurement error was varied. .in, The standard deviation of base station synchronization error is set as follows: The standard deviation of PDOA phase measurement error is fixed as follows: .

[0079] The second set of experiments was used to verify the impact of particle swarm size on joint localization performance. In this experiment, the BS location, the UE location, carrier frequency, observation noise conditions, and other algorithm parameters were kept constant, while the particle swarm size was varied. The standard deviation of the single-link time measurement error is set as follows: The standard deviation of base station synchronization error is set as follows: The standard deviation of PDOA phase measurement error is fixed as follows: The particle swarm size is set sequentially as follows: .

[0080] The third set of experiments was used to verify the impact of PDOA measurement errors on joint positioning performance. In this experiment, the BS location, UE location, carrier frequency, TDOA and Multi-RTT related noise parameters, and particle swarm size were kept constant, while the standard deviation of the PDOA equivalent measurement error was varied. The standard deviation of the single-link time measurement error was set as follows: The standard deviation of base station synchronization error is set as follows: , The standard deviation of PDOA measurement error is set as follows: .

[0081] In each Monte Carlo simulation experiment, the method of the present invention is executed according to the following procedure: 1) Based on the preset BS coordinates and the actual position of the UE to be located, calculate the actual geometric distance between the UE to be located and each BS, and generate noise-free theoretical observations of TDOA, Multi-RTT and PDOA. 2) Based on the set single-link time measurement error, base station synchronization error and PDOA measurement error, zero-mean Gaussian noise is added to the noise-free theoretical observation values ​​to obtain the corresponding TDOA observation vector, Multi-RTT observation vector and PDOA observation vector; 3) Obtain the initial position estimate of the UE to be located based on the TDOA observation vector, and use the initial position estimate as the starting point of the iteration. Use the Gauss-Newton iteration based on the TDOA observation model to locally correct the UE position and obtain the local reference position estimate based on TDOA. 4) Based on the Jacobian matrix and TDOA weight matrix in the local reference position estimation process, calculate the approximate covariance of the local reference position estimation, and determine the perturbation scale for particle swarm initialization accordingly. 5) Initialize the particle swarm with the local reference position estimate based on TDOA as the center, and use the local reference position estimate as the position of one of the initial particles to ensure that the local reference position estimate participates in the subsequent candidate position search; 6) During the particle swarm search process, the joint weighted nonlinear least squares objective function of TDOA, Multi-RTT and PDOA is used as the fitness function to evaluate each candidate position, and the historical best position of each particle and the swarm's population best position are updated according to the evaluation results. 7) When the change in the optimal fitness value of the swarm between two iterations is less than the preset threshold or the preset maximum number of searches is reached, the particle swarm search is terminated, and the optimal position of the swarm is output as the localization result of this simulation experiment. 8) Repeat the Monte Carlo simulation experiment described above. Based on the deviation between the positioning results obtained from each simulation and the actual position of the UE to be positioned, calculate the root mean square error (RMSE) of positioning under different parameter conditions, and evaluate the positioning accuracy and stability of the method of the present invention accordingly.

[0082] Figure 3 Demonstrates measurement errors at different times Comparison of RMSE results for various localization algorithms under different conditions. Figure 3 It is evident that, with As the time difference measurement error increases, the positioning errors of all algorithms gradually rise, indicating that increased time difference measurement error reduces positioning accuracy. Specifically, the RMSE of the iterative TDOA algorithm is higher than that of the iterative Multi-RTT algorithm. This is because TDOA observations are affected by both time difference measurement errors and base station synchronization errors, while Multi-RTT observations can utilize the round-trip distance constraints from the UE to multiple BSs, thus providing more stable positioning results. Furthermore, the RMSE of the Multi-RTT and dual-frequency PDOA fusion method is lower than that of the iterative Multi-RTT algorithm, indicating that dual-frequency PDOA phase observations can provide finer phase difference constraints on top of RTT distance constraints, thereby further improving positioning accuracy. Further comparison shows that the RMSE of the joint positioning method of this invention is lower than that of the Multi-RTT and dual-frequency PDOA fusion method. Both methods use the Gauss-Newton iteration result of TDOA as the particle swarm search center. However, the Multi-RTT and dual-frequency PDOA fusion method mainly uses RTT residuals and dual-frequency PDOA residuals to evaluate candidate positions during the particle swarm search stage. In contrast, the joint localization method of this invention further introduces TDOA residuals, so that the candidate positions are simultaneously constrained by differential time constraints, round-trip distance constraints, and phase difference constraints, thus achieving a lower localization error.

[0083] Figure 4 The RMSE comparison results of the Multi-RTT and dual-frequency PDOA fusion methods and the joint localization method of this invention under different particle numbers (N) are presented. Figure 4 It is evident that when the number of particles is small, the particle swarm search capability is limited, and the algorithm struggles to fully search the optimal region of the joint objective function, resulting in a larger localization error. As the number of particles increases, the RMSE of both methods decreases rapidly and gradually approaches their respective theoretical lower bounds. Once the number of particles reaches a certain scale, the RMSE remains relatively stable, indicating that further increasing the number of particles has limited effect on improving localization accuracy, primarily increasing computational complexity. Further comparison shows that, under the same particle number condition, the RMSE of the joint localization method of this invention is lower than that of the Multi-RTT and dual-frequency PDOA fusion method. This is because the joint localization method of this invention simultaneously introduces three types of residuals—TDOA, Multi-RTT, and dual-frequency PDOA—into the particle swarm fitness function, while the Multi-RTT and dual-frequency PDOA fusion method mainly utilizes RTT residuals and dual-frequency PDOA residuals for search evaluation. Therefore, the observation constraints of the joint localization method of this invention are more complete, achieving higher localization accuracy under the same search scale.

[0084] Figure 5 Demonstrates different dual-frequency PDOA measurement errors Comparison of RMSE results for various localization algorithms under different conditions. Figure 5It is evident that the RMSE of the iterative TDOA algorithm and the iterative Multi-RTT algorithm remains essentially unchanged. This is because they use only TDOA and Multi-RTT observations for localization, respectively, and their results are unaffected by changes in dual-frequency PDOA measurement errors. As the error increases, the RMSE of both the Multi-RTT and dual-frequency PDOA fusion method and the joint positioning method of this invention gradually rises, indicating that the increased measurement error of dual-frequency PDOA weakens the effect of phase constraint on improving positioning accuracy. When the phase difference is small, dual-frequency PDOA can provide finer phase difference constraints, thus the positioning error of the fusion algorithm is significantly lower than that of the standalone TDOA algorithm and the standalone Multi-RTT algorithm. Further comparison shows that, under the same dual-frequency PDOA measurement error conditions, the RMSE of the joint positioning method of this invention is lower than that of the Multi-RTT and dual-frequency PDOA fusion method. This is because the joint positioning method of this invention further introduces TDOA constraints on top of RTT and dual-frequency PDOA constraints, subjecting candidate locations to the combined constraints of multi-source observation information, thereby improving positioning accuracy.

[0085] In summary, without altering the basic hardware architecture of the 5G NR positioning system, this invention improves positioning accuracy and stability under conditions of different time measurement errors, different particle numbers, and different dual-frequency PDOA measurement errors by using weighted fusion of TDOA, Multi-RTT, and dual-frequency PDOA multi-source observations, combined with particle swarm neighborhood search based on the local reference position of TDOA. It is suitable for 5G NR indoor and outdoor positioning scenarios where time measurement errors, synchronization errors, and phase wrapping effects exist.

[0086] Furthermore, it should be understood that although this specification describes embodiments, not every embodiment contains only one independent technical solution. This narrative style is merely for clarity. Those skilled in the art should consider the specification as a whole, and the technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art.

Claims

1. A weighted fusion 5G NR joint positioning method, characterized in that, Includes the following steps: Step 1: Obtain 5G NR wireless observation information by observing the radio frequency transceiver and baseband processing between multiple BSs with known locations and the UE to be located; The wireless observation information includes TDOA observations obtained based on downlink reference signal arrival time measurements, Multi-RTT observations obtained based on bidirectional signal interaction between the BS and the UE, and PDOA raw phase observations at two carrier frequencies, with one of the BSs selected as the reference BS. Step 2: Construct dual-frequency difference PDOA observations based on the original PDOA phase observations at two carrier frequencies; Based on the location of the BS, the location of the reference BS, and the location variables of the UE to be located, construct the TDOA observation model, the Multi-RTT observation model, and the dual-frequency difference PDOA observation model respectively, and construct the corresponding residual vectors according to the TDOA observations, Multi-RTT observations, and dual-frequency difference PDOA observations respectively. Step 3: Construct the TDOA weight matrix, Multi-RTT weight matrix, and dual-frequency difference PDOA weight matrix according to the noise statistical characteristics of various observations, and construct a joint weighted nonlinear least squares objective function from the residual vector and the weight matrix; Step 4: Obtain the initial position estimate of the UE to be located based on the TDOA observation information, and use the initial position estimate as the starting point of the iteration. Use the Gauss-Newton iterative method based on the TDOA observation model to locally correct the UE position and obtain the local reference position estimate based on TDOA. Step 5: Construct a particle swarm search neighborhood centered on the local reference position estimate based on TDOA, and use the particle swarm optimization algorithm within the search neighborhood, with the joint weighted nonlinear least squares objective function as the fitness function to evaluate candidate positions and obtain the swarm optimal position. Step 6: Output the optimal location of the group as the final positioning result of the UE to be located.

2. The weighted fusion 5G NR joint positioning method according to claim 1, characterized in that: In step 2, the TDOA observation model is used to characterize the signal arrival time difference between the non-reference BS and the reference BS, the Multi-RTT observation model is used to characterize the bidirectional signal round-trip propagation time between the UE to be located and each BS, and the dual-frequency difference PDOA observation model is used to characterize the difference in carrier phase difference observation between the same non-reference BS and the reference BS at two carrier frequencies.

3. The weighted fusion 5G NR joint positioning method according to claim 2, characterized in that: The dual-frequency difference PDOA observation model constructs dual-frequency difference phase observations using the original PDOA phase observations at two carrier frequencies, in order to reduce the impact of phase periodicity at a single carrier frequency on the positioning results.

4. The weighted fusion 5G NR joint positioning method according to claim 3, characterized in that: The dual-frequency difference PDOA residual vector is constructed by a phase wrapping function, which is used to map the dual-frequency difference PDOA phase residual to a preset principal value range.

5. The weighted fusion 5G NR joint positioning method according to claim 1, characterized in that: In step 3, the TDOA weight matrix, Multi-RTT weight matrix, and dual-frequency difference PDOA weight matrix are determined by the inverse of the corresponding observation noise covariance matrix. When the observation noises of TDOA, Multi-RTT, and dual-frequency difference PDOA are independent of each other, the joint weight matrix is ​​constructed in a block diagonal form.

6. The weighted fusion 5G NR joint positioning method according to claim 5, characterized in that: TDOA observation noise includes link time measurement noise and base station synchronization error; since different TDOA differential observations share the link time measurement noise of the reference BS, the TDOA observation noise covariance matrix includes non-zero off-diagonal elements.

7. The weighted fusion 5G NR joint positioning method according to claim 1, characterized in that: In step 4, the initial position estimate is obtained by pseudo-linearizing the TDOA observation model and solving it using a two-step weighted least squares method.

8. The weighted fusion 5G NR joint positioning method according to claim 1, characterized in that: In step 4, the Gauss-Newton iterative method includes: calculating the TDOA residual vector and the TDOA Jacobian matrix at the current UE position estimate; calculating the position increment based on the TDOA residual vector, the TDOA Jacobian matrix, and the TDOA weight matrix; and updating the UE position estimate based on the position increment. When the difference between two consecutive UE position estimates is less than a preset convergence threshold, or when the preset maximum number of iterations is reached, the TDOA-based local reference position estimate is output.

9. The weighted fusion 5G NR joint positioning method according to claim 8, characterized in that: In step 5, the perturbation scale of the particle swarm search neighborhood is determined based on the Jacobian matrix and TDOA weight matrix in the TDOA local reference position estimation process; and, in the particle swarm initialization process, the TDOA-based local reference position estimate is set as the position of one of the initial particles.

10. A weighted fusion 5G NR joint positioning system, used to perform the 5G NR joint positioning method as described in any one of claims 1 to 9, characterized in that, include: The observation information acquisition module is used to acquire TDOA observation values, Multi-RTT observation values, and PDOA raw phase observation values ​​at two carrier frequencies between multiple BSs with known locations and the UE through the RF transceiver module, clock synchronization module, and baseband processing module of the BS and the UE, and select one of the BSs as the reference BS. The observation model and residual construction module is used to construct dual-frequency difference PDOA observation values ​​based on the original phase observation values ​​of PDOA under the two carrier frequencies, and to construct TDOA observation model, Multi-RTT observation model and dual-frequency difference PDOA observation model based on the position of the BS, the position of the reference BS and the position variable of the UE to be located, and construct the corresponding residual vectors. The weight matrix and joint objective function construction module is used to construct the TDOA weight matrix, Multi-RTT weight matrix and dual-frequency difference PDOA weight matrix according to the noise statistical characteristics of various observations, and construct the joint weighted nonlinear least squares objective function from the residual vector and the weight matrix. The initial estimation module is used to obtain the initial position estimate of the UE to be located based on TDOA observation information; The local optimization module is used to locally correct the UE position using the initial position estimate as the starting point of the iteration and the Gauss-Newton iterative method based on the TDOA observation model, so as to obtain a local reference position estimate based on TDOA. The particle swarm search module is used to construct a particle swarm search neighborhood centered on the local reference position estimate based on TDOA, and to use the particle swarm optimization algorithm within the search neighborhood, using the joint weighted nonlinear least squares objective function as the fitness function to evaluate candidate positions and obtain the swarm's optimal position. as well as The positioning result output module is used to output the optimal position of the group as the final positioning result of the UE to be located.

Images