Differential positioning method and device based on 5G communication signal

Abstract

The invention relates to the technical field of positioning, in particular to a differential positioning method and device based on 5G communication signals, and the method comprises the following steps: carrying out the data preprocessing of a GNSS original observation value collected by a mobile terminal based on a deep learning method; speed calculation is carried out by fusing sensor data including an inertial measurement unit, pseudo-range single-point positioning, SPV speed measurement and TDCP speed measurement are respectively executed, and different speed measurement results are fused by using a particle filter algorithm to obtain an optimal speed; the obtained optimal speed is substituted into a Kalman filter, and iterative optimization is carried out through the Kalman filter; and evaluating the reliability of the RTK solution according to the effective phase number and the position precision factor, thereby realizing real-time positioning. According to the invention, the anti-interference capability of the system is enhanced, the centimeter-level or even millimeter-level positioning precision can be realized, and the problems that the mobile terminal is difficult to acquire stable high-precision position information and the positioning error may reach several meters or even dozens of meters in the environment in the prior art are solved.

Description

Technical Field

[0001] The present invention relates to the field of positioning technology, and in particular, to a differential positioning method and device based on 5G communication signals. Background Art

[0002] Differential positioning technology based on 5G communication has attracted wide attention due to its potential to achieve high-precision positioning in theory. However, in practical applications, due to the use of omnidirectional linearly polarized antennas with small size, as well as low-cost and low-power GNSS chips, the quality and characteristics of mobile phone GNSS signals have been significantly negatively affected, and many problems have emerged especially in complex environments.

[0003] Traditional GNSS positioning relies on weak signals transmitted by satellites, and the omnidirectional linearly polarized antennas with small size have limitations in receiving signals. The gain of this kind of antenna is relatively low and cannot effectively collect satellite signals like large professional antennas. At the same time, although the omnidirectional design ensures that signals can be received in all directions, it also causes the signals to be scattered in all directions, further weakening the signal intensity. In addition, in order to reduce power consumption and cost, low-cost and low-power GNSS chips have made compromises in signal amplification and processing capabilities and cannot effectively enhance and recover weak satellite signals. Therefore, the intensity of GNSS signals received by mobile phones is far lower than the ideal state, which makes the signals more vulnerable to noise and interference during transmission, resulting in a decrease in positioning accuracy.

[0004] In complex environments, satellite signals will be severely blocked and reflected during propagation. The omnidirectional linearly polarized antenna of the mobile phone cannot effectively receive satellite signals in this environment, resulting in a sharp drop in signal intensity. At the same time, a large number of reflected signals will produce serious multipath effects, increasing the pseudorange multipath error. In addition, the electromagnetic environment in urban canyons is complex, with a large number of radio interference sources, such as base stations, communication equipment, etc. These interferences will further affect the quality of GNSS signals, resulting in poor continuity of carrier phase observations and frequent cycle slips. In this environment, it is very difficult for mobile phones to obtain stable high-precision position information, and the positioning error may reach several meters or even dozens of meters.

[0005] In complex environments, it is still an urgent problem to obtain stable high-precision position information using mobile phones. For this reason, we propose a differential positioning method and device based on 5G communication signals. Summary of the Invention

[0006] Based on the technical problems existing in the background art, the present invention proposes a differential positioning method and device based on 5G communication signals, which has the characteristics of enhancing the anti-interference ability of the system and being able to achieve centimeter-level or even millimeter-level positioning accuracy, and solves the problem that in this environment of the prior art, it is difficult for a mobile terminal to obtain stable high-precision position information, and the positioning error may reach several meters or even dozens of meters.

[0007] The present invention provides the following technical solutions: A differential positioning method based on 5G communication signals, comprising the following steps:

[0008] S1. Data preprocessing, based on a deep learning method, perform data preprocessing on the GNSS raw observations collected by the mobile terminal, extract and analyze the features of the GNSS raw observations to identify gross errors and cycle slips;

[0009] S2. Data fusion, after completing data preprocessing, fuse the sensor data including the inertial measurement unit for speed calculation, respectively perform pseudorange single-point positioning, SPV speed measurement and TDCP speed measurement, and use the particle filter algorithm to fuse different speed measurement results to obtain the optimal speed, providing reliable motion state parameters for subsequent high-precision positioning;

[0010] S3. Kalman filtering and differential positioning calculation, substitute the obtained optimal speed into the Kalman filter, and through the iterative optimization of the Kalman filter, realize the dynamic and accurate estimation of the position and speed of the mobile terminal;

[0011] S4. RTK solution reliability evaluation and ambiguity fixing, after filtering, evaluate the reliability of the RTK solution according to two key indicators of the effective number of phases and the position dilution of precision. If the RTK solution passes the reliability test, enter the ambiguity fixing process and output the high-precision RTK fixed solution to achieve real-time positioning; if it does not meet the reliability test, output the RTD solution as the positioning result.

[0012] Preferably, in the step S1, a convolutional neural network or a recurrent neural network is used to extract and analyze the features of the GNSS raw observations to achieve gross error rejection and cycle slip detection.

[0013] Preferably, in the step S2, the processing processes of the inertial measurement unit and the particle filter algorithm include initialization, particle filter iteration, and optimal speed selection;

[0014] In the initialization stage, use the inertial measurement unit to measure acceleration and angular velocity information and perform particle initialization;

[0015] In the particle filter iteration stage, according to the motion model of the inertial measurement unit and the particle state at the previous moment, predict the particle state at the current moment, and use the speed calculation result of GNSS as the observation value to update the weight of the particle;

[0016] In the optimal speed selection stage, after the iteration and resampling of the particle filter, the current speed is estimated based on the states and weights of the particles, and whether the estimated speed is the optimal value is judged according to a preset index.

[0017] Preferably, in step S3, the time update step is first executed to predict the state parameters and their error covariance at the next moment based on the system dynamics model. Subsequently, the pseudorange difference observation equation and the phase difference observation equation of the mobile terminal and the base station are respectively incorporated into the measurement update link: the pseudorange difference observation equation eliminates the common error by using the difference between the pseudorange observations of the base station and the mobile terminal, and quickly obtains the real-time kinematic difference solution; the phase difference observation equation obtains the real-time kinematic floating-point solution through the differential processing of the carrier phase observations.

[0018] Preferably, in step S3, the adaptive Kalman filtering algorithm is adopted to automatically adjust the parameters of the filter according to the real-time observation data and system state, model the pseudorange difference and phase difference observation equations, and introduce error sources including ionospheric delay and tropospheric delay.

[0019] Preferably, the modeling process of the pseudorange difference observation equation is as follows:

[0020] S31. For the pseudorange observation value ρ rs received by receiver r from satellite s, its observation equation is:

[0021] ρ rs =R rs +c(τ r -τ s )+ε ρrs ;

[0022] where R rs is the true distance from the satellite to the receiver, c is the speed of light, τ r and τ s are the clock biases of the receiver and the satellite respectively, and ε ρrs is the observation error including noise, etc.;

[0023] S32. For the ionospheric delay, dual-frequency observations are used for first-order correction; for the dual-frequency pseudorange observation values ρ r 1s and ρ r 2s, the pseudorange after the first-order ionospheric correction is:

[0024]

[0025] where f1 and f2 are two frequencies, and I0 is a parameter related to the ionospheric electron density; for more accurate modeling, considering the higher-order terms of the ionospheric delay, the corrected pseudorange is further expressed as:

[0026]

[0027] wherein, a i and b i are coefficients determined through experiments or theoretical analysis;

[0028] S33. For tropospheric delay, the Saastamoinen model is adopted for correction; in the Saastamoinen model, the tropospheric delay is expressed as T rs = T dry + T wet ; where T dry is the dry component and T wet is the wet component. The pseudorange after tropospheric delay correction is expressed as:

[0029] ρ″′ r1s = ρ″ r1s - T′ rs ;

[0030] ρ″′ r2s = ρ″ r2s - T′ rs ;

[0031] S34. For the pseudorange differential observation equation of two receivers r1 and r2 receiving the same satellite s, it is:

[0032]

[0033] After considering the above high-order terms such as ionospheric delay and tropospheric delay, the differential pseudorange observation equation becomes:

[0034]

[0035] Preferably, in step S4, the reliability of the RTK solution is evaluated based on two key indicators: one is that the number of effective phase counts needs to be greater than 5 to ensure sufficient observation data to support high-precision positioning; the other is that the position dilution of precision needs to be less than 5, indicating good satellite geometric distribution and guaranteed positioning accuracy;

[0036] If the above conditions are not met, the RTD solution is output as the positioning result; if the RTK solution passes the reliability test, it enters the ambiguity fixing process; the least squares ambiguity decorrelation adjustment method is used to fix the carrier phase ambiguity, the correlation between ambiguity parameters is reduced through decorrelation transformation, and the integer solution is searched by combining the least squares method; after fixing, the reliability of the fixed solution is evaluated through the LAMBDA test. If the test passes, a high-precision RTK fixed solution is output, thereby achieving accurate real-time positioning.

[0037] A differential positioning device based on 5G communication signals, comprising:

[0038] A data preprocessing module, which is used to preprocess the GNSS raw observations collected by the mobile terminal, extract and analyze the features of the GNSS raw observations to eliminate gross errors and detect cycle slips.

[0039] A data fusion module, which is used to fuse the sensor data including the inertial measurement unit for speed calculation, perform pseudorange single point positioning, SPV speed measurement, and TDCP speed measurement respectively, and fuse different speed measurement results using the particle filter algorithm to obtain the optimal speed.

[0040] A data optimization module, which is used to substitute the obtained optimal speed into the Kalman filter, and through the iterative optimization of the Kalman filter, realize the dynamic and accurate estimation of the position and speed of the mobile terminal.

[0041] A reliability evaluation module, which evaluates the reliability of the RTK solution based on two key indicators: the number of valid phase digits and the position dilution of precision. If the RTK solution passes the reliability test, it enters the ambiguity fixing process and outputs a high-precision RTK fixed solution, thereby realizing real-time positioning. If it does not meet the reliability test, it outputs the RTD solution as the positioning result.

[0042] The present invention provides a differential positioning method and device based on 5G communication signals. Through multi-sensor data fusion and particle filtering, even in complex environments such as areas with high-rise buildings in the city or indoor-outdoor transition zones, the reliability of positioning can be significantly improved, and the occurrence of positioning loss can be reduced. Even if the GNSS signal is briefly interrupted, it can continue to provide relatively accurate positioning information relying on the IMU data. Through the optimization of the Kalman filter, the accuracy of the positioning result is significantly improved. In an open environment, centimeter-level or even millimeter-level positioning accuracy can be achieved; in a complex environment, the positioning error can also be controlled within a small range. It can quickly respond to the movement changes of the mobile terminal and adjust the positioning result in a timely manner. Whether it is high-speed movement or frequent start-stop, the accuracy and real-time performance of positioning can be guaranteed. Brief Description of the Drawings

[0043] Figure 1 It is a flowchart of the method of the present invention.

[0044] Figure 2 It is a schematic diagram of the device of the present invention. Detailed Embodiments

[0045] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0046] As Figure 1 shown, the present invention provides a technical solution: a differential positioning method based on 5G communication signals, comprising the following steps:

[0047] S1. Data preprocessing: Based on a deep learning method, perform data preprocessing on the GNSS raw observations collected by the mobile terminal, extract and analyze the features of the GNSS raw observations to identify gross errors and cycle slips; by accurately identifying and removing gross errors and cycle slips, the influence of these abnormal data on subsequent positioning calculations is avoided, and the accumulation of errors is reduced. This makes the positioning result more accurate and reliable. Especially during a long-term positioning process, the positioning accuracy can be significantly improved. After removing the abnormal data, the stability of the GNSS signal is enhanced, and the continuity of the carrier phase observations is better. This helps to improve the success rate of integer ambiguity resolution and lays a foundation for high-precision positioning.

[0048] Use a convolutional neural network or a recurrent neural network to extract and analyze the features of the GNSS raw observations to achieve gross error removal and cycle slip detection. Traditional statistical test methods and algorithms such as polynomial fitting may have misjudgment or missed judgment in complex environments. The deep learning-based method can automatically learn the complex patterns and features in the data, improve the accuracy of gross error removal and cycle slip detection, thereby further improving the quality of the raw data and providing a more reliable data basis for subsequent positioning and velocity measurement.

[0049] Analysis process based on a convolutional neural network (CNN):

[0050] Data preprocessing: Organize the GNSS raw observations, usually including observation data such as pseudorange and carrier phase. Perform normalization processing on the data, map it to a specific interval, such as [0, 1] or [-1, 1], to accelerate model convergence and avoid the influence of too large or too small data on model training. At the same time, divide the data according to a certain time window or observation epoch to form a tensor format suitable for CNN input. For example, form a two-dimensional matrix by combining the observations of multiple consecutive epochs as the input image of CNN.

[0051] Convolutional layer: The core part of CNN is the convolutional layer. It performs convolution operations by sliding the convolution kernel on the input data to extract the local features of the data. For GNSS data, the convolution kernel can capture the local correlation of the data in time and space. For example, by using convolution kernels of different sizes, features of different scales can be extracted. Small convolution kernels may capture short-term signal changes, while large convolution kernels can capture longer-term trends. The output of the convolutional layer is a series of feature maps, and each feature map represents the response of the input data in different feature dimensions.

[0052] Pooling layer: The pooling layer usually follows the convolutional layer immediately and is used to downsample the feature map, reduce the data dimension, and at the same time retain the most important features. Common pooling methods include max pooling and average pooling. Max pooling selects the maximum value within each pooling window as the output, while average pooling takes the average value. The pooling operation can reduce the computational complexity of the model, prevent overfitting, and can further abstract and compress the features.

[0053] Fully connected layer: After feature extraction through multiple convolutional layers and pooling layers, the obtained feature map is flattened into a one-dimensional vector and then input into the fully connected layer. The neurons in the fully connected layer are connected to all neurons in the previous layer. It can synthesize and perform non-linear transformations on the extracted features to obtain the final feature representation. The fully connected layer usually contains multiple hidden layers, and each hidden layer uses an activation function (such as ReLU) to introduce non-linearity and enhance the expressive power of the model.

[0054] Output layer: The output layer is designed according to the specific task. If it is a gross error detection task, the output layer can be a binary classifier, using the Sigmoid activation function to output the probability that each observation is a gross error; if it is a cycle slip detection task, the output layer can be a multi-classifier, using the Softmax activation function to output the probability that each observation belongs to different cycle slip types or normal data. Finally, corresponding decisions are made based on the output results, such as determining observations with probabilities greater than a certain threshold as gross errors or cycle slips.

[0055] Analysis process based on Recurrent Neural Network (RNN):

[0056] Data preprocessing: Similar to CNN, first normalize the GNSS raw observations and arrange them in a time series. Since RNN is a model for processing sequence data, the data needs to be organized into a format suitable for RNN input. For example, the observations for each epoch are used as the input for one time step to form a sequence tensor.

[0057] Recurrent layer: The core of RNN is the recurrent layer, which contains multiple neurons. Each neuron receives the input at the current time step and the hidden state at the previous time step at each time step, and passes the hidden state to the next time step through a recurrent connection. In this way, RNN can utilize the temporal dependencies in the sequence data, remember and utilize historical information. For GNSS data, the recurrent layer can capture the changing trends and dynamic features of the observations over time. Common RNN variants include Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU), which can better handle long-term dependence problems and avoid gradient vanishing or explosion.

[0058] Hidden layer: The output of the recurrent layer can be used as the input of the hidden layer. The hidden layer usually contains multiple neurons, which are used to perform further non-linear transformations and combinations on the features extracted by the recurrent layer. Different activation functions (such as ReLU, tanh, etc.) can be used in the hidden layer to enhance the expressive power of the model and map the input to a more abstract feature space.

[0059] Output layer: Similar to CNN, the output layer is designed according to the specific task. For the tasks of gross error detection and cycle slip detection, the structure and activation function selection of the output layer are the same as those in the output layer of CNN, which are used to output the probabilities of gross errors or cycle slips respectively. Then, corresponding judgments and processing are carried out according to the output results.

[0060] S2. Data fusion: After completing data preprocessing, sensor data including the inertial measurement unit is fused to perform velocity solution. Pseudo-range single-point positioning, SPV velocity measurement, and TDCP velocity measurement are respectively executed. The particle filter algorithm is used to fuse different velocity measurement results to obtain the optimal velocity, providing reliable motion state parameters for subsequent high-precision positioning; the fused velocity information is more accurate and smooth, and can provide more reliable input for the Kalman filter. This makes the positioning trajectory more continuous and smooth, reducing the jumps and fluctuations of the positioning results.

[0061] The processing process of the inertial measurement unit and the particle filter algorithm, including initialization, particle filter iteration, and optimal velocity selection;

[0062] In the initialization stage, the inertial measurement unit is used to measure acceleration and angular velocity information and perform particle initialization;

[0063] In the particle filter iteration stage, according to the motion model of the inertial measurement unit and the particle state at the previous moment, the particle state at the current moment is predicted, and the velocity solution result of GNSS is used as the observation value to update the weights of the particles; after several iterations, the phenomenon of particle weight degradation may occur, that is, the weights of most particles become very small, and only a few particles have larger weights. To solve this problem, resampling operations are carried out. There are various resampling methods, such as roulette wheel method, systematic resampling method, etc. In the resampling process, particles with larger weights have a greater probability of being selected, while particles with smaller weights may be eliminated.

[0064] In the optimal speed selection stage, after the iteration and resampling of the particle filter, the current speed is estimated based on the states and weights of the particles, and whether the estimated speed is the optimal value is judged according to some preset metrics. For example, metrics such as the variance of the estimated speed, consistency with other sensors (such as odometers), the number of solved satellites, and the a posteriori residual can be considered. If the estimated speed meets the preset optimal conditions, it is output as the final optimal speed; otherwise, the iteration of the particle filter can be continued, or relevant parameters and models can be adjusted to further optimize the speed estimation.

[0065] By fusing the IMU and particle filter algorithms, the short-term high-precision measurement characteristics of the IMU and the good processing ability of the particle filter for nonlinear and non-Gaussian systems can be fully utilized to improve the accuracy and stability of speed calculation. Especially when the GNSS signal is interfered or blocked, it can better maintain the continuity and reliability of speed estimation.

[0066] S3. Kalman filtering and differential positioning solution: Substitute the obtained optimal speed into the Kalman filter, and through the iterative optimization of the Kalman filter, the dynamic and accurate estimation of the position and speed of the mobile terminal is realized. The iterative optimization process of the Kalman filter can achieve the real-time and dynamic estimation of the position and speed of the mobile terminal. During the movement, the system can timely track the position change of the terminal and provide accurate positioning results.

[0067] First, execute the time update step to predict the state parameters and their error covariance at the next moment based on the system dynamics model. Subsequently, incorporate the pseudorange differential observation equation and the phase differential observation equation of the mobile terminal and the base station into the measurement update link respectively: The pseudorange differential observation equation uses the difference between the pseudorange observation values of the base station and the mobile terminal to eliminate the common error and quickly obtain the real-time kinematic difference solution; the phase differential observation equation obtains the real-time kinematic floating solution through the differential processing of the carrier phase observation values.

[0068] By adopting the adaptive Kalman filtering algorithm, the parameters of the filter are automatically adjusted according to the real-time observation data and system state, the pseudorange differential and phase differential observation equations are modeled, and error sources including ionospheric delay and tropospheric delay are introduced.

[0069] The modeling process of the pseudorange differential observation equation is as follows:

[0070] S31. For the pseudorange observation value ρ of the receiver r receiving the satellite s rs , its observation equation is:

[0071] ρ rs =R rs +c(τ r -τ s )+ε ρrs ;

[0072] where R rs is the true distance from the satellite to the receiver, c is the speed of light, τ r and τ s are the clock biases of the receiver and the satellite respectively, and ε ρrs is the observation error including noise, etc.;

[0073] S32. For the ionospheric delay, dual-frequency observations are used for first-order correction; for the dual-frequency pseudorange observations ρ r1s and ρ r2s , the pseudorange after first-order ionospheric correction is:

[0074]

[0075] where f1 and f2 are two frequencies, and I0 is a parameter related to the ionospheric electron density; for more accurate modeling, considering the higher-order terms of the ionospheric delay, the corrected pseudorange is further expressed as:

[0076]

[0077]

[0078] where a i and b i are coefficients determined by experiments or theoretical analysis;

[0079] S33. For the tropospheric delay, the Saastamoinen model is used for correction; in the Saastamoinen model, the tropospheric delay is expressed as T rs =T dry +T wet ; where T dry is the dry component, and T wet is the wet component. The pseudorange after tropospheric delay correction is expressed as:

[0080] ρ″′ r1s =ρ″ r1s -T′ rs ;

[0081] ρ″′ r2s =ρ″ r2s -T′ rs ;

[0082] S34. For the pseudorange differential observation equation of two receivers r1 and r2 receiving the same satellite s:

[0083]

[0084] After considering the above higher-order terms such as ionospheric delay and tropospheric delay, the differential pseudorange observation equation becomes:

[0085]

[0086] Similarly, for the modeling of the pseudorange differential observation equation, the modeling of the pseudorange differential observation equation is similar to that of the phase differential observation equation. Through the more accurate modeling of the pseudorange differential and phase differential observation equations above, high-order terms such as ionospheric delay and tropospheric delay are considered, which can improve the positioning accuracy. Especially in high-precision positioning applications, it can better eliminate or weaken the influence of error sources on the observed values.

[0087] S4. RTK solution reliability evaluation and ambiguity fixing. After filtering, the reliability of the RTK solution is evaluated based on two key indicators: the number of valid phase measurements and the position dilution of precision (PDOP). If the RTK solution passes the reliability test, it enters the ambiguity fixing process to output a high-precision RTK fixed solution, thus achieving real-time positioning; if it does not meet the reliability test, an RTD solution is output as the positioning result. Evaluating the reliability of the RTK solution based on the two key indicators of the number of valid phase measurements and PDOP can comprehensively and objectively evaluate the quality of the positioning result. The number of valid phase measurements reflects the richness of the observed data, and PDOP measures the influence of satellite geometry distribution on the positioning accuracy. Through the comprehensive evaluation of these two indicators, the reliability of the RTK solution can be accurately judged.

[0088] The reliability of the RTK solution is evaluated based on two key indicators: one is that the number of valid phase measurements should be greater than 5 to ensure sufficient observed data to support high-precision positioning; the other is that the position dilution of precision should be less than 5, indicating good satellite geometry distribution and guaranteed positioning accuracy.

[0089] If the above conditions are not met, an RTD solution is output as the positioning result; if the RTK solution passes the reliability test, it enters the ambiguity fixing process; the least-squares ambiguity decorrelation adjustment method is used to fix the carrier phase ambiguity. The correlation between ambiguity parameters is reduced through decorrelation transformation, and the integer solution is searched by combining the least-squares method; after fixing, the reliability of the fixed solution is evaluated through the LAMBDA test. If the test passes, a high-precision RTK fixed solution is output, thus achieving accurate real-time positioning.

[0090] If the RTK solution passes the reliability test, a high-precision RTK fixed solution is output; if it does not meet the reliability test, an RTD solution is output as the positioning result. This flexible output method can select the most suitable positioning result according to the actual situation to ensure reliable positioning services in various environments.

[0091] After entering the ambiguity fixing process, an appropriate algorithm (such as the LAMBDA algorithm) is used for ambiguity fixing and reliability testing, which can improve the success rate of ambiguity fixing. Accurate ambiguity fixing is the key to achieving high-precision positioning. Through strict testing, positioning errors caused by incorrect ambiguity fixing can be avoided.

[0092] At the same time, more reliability evaluation indicators are added, such as considering factors such as the signal-to-noise ratio of the observed values and the satellite elevation angle. At the same time, the ambiguity fixing algorithm is improved. For example, an ambiguity search method based on genetic algorithm or simulated annealing algorithm is adopted and optimized in combination with the LAMBDA algorithm.

[0093] Relying solely on the number of valid phase measurements and PDOP as reliability evaluation indicators in the prior art may not be comprehensive enough. Adding indicators such as signal-to-noise ratio and satellite elevation angle can more comprehensively evaluate the quality of the observed data and the reliability of positioning. Improving the ambiguity fixing algorithm can increase the success rate and accuracy of ambiguity fixing. Especially under complex observation conditions, such as low signal-to-noise ratio and severe multipath effects, it can fix the ambiguity more quickly and accurately, thereby improving the accuracy and reliability of the RTK solution and reducing positioning errors caused by incorrect ambiguity fixing.

[0094] As Figure 2 shown, a differential positioning device based on 5G communication signals includes:

[0095] A data preprocessing module for preprocessing the GNSS raw observations collected by the mobile terminal, extracting and analyzing the characteristics of the GNSS raw observations to eliminate gross errors and detect cycle slips;

[0096] A data fusion module for fusing sensor data including an inertial measurement unit for speed calculation, performing pseudorange single-point positioning, SPV speed measurement, and TDCP speed measurement respectively, and using a particle filter algorithm to fuse different speed measurement results to obtain the optimal speed;

[0097] A data optimization module for substituting the obtained optimal speed into a Kalman filter, and through the iterative optimization of the Kalman filter, realizing the dynamic and accurate estimation of the position and speed of the mobile terminal;

[0098] A reliability evaluation module for evaluating the reliability of the RTK solution based on two key indicators, namely the number of valid phase measurements and the position dilution of precision. If the RTK solution passes the reliability test, it enters the ambiguity fixing process and outputs a high-precision RTK fixed solution to achieve real-time positioning; if it does not meet the reliability test, it outputs an RTD solution as the positioning result.

[0099] The deep learning method in the present invention has powerful feature learning capabilities and can extract deep and abstract features from complex GNSS raw observations. Compared with traditional statistical methods, deep learning models can automatically discover hidden patterns and regularities in data, accurately distinguish normal observations from outliers such as gross errors and cycle slips. Deep learning models can be trained with a large amount of data to adapt to different observation environments and signal characteristics. Whether in complex environments such as urban canyons and forests or under human interference, the model can maintain a high recognition accuracy.

[0100] Fusing sensor data such as that from an inertial measurement unit (IMU) can make full use of the complementary information of different sensors. GNSS positioning has high accuracy in open environments, but positioning inaccuracies occur in cases of severe signal blockage or interference; while the IMU can provide short-term high-precision motion information and is not affected by signal blockage. Through data fusion, when GNSS signals are poor, IMU data can be used for assisted positioning to improve the continuity and reliability of positioning. In complex environments such as areas with high-rise buildings in the city or indoor-outdoor transition zones, this solution can significantly improve the reliability of positioning and reduce the occurrence of positioning losses. Even when GNSS signals are briefly interrupted, the system can rely on IMU data to continue providing relatively accurate positioning information.

[0101] Through the optimization of the Kalman filter, the accuracy of the positioning result is significantly improved. In open environments, centimeter-level or even millimeter-level positioning accuracy can be achieved; in complex environments, the positioning error can also be controlled within a small range. The system can quickly respond to the movement changes of the mobile terminal and timely adjust the positioning result. Whether it is high-speed movement or frequent starts and stops, the accuracy and real-time nature of positioning can be guaranteed.

[0102] Through reliable evaluation and flexible output methods, the accuracy and reliability of the positioning result can be guaranteed. In complex environments, even if the RTK solution is unreliable, the RTD solution can provide relatively accurate positioning information to meet the needs of general applications. Accurate ambiguity fixing and reliability testing can improve the stability of positioning and reduce the fluctuations and errors of the positioning result. During a long-term positioning process, high positioning accuracy and reliability can be maintained.

[0103] The above is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention, according to the technical solution and inventive concept of the present invention, making equivalent replacements or changes, should be covered by the protection scope of the present invention.

Claims

1. A differential positioning method based on 5G communication signals, characterized in that: It includes the following steps: S1. Data preprocessing: Based on deep learning methods, perform data preprocessing on the GNSS raw observations collected by the mobile terminal, extract and analyze the features of the GNSS raw observations to identify gross errors and cycle slips; S2. Data fusion: After completing data preprocessing, fuse the sensor data including the inertial measurement unit to perform velocity solution, respectively execute pseudorange single-point positioning, SPV velocity measurement, and TDCP velocity measurement, and use the particle filter algorithm to fuse different velocity measurement results to obtain the optimal velocity, providing reliable motion state parameters for subsequent high-precision positioning; S3. Kalman filtering and differential positioning solution: Substitute the obtained optimal velocity into the Kalman filter, and through the iterative optimization of the Kalman filter, realize the dynamic and accurate estimation of the position and velocity of the mobile terminal; S4. RTK solution reliability evaluation and ambiguity fixing: After filtering, evaluate the reliability of the RTK solution according to the effective number of phases and the position dilution of precision. If the RTK solution passes the reliability test, enter the ambiguity fixing process and output the high-precision RTK fixed solution to achieve real-time positioning; if it does not meet the reliability test, output the RTD solution as the positioning result.

2. The differential positioning method based on 5G communication signals according to claim 1, characterized in that: In the step S1, use a convolutional neural network or a recurrent neural network to extract and analyze the features of the GNSS raw observations to achieve gross error rejection and cycle slip detection.

3. A differential positioning method based on 5G communication signals according to claim 1, characterized in that: In the step S2, the processing process of the inertial measurement unit and the particle filter algorithm includes initialization, particle filter iteration, and optimal velocity selection; In the initialization stage, use the inertial measurement unit to measure acceleration and angular velocity information and perform particle initialization; In the particle filter iteration stage, according to the motion model of the inertial measurement unit and the particle state at the previous moment, predict the particle state at the current moment, and use the velocity solution result of GNSS as the observation value to update the particle weights; In the optimal velocity selection stage, after the particle filter iteration and resampling, estimate the current velocity according to the particle state and weight, and judge whether the estimated velocity is the optimal value according to the preset index.

4. The differential positioning method based on 5G communication signals according to claim 3, wherein: In the step S3, first execute the time update step, predict the state parameters and their error covariance at the next moment based on the system dynamics model. Subsequently, incorporate the pseudorange differential observation equation and the phase differential observation equation of the mobile terminal and the base station into the measurement update link respectively: the pseudorange differential observation equation uses the difference between the pseudorange observations of the base station and the mobile terminal to eliminate the common error and quickly obtain the real-time kinematic differential solution; the phase differential observation equation obtains the real-time kinematic floating-point solution through the differential processing of the carrier phase observations.

5. A differential positioning method based on 5G communication signals according to claim 4, characterized in that: In the step S3, adopt an adaptive Kalman filter algorithm, automatically adjust the filter parameters according to the real-time observation data and system state, model the pseudorange differential and phase differential observation equations, and introduce error sources including ionospheric delay and tropospheric delay.

6. The differential positioning method based on 5G communication signals according to claim 5, characterized in that: The modeling process of the pseudorange differential observation equation is as follows: S31. For the pseudorange observation value ρ of satellite s received by receiver r rs , its observation equation is: ρ rs = R rs + c(τ r - τ s ) + ε ρrs ; where R rs is the true distance from the satellite to the receiver, c is the speed of light, τ r and τ s are the clock biases of the receiver and the satellite respectively, and ε ρrs is the observation error including noise, etc.; S32. For ionospheric delay, dual-frequency observations are used for first-order correction; for dual-frequency pseudorange observations ρ r1s and ρ r2s , the pseudorange after first-order ionospheric correction is as follows: where f1 and f2 are two frequencies, I0 is a parameter related to the ionospheric electron density; for more accurate modeling, considering the higher-order terms of the ionospheric delay, the corrected pseudorange is further expressed as: where a i and b i are coefficients determined by experiments or theoretical analysis; S33. For tropospheric delay, the Saastamoinen model is used for correction; in the Saastamoinen model, the tropospheric delay is expressed as T rs = T dry + T wet ; where T dry is the dry component, and T wet is the wet component. The pseudorange after tropospheric delay correction is expressed as: ρ″′ r1s = ρ″ r1s - T′ rs ; ρ″′ r2s = ρ″ r2s - T′ rs ; S34. For the pseudorange differential observation equations of two receivers r1 and r2 receiving the same satellite s, they are as follows: After considering high-order terms such as the above ionospheric delay and tropospheric delay, the differential pseudorange observation equation becomes:

7. A differential positioning method based on 5G communication signals according to claim 1, characterized in that: In step S4, the reliability of the RTK solution is evaluated based on the effective number of phases and the position dilution of precision. If the conditions are not met, the RTD solution is output as the positioning result; if the RTK solution passes the reliability test, it enters the ambiguity fixing process; the least-squares ambiguity decorrelation adjustment method is used to fix the carrier phase ambiguity, the correlation between ambiguity parameters is reduced through decorrelation transformation, and the integer solution is searched by combining the least-squares method; after fixing, the reliability of the fixed solution is evaluated through the LAMBDA test. If the test passes, the high-precision RTK fixed solution is output.

8. A differential positioning device based on 5G communication signals, comprising: A data preprocessing module for preprocessing the GNSS raw observations collected by the mobile terminal, extracting and analyzing the features of the GNSS raw observations to eliminate gross errors and detect cycle slips; A data fusion module for fusing sensor data including an inertial measurement unit to perform speed calculation, respectively performing pseudorange single-point positioning, SPV speed measurement, and TDCP speed measurement, and using the particle filter algorithm to fuse different speed measurement results to obtain the optimal speed; A data optimization module for substituting the obtained optimal speed into the Kalman filter, and through the iterative optimization of the Kalman filter, realizing the dynamic and accurate estimation of the position and speed of the mobile terminal; A reliability evaluation module for evaluating the reliability of the RTK solution based on the effective number of phases and the position dilution of precision. If the RTK solution passes the reliability test, it enters the ambiguity fixing process and outputs the high-precision RTK fixed solution to achieve real-time positioning; if the reliability test is not satisfied, the RTD solution is output as the positioning result.

Images