A high-precision positioning method and system in a tunnel without GNSS

Abstract

The application discloses a high-precision positioning method and system in a tunnel under a GNSS-free environment, and relates to the technical field of vehicle positioning. The application comprises the following steps: extracting inertial measurement disturbance characteristics by using a vibration mode decomposition method, and determining a wheel slip state according to rotation pulse data output by a wheel distance measuring sensor; decoupling the inertial measurement disturbance characteristics by using a frequency response suspension vibration transmission model based on suspension damping coefficient and suspension stiffness coefficient fitting, and determining suspension displacement; compensating the rotation pulse data by using a nonlinear mapping relationship between the wheel slip rate and the suspension displacement, and determining a vehicle driving distance; inputting the vehicle driving distance and the suspension displacement into an extended state space unscented Kalman filter model, and calculating automobile positioning coordinates; and the application effectively suppresses inertial navigation disturbance and wheel slip error, and realizes high-precision autonomous positioning of an automobile in a tunnel under a GNSS-free environment.

Description

Technical Field

[0001] This invention relates to the field of vehicle positioning technology, specifically to a high-precision positioning method and system in tunnels without GNSS. Background Technology

[0002] With the development of transportation infrastructure, tunnels, as an important component of highways, place high demands on vehicle positioning accuracy for safe and efficient operation. However, inside tunnels, satellite navigation signals have difficulty penetrating effectively due to the tunnel structure's obstruction, making it impossible for existing Global Navigation Satellite Systems (GNSS) to provide vehicles with continuous and reliable positioning information. To overcome these problems, a combination of inertial navigation and wheel ranging technologies is currently widely used to achieve autonomous positioning within tunnels.

[0003] However, in practical applications, the measurement data output by inertial navigation systems is easily affected by vehicle vibrations caused by road surface excitation, making it difficult to accurately separate the effective measurement information that truly reflects the vehicle's motion state, resulting in a rapid accumulation of positioning errors. Simultaneously, wheel distance sensors are susceptible to wheel slippage under different road conditions, significantly increasing the calculation error of vehicle distance and further reducing positioning accuracy. Furthermore, existing data fusion methods, when dealing with the multi-degree-of-freedom motion of vehicles, typically fail to effectively consider the dynamic constraints of the vehicle's longitudinal, lateral, and vertical motions, making it difficult to achieve high-precision positioning estimation and limiting their reliable application in real-world environments.

[0004] Therefore, in the context of tunnel environments without GNSS signals, how to reasonably and effectively eliminate measurement errors caused by inertial measurement disturbances and wheel slippage, and fully integrate the multi-degree-of-freedom motion characteristics of vehicles to achieve high-precision autonomous vehicle positioning, has become an urgent technical problem to be solved. Summary of the Invention

[0005] The purpose of this invention is to provide a high-precision positioning method and system in tunnels without GNSS, so as to solve the problems in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: In a first aspect, the present invention provides a high-precision positioning method in a tunnel in a GNSS-free environment, characterized in that it includes: Based on the triaxial acceleration and triaxial angular velocity output by the inertial navigation unit, the inertial measurement disturbance characteristics are extracted using vibration mode decomposition, and the wheel slip state is determined based on the rotation pulse data output by the wheel distance sensor. The inertial measurement disturbance characteristics are decoupled using a frequency response suspension vibration transmission model fitted based on the suspension damping coefficient and suspension stiffness coefficient, and the suspension displacement is determined. Based on the wheel slip state, the nonlinear mapping relationship between wheel slip ratio and suspension displacement is used to compensate for the rotation pulse data and determine the vehicle travel distance. The vehicle's travel distance and suspension displacement are input into an extended state-space unscented Kalman filter model that integrates longitudinal velocity constraints, lateral sideslip constraints, and vertical vibration displacement constraints to calculate the vehicle's positioning coordinates.

[0007] Secondly, the present invention provides a high-precision positioning system in a tunnel without GNSS, implemented based on the aforementioned high-precision positioning method in a tunnel without GNSS, including: The data acquisition module is used to extract inertial measurement disturbance features based on the triaxial acceleration and triaxial angular velocity output by the inertial navigation unit using vibration mode decomposition, and to determine the wheel slip state based on the rotation pulse data output by the wheel distance sensor. The feature decoupling module is used to decouple the inertial measurement disturbance features by using a frequency response suspension vibration transmission model fitted based on the suspension damping coefficient and suspension stiffness coefficient, and to determine the suspension displacement. The pulse compensation module is used to compensate for the rotational pulse data based on the wheel slip state and the nonlinear mapping relationship between the wheel slip ratio and the suspension displacement, so as to determine the vehicle travel distance. The positioning module is used to calculate the vehicle's positioning coordinates by integrating the vehicle's driving distance and suspension displacement inputs with an extended state-space unscented Kalman filter model that incorporates vehicle longitudinal velocity constraints, lateral sideslip constraints, and vertical vibration displacement constraints.

[0008] The technical effects and advantages provided by the present invention in the above technical solution are as follows: This invention extracts inertial measurement disturbance features by employing vibration mode decomposition, effectively separating suspension vibration interference caused by road surface excitation from the acceleration and angular velocity data output by the inertial navigation unit. This reduces disturbance errors in inertial measurement data, improves the effectiveness and accuracy of inertial navigation measurement data, and lays the foundation for high-precision calculation of vehicle displacement.

[0009] This invention constructs a suspension vibration transmission model and uses this model to decouple the inertial measurement disturbance characteristics, effectively obtaining suspension displacement information that truly reflects the vertical motion characteristics of the vehicle. This accurately captures the true characteristics of road surface excitation, reduces vehicle displacement observation errors, and provides reliable data support for the accurate calculation of vehicle travel distance.

[0010] This invention establishes a nonlinear mapping relationship between wheel slip ratio and suspension displacement, and effectively compensates for rotational pulse data based on this mapping relationship. This significantly reduces the adverse effects of wheel slip on vehicle distance calculation, improves the reliability of vehicle distance measurement results, and further combines an extended state-space unscented Kalman filter model with multi-degree-of-freedom motion constraints to ultimately achieve high-precision autonomous vehicle positioning in tunnels without GNSS. Attached Figure Description

[0011] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0012] Figure 1 This is a flowchart of a high-precision positioning method in a tunnel without GNSS according to the present invention; Figure 2 This is a framework diagram of a high-precision positioning system for tunnels in GNSS-free environments according to the present invention. Detailed Implementation

[0013] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, they are provided to make the description of this application more complete and comprehensive, and to fully convey the concept of the exemplary embodiments to those skilled in the art. The drawings are merely illustrative illustrations of this application and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and therefore repeated descriptions of them will be omitted.

[0014] Furthermore, the described features, structures, or characteristics can be combined in any suitable manner in one or more exemplary embodiments. Numerous specific details are provided in the following description to give a full understanding of the exemplary embodiments disclosed in this application. However, those skilled in the art will recognize that the technical solutions disclosed in this application can be practiced with one or more specific details omitted, or other methods, components, steps, etc., can be employed. In other instances, well-known structures, methods, implementations, or operations are not shown or described in detail to avoid obscuring various aspects of the disclosure of this application.

[0015] Example 1 like Figure 1 As shown, this embodiment discloses a high-precision positioning method in a tunnel in a GNSS-free environment, including: S101: Based on the triaxial acceleration and triaxial angular velocity output by the inertial navigation unit, the inertial measurement disturbance characteristics are extracted using vibration mode decomposition, and the wheel slip state is determined based on the rotation pulse data output by the wheel distance sensor. In specific implementations, the inertial navigation unit is preferably a high-precision MEMS inertial navigation device, rigidly mounted at the vehicle's center of gravity. This inertial navigation unit includes at least a three-axis MEMS accelerometer, a three-axis MEMS gyroscope, and a data acquisition and processing module, used to output real-time three-axis acceleration and three-axis angular velocity data of the vehicle while driving in the tunnel. This data constitutes the basic dataset characterizing the vehicle's motion state and its response to road surface excitation.

[0016] Furthermore, the wheel distance sensor is installed on the vehicle's drive wheel axle or driven wheel axle, specifically employing a high-precision photoelectric rotary encoder to collect wheel rotation pulse signals in real time. These rotation pulse signals reflect the wheel's rotational motion state and serve as the basis for subsequent vehicle distance estimation.

[0017] Specifically, the extraction of inertial measurement disturbance features using vibration mode decomposition includes: Based on the original vibration signals of the inertial navigation unit collected when the vehicle passes through the tunnel entrance section, the target frequency band closely related to the suspension response is determined; In practice, when a vehicle enters the tunnel entrance area, the inertial navigation unit initiates high-frequency sampling (sampling rate set to, for example, 200Hz) to collect real-time acceleration and angular velocity time-series data in the longitudinal (X-axis), lateral (Y-axis), and vertical (Z-axis) directions of the vehicle. The processing module uses a vector synthesis algorithm to fuse the three-axis accelerations to obtain a comprehensive inertial vibration signal characterizing the overall excitation state of the vehicle.

[0018] Furthermore, the comprehensive inertial vibration signal is subjected to frequency domain transformation processing, and its spectral distribution is analyzed using Fast Fourier Transform (FFT). The specific calculation expression is as follows:

[0019] In the formula: Represents the amplitude of spectral lines in the frequency domain. The amplitude of the original vibration signal at discrete time points is represented by N, which is the total number of sampling points within the sampling window.

[0020] The power spectral density distribution of the vehicle suspension system's vibration response is obtained through spectral analysis, and the energy proportion of different frequency bands is calculated. A continuous frequency range with an energy proportion greater than a preset threshold (e.g., 80% of the total energy) is selected as the target frequency band. For example, in one implementation, analysis determined that the 5Hz to 25Hz frequency band concentrated the main energy of the vehicle suspension response; therefore, 5Hz to 25Hz was determined as the target frequency band for subsequent refined modal analysis.

[0021] Within the target frequency band, a vibration mode screening algorithm based on the local concentration of modal energy is used to extract effective vibration modes associated with the suspension response; The extraction of effective vibration modes using a vibration mode screening algorithm based on the local concentration of modal energy includes: The target frequency band vibration signal is decomposed into multiple sub-frequency band signals, and the energy distribution of each sub-frequency band signal is determined. In the specific implementation process, the variational mode decomposition (VMD) algorithm is used to non-recursively decompose the vibration signal within the target frequency band into K intrinsic mode functions (IMFs) with different center frequencies. For each IMF component obtained from the decomposition, the corresponding sub-frequency band signal energy value is calculated using the following formula: In the formula, This represents the energy value of the j-th intrinsic mode function (IMF). Let N be the amplitude of the j-th IMF at discrete time point n, and N be the signal length.

[0022] Identify characteristic sub-frequency bands that are highly correlated with suspension structure characteristics based on the energy concentration of each sub-frequency band; In practice, the energy concentration index is defined as the energy value of a single sub-band. Total energy of all sub-bands The ratio. Set an energy concentration threshold (e.g., 20%). When the energy concentration index of a sub-band is greater than or equal to this threshold, it is marked as a characteristic sub-band.

[0023] For example, if five IMF components are obtained from the decomposition, and the energy proportion of the second IMF component is 45%, which is significantly higher than that of the other components, then the frequency band corresponding to the second IMF component is determined to be the characteristic sub-frequency band.

[0024] Vibration mode clustering is performed on characteristic sub-frequency bands to obtain effective vibration modes associated with suspension response; In practice, the K-means clustering algorithm is used to classify and aggregate the selected feature sub-bands. A two-dimensional feature vector containing "center frequency" and "energy density" is constructed, and feature sub-bands with close distances are grouped into the same mode cluster. After clustering, several mode clusters with stable cluster centers and significant energy are output and defined as effective vibration modes. For example, cluster analysis yielded two effective vibration mode clusters with center frequencies of 8Hz and 13Hz, respectively, which correspond to the sprung mass mode and unsprung mass mode of the suspension.

[0025] Inertial measurement perturbation characteristics are constructed using the effective vibration modes to obtain inertial measurement perturbation characteristics limited by the local concentration of modal energy; In specific implementation, the key parameters of each effective vibration mode cluster are extracted and combined to form an inertial measurement disturbance feature vector. The feature vector includes: (1) Modal energy value: the energy amplitude of each effective vibration mode cluster; (2) Modal center frequency: the frequency center of each effective vibration mode cluster; (3) Modal distribution divergence: characterizing the degree of dispersion of modal energy in the frequency domain.

[0026] For example, in a specific implementation, the obtained inertial measurement disturbance characteristics are: modal center frequency = [8Hz, 13Hz], modal energy value = [350J, 420J].

[0027] Furthermore, determining the wheel slip state based on the rotational pulse data output by the wheel distance sensor is achieved through the physical mapping relationship between the inertial measurement disturbance characteristics and the road surface roughness and adhesion coefficient. Since road surface roughness directly affects the vibration energy of the vehicle suspension and also determines the adhesion limit between the tire and the road surface, the wheel slip trend can be indirectly inferred through vibration characteristics.

[0028] The specific discrimination logic is as follows: Calculate the total energy of all valid vibration mode families. Preset slip state energy threshold range ,in, This is the low slip limit. This represents the high slip limit.

[0029] 1) If This is determined to be a low slip condition, indicating that the road surface is smooth and the wheel is in a stable rolling area; (2) If This is determined to be a medium slip state, characterized by obvious joints or rough textures on the road surface and slight slippage of the wheels. (3) If This condition is classified as high slippage, indicating that the road surface is severely bumpy or slippery, and there is a significant risk of wheel slippage.

[0030] For example, setting If the current calculation yields... If this is determined, the current wheel slip state of the vehicle is identified as mid-slip. This determined state result will be passed as a control command to subsequent steps for selecting a matching nonlinear compensation model.

[0031] S102: Decouple the inertial measurement disturbance characteristics by using a frequency response suspension vibration transmission model fitted based on the suspension damping coefficient and suspension stiffness coefficient, and determine the suspension displacement. In this step, to obtain the true vertical displacement response of the vehicle suspension system in a tunnel environment (which reflects the geometric characteristics of road surface excitation), a vibration transmission model that accurately reflects the mechanical properties of the suspension needs to be constructed. This embodiment equates the vehicle suspension system to a two-degree-of-freedom spring-damped-mass mechanical system. The initial transfer function is derived using theoretical vibration equations, and key parameters are identified and corrected based on measured data. Finally, the corrected model is used to dynamically decouple the inertial signal.

[0032] Specifically, the method of decoupling inertial measurement disturbance characteristics using a frequency response suspension vibration transmission model fitted based on suspension damping coefficient and suspension stiffness coefficient includes: Based on the theoretical vibration equation of the suspension structure, the initial vibration transfer function is established; In practical implementation, based on the principles of vehicle dynamics, a coupled vibration differential equation is established for the sprung mass (vehicle body) and the unsprung mass (wheels and axles). Definition For the sprung mass, Unsprung mass; These are the vertical displacements (relative to the static equilibrium position) of the sprung mass and the unsprung mass, respectively. This refers to the suspension damping coefficient; This is the suspension stiffness coefficient; This refers to the tire stiffness coefficient. The road surface is given a vertical excitation. The two-degree-of-freedom dynamic equations are established as follows:

[0033]

[0034] By performing a Laplace transform on the above equations, the road surface excitation is derived. Vertical displacement of the sprung mass initial vibration transfer function : .

[0035] This transfer function describes the frequency response relationship between the road excitation input and the sprung mass displacement of the vehicle, providing a basic physical model for subsequent decoupling of inertial disturbance characteristics.

[0036] Collect suspension displacement response samples of a vehicle when it passes through a road surface joint at a typical tunnel entrance section; In practical implementation, in order to address the unknown or uncertain parameters in the above theoretical model (mainly the suspension damping coefficient) and stiffness coefficient Calibration requires collecting measured data under known operating conditions.

[0037] A typical tunnel entrance road surface joint area was selected as the test scenario, and high-precision displacement sensors (such as linear variable differential transformers (LVDTs) were installed at key nodes of the vehicle suspension. As the vehicle passed through the joint at a predetermined speed, the actual vertical displacement data of the suspension was simultaneously collected, forming a suspension displacement response sample sequence. This sample sequence contains the true dynamic response characteristics of the vehicle suspension system to step or pulse excitations.

[0038] The initial vibration transfer function is modified by using suspension displacement response samples to obtain a suspension vibration transfer model defined by the displacement response samples. The step of correcting the parameters of the initial vibration transfer function using suspension displacement response samples includes: Time-frequency analysis was performed on suspension displacement response samples collected from typical tunnel entrance sections to determine the time-frequency characteristics of the suspension displacement response; In practice, the short-time Fourier transform (STFT) is used to process the collected suspension displacement response samples to obtain the joint distribution characteristics of the signal in the time and frequency domains. Through time-frequency analysis, the main frequency range (i.e., natural frequency) where the suspension system resonates and the time constant of vibration decay are identified.

[0039] Based on the time-frequency characteristics of the suspension displacement response, a parametric sensitivity model of the initial vibration transfer function is constructed; Specifically, the effect of the transfer function amplitude-frequency characteristic on the suspension damping coefficient is calculated. and stiffness coefficient Using the partial derivatives, establish a parameter sensitivity model:

[0040] Sensitivity analysis quantifies the impact of each parameter on the model's output accuracy, thereby determining the priority of parameter correction and the search space.

[0041] The optimal correction range for suspension damping coefficient and stiffness coefficient is determined by using a parametric sensitivity model. In practice, a sensitivity threshold (e.g., 0.1) is set to filter out highly sensitive parameters. Based on the vehicle's factory specifications and the results of sensitivity analysis, constraint boundaries for parameter optimization are set. For example, the suspension stiffness coefficient is determined. The correction range is the nominal value Suspension damping coefficient The correction range is the nominal value .

[0042] The initial vibration transfer function is updated based on the optimal correction range to obtain the suspension vibration transfer model defined by the displacement response samples; In specific implementation, nonlinear least squares method or particle swarm optimization (PSO) algorithm is used, with minimizing the mean square error between the measured displacement response samples and the theoretical output of the model as the objective function, and within the optimal correction range... and Perform iterative optimization.

[0043] When the objective function converges, the optimal parameter set is obtained. Substituting this into the initial vibration transfer function, we obtain the modified suspension vibration transfer model. .

[0044] The suspension displacement is obtained by decoupling the inertial measurement disturbance characteristics using a suspension vibration transmission model defined by displacement response samples. In practice, this step uses the "effective vibration mode" (containing a specific center frequency and frequency band range) determined in S101 as the filtering basis. Combined with the dynamic model corrected in this step, the suspension displacement is decoupled from the vertical acceleration signal output by the inertial navigation unit. The specific process is as follows: Signal Reconstruction and Filtering: Based on the center frequency and bandwidth of the effective vibration modes output by S101, a bandpass filter is designed to process the raw vertical acceleration signal acquired by the inertial navigation unit. Filtering is performed to extract the suspension response component caused by road surface excitation, which is recorded as the effective acceleration signal. .

[0045] Constructing an inverse system decoupling model: Based on the modified transfer model (Characterizing the transmission relationship from road surface excitation to suspension displacement), and combining the physical differential properties of acceleration and displacement, an inverse system transfer function is established from effective acceleration input to suspension displacement output. :

[0046] In the formula: The transfer function from road surface excitation to vehicle body acceleration (in theory) Using this inverse system model, the corresponding displacement component can be extracted from the acceleration signal. In practical digital signal processing, this inverse system model behaves as an integral filter with specific frequency response characteristics.

[0047] Displacement calculation: The effective acceleration signal Input the inverse system decoupling model, perform convolution operations in the time or frequency domain, and output the true vertical displacement of the vehicle suspension. .

[0048] Through the above steps, this step successfully separated the suspension displacement signal reflecting the road surface geometry from the inertial measurement data. This suspension displacement... This data will be used as key data and simultaneously transmitted to S103 for quantitative calculation of wheel slip ratio, and to S104 as vertical observation constraints for the Kalman filter.

[0049] S103: Based on the wheel slip state, the nonlinear mapping relationship between wheel slip ratio and suspension displacement is used to compensate for the rotation pulse data and determine the vehicle travel distance; In practical implementation, this step aims to solve the problem of inaccurate mileage calculation caused by wheel slippage due to wet or bumpy road surfaces in GNSS-free environments. This invention innovatively constructs a graded compensation strategy: using the "wheel slippage state" (low, medium, high) determined in S101 as the model selection criterion, and using the "suspension displacement" obtained through decoupling in S102 as a quantitative input reflecting the road surface excitation intensity, the real-time slippage rate is calculated through a preset nonlinear regression model, thereby correcting the wheel rotation pulse data.

[0050] Specifically, the method of compensating for rotational pulse data using the nonlinear mapping relationship between wheel slip ratio and suspension displacement includes: Construct a sample set that correlates suspension displacement features with wheel slip ratio under different wheel slip states; In practice, during vehicle calibration or offline training, driving data is collected under various typical road conditions (including dry, wet, flooded, and gravel conditions). For each data sample collected, the following processing is performed: State marking: Using the method described in S101, calculate the vibration energy index of the inertial measurement disturbance characteristics. Based on this, the samples are labeled as "low slip state", "medium slip state" or "high slip state".

[0051] Feature extraction: For the same time window, extract the suspension displacement signal obtained by decoupling using the method described in S102. The statistical characteristic values ​​are calculated. This embodiment constructs a multidimensional feature vector as the input feature. The feature vector includes at least: the root mean square value (RMS) of the suspension displacement (characterizing the geometric excitation intensity of the road surface) and the statistical feature value of the vehicle's longitudinal acceleration (characterizing the driving / braking intensity).

[0052] Truth value correlation: The true slip rate measured by a high-precision reference device (such as a differential GNSS / INS integrated navigation system) is used as the output label. .

[0053] Through the above processing, three independent association sample sets were constructed: low-slip association sample set. Mid-glide correlation sample set and high-slip correlation sample set .

[0054] Using the support vector regression algorithm, nonlinear mapping relationships under specific states are constructed based on each associated sample set; This embodiment uses the Support Vector Regression (SVR) algorithm to train regression models for the three sample sets mentioned above, in order to establish an accurate mapping between "suspension displacement characteristics → wheel slip ratio".

[0055] Taking the construction of a nonlinear mapping relationship under the "medium-slip state" as an example, the specific steps are as follows: Determine the kernel function type; considering the complex nonlinear coupling relationship between suspension displacement and slip ratio, the radial basis function (RBF) is selected as the kernel function, and its definition is:

[0056] In the formula: The input is the suspension displacement feature vector. The kernel width parameter is determined through cross-validation to find the optimal value.

[0057] Calculate the support vector parameters; construct the following optimization objective function and solve for the SVR model parameters:

[0058] The constraints are satisfied:

[0059]

[0060] In the formula: The penalty coefficient is... For insensitive loss parameters, These are slack variables. The Lagrange multipliers are obtained by solving a quadratic programming problem. and bias .

[0061] Establish a nonlinear mapping expression; based on the solution results, generate a slip ratio prediction function under the "medium slip state". :

[0062] Similarly, using low-slip sample sets The model is obtained through training Using high-slip sample sets The model is obtained through training Generally, under the same suspension displacement input, the slip ratio output by the high-slip model is greater than that of the low-slip model.

[0063] Calculate the real-time wheel slip ratio based on the real-time determined wheel slip state and suspension displacement; During the real-time operation of the vehicle, the following logical steps are executed: Input Acquisition: Receives the real-time "wheel slip status" flag output by S101 and the real-time suspension displacement data output by S102. .

[0064] Feature calculation: A sliding window process is applied to the suspension displacement data, and the root mean square value within the current window is calculated. .

[0065] Model invocation and prediction: If the flag is in "low glide state", call Calculate the predicted slip ratio ; If the flag is in "middle-slide state", call Calculate the predicted slip ratio ; If the flag is in "high glide state", call Calculate the predicted slip ratio .

[0066] The vehicle travel distance is obtained by compensating the rotational pulse data with the predicted wheel slip ratio. Based on the physical definition of wheel slip ratio, a travel distance compensation formula is established.

[0067] In practice, the measured travel distance is first obtained by multiplying the number of rotation pulses output by the wheel distance sensor by the wheel circumference. Subsequently, the calculated predicted slip ratio was used. (Values ​​are decimals), calculate the actual vehicle travel distance after compensation. The calculation formula is: ).

[0068] For example, suppose in a certain implementation, S101 determines that the current vehicle is in a "medium slip state" (corresponding to the wet road surface joint area). The root mean square value of the suspension displacement output by S102 is 12mm. This is then input... The model outputs a predicted slip ratio. (i.e., 2.5%). If the original distance measured by the rotating pulse during this period... The actual vehicle travel distance after compensation is: .

[0069] Through the above steps, the present invention effectively eliminates wheel slippage errors caused by changes in road conditions. The calculated actual vehicle travel distance... It will serve as a high-precision longitudinal position observation benchmark, directly transmitted to S104, and used to construct the velocity observation equations for the Kalman filter.

[0070] S104: The extended state-space unscented Kalman filter model, which integrates vehicle travel distance and suspension displacement inputs with vehicle longitudinal velocity constraints, lateral sideslip constraints, and vertical vibration displacement constraints, is used to calculate the vehicle positioning coordinates. In the specific implementation process, in order to achieve high accuracy and stability of vehicle positioning in tunnels without GNSS, this step constructs an extended state-space unscented Kalman filter (UKF) model that integrates multiple dynamic constraints. This step uses the "actual vehicle travel distance" obtained in step S103 after accurate compensation of wheel slip ratio to derive the longitudinal velocity observation value, and uses the "suspension displacement" obtained in step S102 as the vertical vibration constraint observation value to achieve accurate estimation of the vehicle's three-dimensional positioning coordinates.

[0071] Specifically, the calculation of vehicle positioning coordinates includes: Based on the vehicle dynamics equations, kinematic constraint expressions for the vehicle's longitudinal velocity, lateral sideslip, and vertical vibration displacement are established respectively. In practical implementation, vehicle motion is abstracted as a three-degree-of-freedom rigid body dynamics model. To accurately describe the complex motion of the vehicle within the tunnel, this embodiment establishes velocity update equations in the body frame (which moves with the vehicle) and position update equations in the navigation frame (usually defined as the northeast-sky coordinate system or the tunnel local coordinate system). The specific dynamic constraint equations are as follows: (1) Longitudinal and lateral velocity constraint equations (including Coriolis force terms):

[0072]

[0073] (2) Vertical direction motion constraint equations:

[0074] (3) Position update equation (from vehicle coordinate system to navigation coordinate system):

[0075]

[0076]

[0077] In the formula: These represent the longitudinal, lateral, and vertical velocities of the vehicle in the vehicle coordinate system, respectively. These represent the three-axis accelerations measured by the inertial navigation unit (with gravity components and zero bias removed); This represents the yaw rate of the vehicle about its vertical axis. This indicates the vehicle's three-dimensional position in the navigation coordinate system; Indicates the vehicle's yaw angle.

[0078] Based on the kinematic constraint expression, and combined with the vehicle travel distance and suspension displacement, a system observation expression in the extended state space is constructed; In practical implementation, the state vector of the extended state space is first defined. This vector contains the vehicle's three-dimensional position, three-dimensional velocity, and heading angle:

[0079] Secondly, construct the system's observation vectors. To address the location update lag and path dependency issues caused by directly using accumulated distance as an observation, this embodiment adopts the following observation construction strategy: Longitudinal velocity observation construction: The "compensated actual vehicle travel distance" output in step S103 "Time difference processing is performed to obtain high-precision longitudinal velocity observations." The calculation formula is: In the formula, This represents the sampling time interval.

[0080] Vertical constraint observation construction: Given that the long-period average vertical velocity of a vehicle traveling inside a tunnel approaches zero, a virtual vertical velocity observation value of 0 is introduced. At this time, the suspension displacement obtained by decoupling S102... This will serve as the basis for assessing the credibility of the "vertical zero velocity hypothesis" and will be used for subsequent adaptive adjustment of observation noise.

[0081] In summary, the system observation vector is defined as:

[0082] in, The first term is the actual longitudinal velocity obtained from S103, and the second term 0 is the virtual vertical velocity observation (based on the non-integrity constraint that the average vertical velocity of the vehicle in the tunnel approaches zero).

[0083] Establish a nonlinear mapping relationship (observation equation) between state variables and observed variables:

[0084] Specific mapping function The construction is as follows:

[0085] Through the above construction, the algorithm utilizes Correcting the longitudinal velocity in the system state The vertical velocity in the system state is corrected using the virtual observation value 0. This effectively suppresses integral drift of inertial navigation in the vertical channel.

[0086] By fusing the system observation expression and the kinematic constraint expression, an extended state-space system state model is formed; In practice, the continuous dynamic equations are discretized (e.g., using the Euler method or Runge-Kutta method) to construct discrete-time extended state-space state equations:

[0087] in, Let k be the state variable at time k. The control input measured by the inertial navigation unit at time k; The system state noise vector follows a Gaussian distribution.

[0088] The specific expression for the discrete state equation is:

[0089]

[0090]

[0091]

[0092]

[0093]

[0094]

[0095] In the formula: The time step is used for discretization; in practice, it can be set to, for example, between 0.01s and 0.05s.

[0096] The system state model is processed by the unscented Kalman filter algorithm to obtain an extended state-space unscented Kalman filter model constrained by kinematic constraint expressions; The process of using the unscented Kalman filter algorithm to process the system state model includes: Determine the initial state variables based on the extended state-space system state model; In specific implementation, the initial state vector at the moment the vehicle enters the tunnel entrance is defined. (Positioning and velocity information at the last moment is typically provided by a GNSS system outside the tunnel). Simultaneously, the initial state covariance matrix is ​​set. The process noise covariance matrix Q.

[0097] Sigma sampling points for determining state variables based on initial state variables; In practical implementation, unscented transform (UT) is used to generate... Sigma sampling points ( Here, the dimension of the state is... =7), with an approximate nonlinear distribution: .

[0098] By substituting Sigma sampling points into the extended state-space system state model, the state variables and covariance matrix are predicted. In practice, each of the aforementioned Sigma sampling points is substituted into the aforementioned state equation. Calculate the predicted Sigma point for the next time step:

[0099] Accordingly, the prediction covariance matrix is ​​expressed as:

[0100] In the formula: The system process noise covariance matrix has a value determined based on specific experimental data; for example, a typical value range is between 0.001 and 0.01. The weighting coefficients... These are the mean weight and the covariance weight, respectively, and are defined as follows:

[0101]

[0102]

[0103] Among them, parameters The value is usually set to 2 to optimize filtering performance.

[0104] An adaptive observation noise covariance matrix is ​​constructed based on suspension displacement, and the filter gain matrix is ​​calculated. In practical implementation, an adaptive observation noise covariance matrix is ​​constructed. Specifically, define For diagonal matrices: ,in, The variance of longitudinal velocity observation noise (take a fixed value, such as 0.01); The observation noise variance for the vertical channel is calculated based on the suspension displacement output by S102. Perform real-time adaptive computation:

[0105] In the formula: The variance of base noise when the road surface is flat; This represents the root mean square value of the suspension displacement at the current moment; These are adaptive weighting coefficients, determined experimentally.

[0106] It should be understood that when a vehicle travels over a bumpy road ( When the vertical motion of the vehicle body is large, the error of the assumption that "the vertical velocity is 0" increases, and the algorithm automatically increases the error accordingly. Reduce the filter's weight on the virtual observation; when the road surface is flat, reduce... Strengthen constraints.

[0107] Then, by weighted summing of the predicted observation Sigma points, the predicted observation vector is obtained:

[0108] Further calculation of the predicted observation covariance matrix:

[0109] Then calculate the cross-covariance matrix between the state variables and the observed variables:

[0110] Finally, the filter gain matrix is ​​determined based on the above calculation results: .

[0111] The predicted state variables and covariance matrix are updated using the filter gain matrix to obtain an extended state-space unscented Kalman filter model constrained by kinematic constraints. In practice, the state update formula is as follows:

[0112] The formula for updating the covariance matrix is:

[0113] In the formula: The actual observation vector of the system at time k is the longitudinal velocity output by S103. It consists of a virtual vertical velocity of 0; The updated state variables include the vehicle's real-time position, speed, and yaw angle information; The updated state covariance matrix represents the uncertainty of the estimated state.

[0114] Using the extended state-space unscented Kalman filter model described above, the vehicle's location state at each time step can be estimated accurately in real time.

[0115] The vehicle's positioning coordinates are calculated based on the extended state-space unscented Kalman filter model constrained by kinematic constraints. In practical implementation, the updated state variables mentioned above will be... The system analyzes the data to obtain the real-time vehicle location coordinates, i.e.: ,in, Let represent the three-dimensional position coordinates of the vehicle at time k in the tunnel coordinate system.

[0116] For example, in a certain actual test, the real-time position coordinates of the vehicle after the above model update are as follows: , The above location coordinates are the real-time positioning results of the vehicle in a tunnel environment without GNSS, possessing extremely high accuracy and stability.

[0117] By using a complete extended state-space unscented Kalman filter model, real-time and high-precision positioning of vehicles in tunnels without GNSS was achieved, providing reliable technical support for the safe and efficient operation of vehicles in tunnels.

[0118] Example 2 like Figure 2 As shown in the example, the parts not detailed in this embodiment are as shown in Example 1. This embodiment discloses a high-precision positioning system in a tunnel in a GNSS-free environment, including: The data acquisition module 201 is used to extract inertial measurement disturbance features by means of vibration mode decomposition based on the triaxial acceleration and triaxial angular velocity output by the inertial navigation unit, and to determine the wheel slip state based on the rotation pulse data output by the wheel distance sensor. The feature decoupling module 202 is used to decouple the inertial measurement disturbance features by using a frequency response suspension vibration transmission model fitted based on the suspension damping coefficient and suspension stiffness coefficient, and to determine the suspension displacement. The pulse compensation module 203 is used to compensate for the rotational pulse data based on the wheel slip state and the nonlinear mapping relationship between the wheel slip ratio and the suspension displacement, so as to determine the vehicle travel distance. The positioning module 204 is used to calculate the vehicle positioning coordinates by integrating the vehicle's driving distance and suspension displacement inputs into an extended state-space unscented Kalman filter model that incorporates vehicle longitudinal velocity constraints, lateral sideslip constraints, and vertical vibration displacement constraints.

[0119] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters, weights, and thresholds in the formulas are set by those skilled in the art according to the actual situation.

[0120] The foregoing has only described certain exemplary embodiments of the present invention by way of illustration. Undoubtedly, those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the foregoing drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. A high-precision positioning method in a tunnel without GNSS, characterized in that, include: Based on the triaxial acceleration and triaxial angular velocity output by the inertial navigation unit, the inertial measurement disturbance characteristics are extracted using vibration mode decomposition, and the wheel slip state is determined based on the rotation pulse data output by the wheel distance sensor. The inertial measurement disturbance characteristics are decoupled using a frequency response suspension vibration transmission model fitted based on the suspension damping coefficient and suspension stiffness coefficient, and the suspension displacement is determined. Based on the wheel slip state, the nonlinear mapping relationship between wheel slip ratio and suspension displacement is used to compensate for the rotation pulse data and determine the vehicle travel distance. The vehicle's travel distance and suspension displacement are input into an extended state-space unscented Kalman filter model that integrates longitudinal velocity constraints, lateral sideslip constraints, and vertical vibration displacement constraints to calculate the vehicle's positioning coordinates.

2. The high-precision positioning method in a tunnel without GNSS environment according to claim 1, characterized in that, The extraction of inertial measurement disturbance features using vibration mode decomposition includes: Based on the original vibration signals of the inertial navigation unit collected when the vehicle passes through the tunnel entrance section, the target frequency band closely related to the suspension response is determined; Within the target frequency band, a vibration mode screening algorithm based on the local concentration of modal energy is used to extract effective vibration modes associated with the suspension response; Inertial measurement perturbation characteristics are constructed using the effective vibration modes to obtain inertial measurement perturbation characteristics limited by the local concentration of modal energy.

3. The high-precision positioning method in a tunnel without GNSS environment according to claim 2, characterized in that, The method for extracting effective vibration modes using a vibration mode screening algorithm based on the local concentration of modal energy includes: The target frequency band vibration signal is decomposed into multiple sub-frequency band signals, and the energy distribution of each sub-frequency band signal is determined. Identify characteristic sub-frequency bands that are highly correlated with suspension structure characteristics based on the energy concentration of each sub-frequency band; Vibration mode clustering is performed on characteristic sub-frequency bands to obtain effective vibration modes associated with suspension response.

4. The high-precision positioning method in a tunnel without GNSS environment according to claim 3, characterized in that, The method of decoupling inertial measurement disturbance characteristics using a frequency response suspension vibration transmission model fitted based on suspension damping coefficient and suspension stiffness coefficient includes: Based on the theoretical vibration equation of the suspension structure, the initial vibration transfer function is established. Collect suspension displacement response samples of a vehicle when it passes through a road surface joint at a typical tunnel entrance section; The initial vibration transfer function is modified by using suspension displacement response samples to obtain a suspension vibration transfer model defined by the displacement response samples. The suspension displacement is obtained by decoupling the inertial measurement disturbance characteristics using a suspension vibration transmission model defined by displacement response samples.

5. The high-precision positioning method in a tunnel without GNSS environment according to claim 4, characterized in that, The parameter correction of the initial vibration transfer function using suspension displacement response samples includes: Time-frequency analysis was performed on suspension displacement response samples collected from typical tunnel entrance sections to determine the time-frequency characteristics of the suspension displacement response; Based on the time-frequency characteristics of the suspension displacement response, a parametric sensitivity model of the initial vibration transfer function is constructed; The optimal correction range for suspension damping coefficient and stiffness coefficient is determined by using a parametric sensitivity model. The initial vibration transfer function is updated based on the optimal correction range to obtain the suspension vibration transfer model defined by the displacement response samples.

6. The high-precision positioning method in a tunnel without GNSS environment according to claim 5, characterized in that, The method of compensating for rotational pulse data using the nonlinear mapping relationship between wheel slip ratio and suspension displacement includes: Construct a sample set that correlates suspension displacement features with wheel slip ratio under different wheel slip states; Using the support vector regression algorithm, nonlinear mapping relationships under specific states are constructed based on each associated sample set; Calculate the real-time wheel slip ratio based on the real-time determined wheel slip state and suspension displacement; The vehicle travel distance is obtained by compensating the rotational pulse data with the predicted wheel slip ratio.

7. The high-precision positioning method in a tunnel without GNSS environment according to claim 6, characterized in that, The calculation of vehicle positioning coordinates includes: Based on the vehicle dynamics equations, kinematic constraint expressions for the vehicle's longitudinal velocity, lateral sideslip, and vertical vibration displacement are established respectively. Based on the kinematic constraint expression, and combined with the vehicle travel distance and suspension displacement, a system observation expression in the extended state space is constructed; By fusing the system observation expression and the kinematic constraint expression, an extended state-space system state model is formed; The system state model is processed by the unscented Kalman filter algorithm to obtain an extended state-space unscented Kalman filter model constrained by kinematic constraint expressions; The vehicle's positioning coordinates are calculated based on the extended state-space unscented Kalman filter model constrained by kinematic constraints.

8. The high-precision positioning method in a tunnel without GNSS environment according to claim 7, characterized in that, The process of using the unscented Kalman filter algorithm to process the system state model includes: Determine the initial state variables based on the extended state-space system state model; Sigma sampling points for determining state variables based on initial state variables; By substituting Sigma sampling points into the extended state-space system state model, the state variables and covariance matrix are predicted. An adaptive observation noise covariance matrix is ​​constructed based on suspension displacement, and the filter gain matrix is ​​calculated. The predicted state variables and covariance matrix are updated using the filter gain matrix, resulting in an extended state-space unscented Kalman filter model constrained by kinematic constraints.

9. A high-precision positioning system in a tunnel without GNSS, implemented based on any one of claims 1-8, characterized in that, include: The data acquisition module is used to extract inertial measurement disturbance features based on the triaxial acceleration and triaxial angular velocity output by the inertial navigation unit using vibration mode decomposition, and to determine the wheel slip state based on the rotation pulse data output by the wheel distance sensor. The feature decoupling module is used to decouple the inertial measurement disturbance features by using a frequency response suspension vibration transmission model fitted based on the suspension damping coefficient and suspension stiffness coefficient, and to determine the suspension displacement. The pulse compensation module is used to compensate for the rotational pulse data based on the wheel slip state and the nonlinear mapping relationship between the wheel slip ratio and the suspension displacement, so as to determine the vehicle travel distance. The positioning module is used to calculate the vehicle's positioning coordinates by integrating the vehicle's driving distance and suspension displacement inputs with an extended state-space unscented Kalman filter model that incorporates vehicle longitudinal velocity constraints, lateral sideslip constraints, and vertical vibration displacement constraints.

Images