Adaptive multi-source information fusion positioning method and system based on lane line constraint

Abstract

The application discloses a kind of adaptive multi-source information fusion positioning method and system based on lane line constraint, belong to the field of automatic driving technology.The method fuses the data of inertial measurement unit, satellite positioning module, wheel speed meter and lane line detection module by extending Kalman filter, constructs the state model including position, speed, attitude error and sensor zero offset, and according to whether there is high-precision map, flexibly uses position error or lateral deviation as lane line constraint observation model.In addition, the application also introduces adaptive observation noise adjustment mechanism based on lane line detection quality and degradation processing mechanism based on chi-square detection.The application effectively solves the positioning drift problem in satellite signal shielding environment, improves the vehicle lateral positioning accuracy and the robustness of system, realizes low-cost, high-precision continuous positioning.

Description

Technical Field

[0001] This invention relates to the field of autonomous driving technology, and more specifically, to an adaptive multi-source information fusion localization method and system based on lane line constraints. Background Technology

[0002] Autonomous driving technology has been widely used on various unmanned transportation platforms. Its application scenarios are complex and varied, covering urban canyons, tunnels, elevated roads, and indoor scenarios such as various factories, which puts forward higher requirements for the accuracy and reliability of the fusion positioning system.

[0003] In the aforementioned complex environments, existing algorithms struggle to simultaneously meet the combined demands of low cost, high accuracy, and high reliability. Satellites are susceptible to obstruction, multipath effects, and interference, leading to decreased positioning accuracy and an inability to provide continuous and stable navigation information. While inertial measurement units (IMUs) can provide continuous attitude, velocity, and position information for short periods when satellite signals are lost, their errors accumulate rapidly over time due to factors such as zero bias, calibration errors, and temperature drift, making it impossible to maintain accuracy in the long term. Wheel speed sensors can improve pose accuracy to some extent, but their errors also increase with mileage, especially under conditions of tire slippage or vehicle drift, where the errors diverge even faster. Visual sensors can acquire environmental information such as lane lines and are sensitive to lateral vehicle displacement, but their observation stability is insufficient under conditions of changing lighting, obstruction, or unclear road markings. Therefore, in complex environments, there is an urgent need for a robust navigation method that can integrate lane line observation information into traditional navigation algorithms, possess adaptive weight adjustment capabilities, and an abnormal condition degradation handling mechanism, achieving a balance of low cost, high accuracy, and high reliability without adding additional sensors.

[0004] Currently, the commonly used low-cost multi-source fusion positioning methods in various unmanned transportation platforms mainly include inertial and satellite combined navigation, visual-assisted methods, and multi-sensor fusion methods.

[0005] The combined inertial and satellite navigation method uses an IMU for state prediction and satellite information for periodic correction, and is currently the most widely used approach. However, this method is highly dependent on satellite signals, lacks effective alternative observation sources in signal-limited environments, and relying solely on inertial navigation leads to rapid error accumulation and a significant decrease in system accuracy.

[0006] Lane line assist methods acquire lane line information through visual sensors and are typically used for driving assistance or route planning. However, most existing technologies do not introduce lane line information as an observation into navigation filters, but only use it as reference information, making it difficult to directly constrain navigation errors, especially lateral errors.

[0007] In summary, existing technologies still have significant shortcomings in terms of in-depth utilization of lane line information, adaptive fusion capabilities, and handling of abnormal operating conditions. Summary of the Invention

[0008] One objective of this invention is to provide an adaptive multi-source information fusion positioning method and system based on lane line constraints, in order to solve the problems of existing technologies in terms of in-depth utilization of lane line information, adaptive fusion capability, and handling of abnormal working conditions.

[0009] According to a first aspect of the present invention, an adaptive multi-source information fusion localization method based on lane line constraints is provided, the method comprising the following steps:

[0010] Acquire inertial data output by the inertial measurement unit, absolute position data output by the satellite positioning module, and lane line observation data output by the lane line detection module;

[0011] Based on the availability of high-precision map data, construct a corresponding lane line constraint model, and convert the lane line observation data into a position error observation vector or a lateral deviation observation vector.

[0012] An extended Kalman filter is constructed, and the state is predicted using the inertial data. The state is updated using the absolute position data and the position error observation vector or the lateral deviation observation vector.

[0013] In the state update process, the observation noise covariance matrix is ​​dynamically adjusted according to the lane line detection quality, and k-square detection is performed based on the observation residuals.

[0014] When an observation anomaly is detected, a degradation processing mechanism is activated to either block the abnormal observation or perform only a time update in order to output the fused positioning result.

[0015] Optionally, the step of constructing a corresponding lane constraint model based on the availability of high-precision map data specifically includes:

[0016] If a high-precision map is available, the reference position of the lane centerline in the high-precision map can be obtained. Combined with the ratio of the left and right lane lines obtained by the visual sensor, the planar position error vector of the vehicle relative to the lane centerline can be calculated and projected onto the vehicle coordinate system to obtain the lateral error and longitudinal error, and the position error observation equation can be constructed.

[0017] Without the assistance of high-precision maps, the lateral deviation of the vehicle relative to the center of the lane is calculated based on the geometric relationship between the left and right lane lines measured by visual sensors, and a lateral deviation observation equation is constructed.

[0018] Optionally, if a high-precision map is available, the planar position error vector of the vehicle relative to the lane centerline is calculated, specifically including:

[0019] Vehicle position Lane reference position Mapping to a geographic coordinate system yields a relative position vector:

[0020] ,in, It is the planar position error vector of the vehicle relative to the lane reference point;

[0021] Combined with the vehicle's current heading angle Construct the vehicle coordinate system orientation basis vectors and project the error vectors onto the vehicle coordinate system to obtain the lateral error. With longitudinal error :

[0022] ,in, This represents the forward unit vector of the vehicle. This represents the lateral unit vector of the vehicle. This is the accumulated historical lateral offset compensation value, used to eliminate inertial integration errors or historical measurement biases, when satellite signal is good. The lane line observation error is avoided by using satellite and lane reference positions and deducting them when calculating lateral deviation.

[0023] Optionally, if no high-precision map is available, the lateral deviation of the vehicle relative to the lane center is calculated, specifically including:

[0024] Acquire information on the left and right lane lines of the vehicle as measured by a vision sensor. , lane width ;

[0025] Calculate the lateral deviation of the vehicle relative to the lane center :

[0026] ,in, This indicates that the vehicle is veering to the right. This indicates that the vehicle is veering to the left.

[0027] Optionally, an extended Kalman filter is constructed to perform state prediction using the inertial data, specifically including:

[0028] A 15-dimensional EKF state vector is constructed, which includes the position error, velocity error, attitude misalignment angle, accelerometer bias, and gyroscope drift of the carrier.

[0029] Perform EKF state prediction, and calculate the prior state estimate at time k based on the measurement value of the inertial measurement unit (IMU) through the state transition function;

[0030] Perform EKF state update, sequentially fuse satellite position observation, wheel speed meter speed observation and lane line lateral deviation observation, calculate Kalman gain and correct prior state estimate to obtain posterior state estimate at time k;

[0031] The original navigation solution of the vehicle is corrected based on the posterior state estimate, and the final positioning result is output.

[0032] Optionally, the step of dynamically adjusting the observation noise covariance matrix based on lane detection quality specifically includes:

[0033] Acquiring lane line image quality indicators Consistency index of left and right lane lines and the continuous effective duration index of lane line information The lane line ratio confidence level α is obtained by weighting the various indicators according to the preset weights, where α ∈ [0.2, 1].

[0034] The confidence level of the lane line proportion Introducing the observation noise covariance matrix R k The formula used in the calculation is as follows:

[0035] ;

[0036] in, For the reference noise covariance, when When approaching the upper limit, increase the weight of the observation; when When the value approaches 0, the weight of the observation is reduced.

[0037] Optionally, the step of performing k-square detection based on the observed residuals specifically includes:

[0038] The k-square values ​​λ1 for satellite position, λ2 for wheel speed meter, and λ3 for lane line position are calculated separately using the following formulas:

[0039] Satellite location item Kafang ;

[0040] Wheel speed meter speed ka square ;

[0041] Lane line position (click) ;

[0042] In the formula, , Indicates the first To the One element, express The To the Line and number To the A matrix block composed of columns;

[0043] because obey The distribution and fault determination criteria are as follows:

[0044] ;in It is a pre-set threshold.

[0045] Optionally, the initiation of the degradation processing mechanism specifically includes:

[0046] When one or both of the satellite, wheel speedometer, or lane line detection results are abnormal, the abnormal observation is stopped from participating in the filter update, and only the normal observation is used for the status update.

[0047] When the detection results of satellite, wheel speedometer, and lane line are all abnormal, the state update step is stopped, and only the time update step is executed to maintain the inertial prediction mode until the observation results are detected to return to normal.

[0048] According to a second aspect of the present invention, an adaptive multi-source information fusion positioning system based on lane line constraints is also provided, the system comprising a sensor module and a computing unit;

[0049] The sensor module includes an inertial measurement unit, a satellite positioning module, a wheel speed sensor, and a lane detection module. The inertial measurement unit and the lane detection module are both connected to the computing unit via an RS422 interface. The satellite positioning module is connected to the computing unit via an RS232 interface. The wheel speed sensor is connected to the computing unit via a CAN interface.

[0050] The computing unit includes a microcontroller configured to perform the following controls:

[0051] Acquire inertial data output by the inertial measurement unit, absolute position data output by the satellite positioning module, wheel speed data output by the wheel speed meter, and lane line observation data output by the lane line detection module;

[0052] Based on the availability of high-precision map data, construct a corresponding lane line constraint model and convert lane line observation data into position error observation vectors or lateral deviation observation vectors.

[0053] An extended Kalman filter is constructed to predict the state using inertial data and wheel speed data, and to update the state using absolute position data and position error observation vectors or lateral deviation observation vectors.

[0054] During the state update process, the observation noise covariance matrix is ​​dynamically adjusted based on the lane line detection quality, and chi-square detection is performed based on the observation residuals. When an observation anomaly is detected, a degradation processing mechanism is activated to shield the abnormal observation or perform only time updates in order to output the fused positioning results.

[0055] According to a third aspect of the present invention, a computer-readable storage medium is also provided, on which a computer program is stored, which, when executed by a processor, implements the adaptive multi-source information fusion positioning method based on lane line constraints as described in the first aspect of the present invention.

[0056] The adaptive multi-source information fusion localization method and system based on lane line constraints disclosed herein have the following technical advantages:

[0057] This invention transforms the lane line geometry information acquired by a visual sensor into position constraint observations that can be used for filtering and estimation, and introduces them into a multi-source fusion positioning system, thereby suppressing position error divergence in environments where satellite signals are limited or unavailable.

[0058] With the assistance of high-precision maps, the absolute position information of the vehicle relative to the lane centerline is obtained by matching lane lines with the lane centerline in the high-precision map, and position error observation is constructed to directly constrain the pose, thereby improving positioning accuracy and enhancing the overall accuracy and stability of the navigation system. Without the assistance of high-precision maps, modeling is performed based on the geometric relationship of lane lines, converting the lane line detection results into a lateral deviation observation of the vehicle relative to the lane center, and introducing a filter to effectively constrain and correct lateral position errors. The method and system of this invention can operate independently of a map, providing reliable constraints on lateral accuracy while simultaneously improving overall navigation accuracy.

[0059] In the observation modeling and fusion process, an adaptive observation noise adjustment mechanism based on lane line detection quality is introduced. The observation noise covariance is dynamically adjusted according to lane line image quality and continuity indicators to achieve adaptive weighted fusion of multi-source information. Simultaneously, to address potential anomalies, jumps, or loss in lane line observations, a ka-square detection method is used to identify and remove abnormal observations, ensuring stable filter operation and preventing filter divergence caused by abnormal observations. This maintains the continuous navigation capability and overall reliability of the fused positioning system, thereby guaranteeing the system's accuracy and reliability.

[0060] Therefore, this invention can obtain high-precision absolute positioning results when a high-precision map is available, and can also operate independently when a high-precision map is unavailable. By introducing lateral deviation observation, it can effectively improve positioning accuracy and has good environmental adaptability and engineering practical value.

[0061] Other features and advantages of the invention will become clear from the following detailed description of exemplary embodiments of the invention with reference to the accompanying drawings. Attached Figure Description

[0062] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments of the invention and, together with their description, serve to explain the principles of the invention.

[0063] Figure 1 This is a flowchart illustrating the adaptive multi-source information fusion localization method based on lane line constraints provided in an embodiment of the present invention.

[0064] Figure 2 The hardware connection diagram of the adaptive multi-source information fusion positioning system based on lane line constraints provided in the embodiments of the present invention;

[0065] Figure 3 This is a comparison chart of position errors of different fusion schemes under conditions of effective and ineffective lane line observation in the embodiments of the present invention. Detailed Implementation

[0066] Various exemplary embodiments of the present invention will now be described in detail with reference to the accompanying drawings. It should be noted that, unless otherwise specifically stated, the relative arrangement, numerical expressions, and values ​​of the components and steps set forth in these embodiments do not limit the scope of the invention.

[0067] The following description of at least one exemplary embodiment is merely illustrative and is in no way intended to limit the invention or its application or use.

[0068] Techniques, methods, and equipment known to those skilled in the art may not be discussed in detail, but where appropriate, such techniques, methods, and equipment should be considered part of the specification.

[0069] In all the examples shown and discussed herein, any specific values ​​should be interpreted as merely exemplary and not as limitations. Therefore, other examples of exemplary embodiments may have different values.

[0070] This invention proposes an embodiment of an adaptive multi-source information fusion localization method based on lane line constraints, specifically, as follows: Figure 1 As shown, it includes the following steps:

[0071] Acquire inertial data output by the inertial measurement unit, absolute position data output by the satellite positioning module, and lane line observation data output by the lane line detection module;

[0072] Based on the availability of high-precision map data, construct a corresponding lane line constraint model, and convert the lane line observation data into a position error observation vector or a lateral deviation observation vector.

[0073] An extended Kalman filter is constructed, and the state is predicted using the inertial data. The state is updated using the absolute position data and the position error observation vector or the lateral deviation observation vector.

[0074] In the state update process, the observation noise covariance matrix is ​​dynamically adjusted according to the lane line detection quality, and k-square detection is performed based on the observation residuals.

[0075] When an observation anomaly is detected, a degradation processing mechanism is activated to either block the abnormal observation or perform only a time update in order to output the fused positioning result.

[0076] In this embodiment of the invention, after detecting lane line information, it is determined whether corresponding high-precision map data exists based on the current location. If a map exists, the planar position error and observation error of the lane line are calculated based on the high-precision map; if no map exists, the lateral deviation of the lane line and its observations are calculated. Subsequently, EKF fusion of IMU / wheel speed sensor / satellite / lane line is performed. If lane line information is unavailable, it degenerates to EKF fusion of IMU / wheel speed sensor / satellite to ensure continuous navigation capability. The step of constructing a corresponding lane line constraint model based on the existence of high-precision map data specifically includes:

[0077] If a high-precision map is available, the reference position of the lane centerline in the high-precision map can be obtained. Combined with the ratio of the left and right lane lines obtained by the visual sensor, the planar position error vector of the vehicle relative to the lane centerline can be calculated and projected onto the vehicle coordinate system to obtain the lateral error and longitudinal error, and the position error observation equation can be constructed.

[0078] Without the assistance of high-precision maps, the lateral deviation of the vehicle relative to the center of the lane is calculated based on the geometric relationship between the left and right lane lines measured by visual sensors, and a lateral deviation observation equation is constructed.

[0079] In this embodiment of the invention, when a high-precision map is available, the vehicle control center should coincide with the lane center of the high-precision map by default. If a high-precision map is available, the planar position error vector of the vehicle relative to the lane centerline is calculated, specifically including:

[0080] Lane line measurement: The vehicle position is obtained by combining the ratio of the left and right lane lines acquired by the visual sensor with a high-precision map;

[0081] Obtain vehicle navigation status: fuse values ​​calculated by the positioning system, including attitude, speed, and position;

[0082] Vehicle position Lane reference position Mapping to a geographic coordinate system yields a relative position vector:

[0083] ,in, It is the planar position error vector of the vehicle relative to the lane reference point;

[0084] Combined with the vehicle's current heading angle Construct the vehicle coordinate system orientation basis vectors and project the error vectors onto the vehicle coordinate system to obtain the lateral error. With longitudinal error :

[0085] ,in, This represents the forward unit vector of the vehicle. This represents the lateral unit vector of the vehicle. This is the accumulated historical lateral offset compensation value, used to eliminate inertial integration errors or historical measurement biases, when satellite signal is good. The position is obtained through satellite and lane reference positions and deducted when calculating lateral deviation to avoid the accumulation of lane line observation errors. It should be noted that in calculating the planar position error vector, when the satellite signal quality meets a preset threshold, the position deviation between the current satellite positioning data and the position data from the fused positioning system is calculated. Subsequently, the fused positioning system is introduced to correct this deviation, thereby achieving alignment between the high-precision map and the position calculated by the fused positioning system. When the satellite signal quality does not meet the preset threshold, historical lateral deviation compensation values ​​are calculated. Subsequently, a fusion positioning system was introduced to correct the lateral deviation in order to eliminate the integral drift of inertial navigation.

[0086] In this embodiment of the invention, without a high-precision map, the vehicle control center is assumed to coincide with the lane center. Without the assistance of a high-precision map, the lateral deviation of the vehicle relative to the lane center can only be corrected based on the lane line results and the fused positioning results, specifically including:

[0087] Lane line measurement: Lane line information for the left and right sides of the vehicle measured by a vision sensor. , lane width ;

[0088] Obtain vehicle navigation status: fuse values ​​calculated by the positioning system, including attitude, speed, and position;

[0089] Acquire information on the left and right lane lines of the vehicle as measured by a vision sensor. , lane width ;

[0090] Calculate the lateral deviation of the vehicle relative to the lane center :

[0091] ,in, This indicates that the vehicle is veering to the right. This indicates that the vehicle is veering to the left.

[0092] In this embodiment of the invention, an extended Kalman filter is constructed, and state prediction is performed using the inertial data, specifically including:

[0093] Construct a 15-dimensional EKF state vector X, which includes the carrier's position error, velocity error, attitude misalignment angle, accelerometer bias, and gyroscope drift; specifically, ,in, , , These are latitude error, altitude error, and longitude error, respectively. , , These are the northward velocity error, the celestial velocity error, and the eastward velocity error, respectively. , , These are the northward misalignment angle, the celestial misalignment angle, and the eastward misalignment angle, respectively. , , The accelerometer zero biases are for the x-axis, y-axis, and z-axis, respectively. , , These represent the gyroscope drift along the x, y, and z axes, respectively.

[0094] EKF state prediction is performed by calculating the prior state estimate at time k based on the measurements from the inertial measurement unit (IMU) using a state transition function. Specifically, the IMU is used for state prediction, completing the forward prediction of the inertial navigation state by integrating angular velocity and acceleration. The specific formula is as follows: ,in, This data was obtained from IMU measurements.

[0095] Perform EKF state update by sequentially fusing satellite position observations, wheel speedometer observations, and lane lateral deviation observations. Calculate the Kalman gain and correct the prior state estimate to obtain the posterior state estimate at time k. The specific formula is: Observation In addition to satellite position observation and wheel speed measurement, it also includes lane line position / lateral deviation observation, which was mentioned in the previous section.

[0096] When aided by a high-precision map, the location observation matrix block is The observation equation is: ,in, , These represent the positional errors in the North-East coordinate system; For noise observation, the level is typically in the centimeter range.

[0097] Without high-precision map assistance, the position observation vector is The observation equation is: ,in, This is adaptive observation noise.

[0098] The original navigation solution of the carrier is corrected based on the posterior state estimate, and the final positioning result is output. During the update process, the observation noise covariance is adaptively adjusted according to the satellite signal quality or lane line signal quality. When observations are abnormal or lost, the system initiates degradation processing strategies, including covariance expansion or observation masking, to prevent the EKF state from diverging. This enables dynamic weighted fusion of different sensors, ensuring the continuity and robustness of the fused positioning system.

[0099] In this embodiment of the invention, the step of dynamically adjusting the observation noise covariance matrix based on lane detection quality specifically includes:

[0100] Acquiring lane line image quality indicators Consistency index of left and right lane lines and the continuous effective duration index of lane line information , , , All information is known from the lane line detection module. The indicators are weighted according to preset weights to obtain the lane line ratio confidence level α, where α∈[0.2,1]. The calculation method is as follows:

[0101] ,

[0102] in, , , The weights for each indicator are determined through experimental calibration or empirical setting to meet the requirements. In this invention patent, Take 0.6, Take 0.25, Take 0.15.

[0103] An adaptive observation noise adjustment mechanism is introduced, which dynamically updates the observation noise covariance matrix of the EKF based on sensor observation quality, avoiding state divergence caused by abnormal observations and improving fusion accuracy. The lane line proportion confidence level is then used. Introducing the observation noise covariance matrix R k The formula used in the calculation is as follows:

[0104] ;

[0105] in, For the reference noise covariance, when When approaching the upper limit, increase the weight of the observation; when When the value approaches 0, the weight of the observation is reduced.

[0106] In this embodiment of the invention, a degradation processing mechanism is introduced, which detects the calorie content. This involves shielding abnormal observations and maintaining the inertial prediction mode. Specifically, the step of performing k-square detection based on the observation residuals includes:

[0107] The k-square values ​​λ1 for satellite position, λ2 for wheel speed meter, and λ3 for lane line position are calculated separately using the following formulas:

[0108] Satellite location item Kafang ;

[0109] Wheel speed meter speed ka square ;

[0110] Lane line position (click) ;

[0111] In the formula, , Indicates the first To the One element, express The To the Line and number To the A matrix block composed of columns;

[0112] because obey The distribution and fault determination criteria are as follows:

[0113] ;in It is a pre-set threshold.

[0114] In this embodiment of the invention, the initiation of the degradation processing mechanism specifically includes:

[0115] When one or both of the satellite, wheel speedometer, or lane line detection results are abnormal, the abnormal observation is stopped from participating in the filter update, and only the normal observation is used for the status update.

[0116] When the detection results from satellite, wheel speedometer, and lane markings are all abnormal, the status update step is stopped, and only the time update step is executed, maintaining the inertial prediction mode until the observation results are detected to have returned to normal. Through the above mechanism, the system can still continuously output navigation status when the satellite signal is poor, the wheel speedometer is abnormal, or the lane markings are invalid, ensuring the vehicle's continuous navigation capability in tunnels, urban canyons, or complex road environments.

[0117] This invention also provides an embodiment of an adaptive multi-source information fusion positioning system based on lane line constraints, specifically as follows: Figure 2 As shown, the system includes a sensor module and a computing unit;

[0118] The sensor module includes an inertial measurement unit (IMU), a satellite positioning module, a wheel speed sensor, and a lane detection module. The IMU and lane detection module are connected to the computing unit via an RS422 interface, the satellite positioning module via an RS232 interface, and the wheel speed sensor via a CAN interface. The IMU continuously outputs three-axis angular velocity and acceleration for forward prediction in inertial navigation. The satellite positioning module outputs position information for state correction. The lane detection module is a visual sensor used to detect the boundaries of the vehicle's left and right lane lines.

[0119] The computing unit includes a microcontroller configured to perform the following controls:

[0120] Acquire inertial data output by the inertial measurement unit, absolute position data output by the satellite positioning module, wheel speed data output by the wheel speed meter, and lane line observation data output by the lane line detection module;

[0121] Depending on whether high-precision map data is available, a corresponding lane line constraint model is constructed, and the lane line observation data is converted into position error observation vectors or lateral deviation observation vectors. If high-precision map data is available, the position error observation vector of the vehicle center is output; if high-precision map data is not available, the lateral deviation observation vector is output.

[0122] An extended Kalman filter is constructed to predict the state using inertial data and wheel speed data, and to update the state using absolute position data and position error observation vectors or lateral deviation observation vectors.

[0123] During the state update process, the observation noise covariance matrix is ​​dynamically adjusted based on the lane line detection quality, and chi-square detection is performed based on the observation residuals. When an observation anomaly is detected, a degradation processing mechanism is activated to shield the abnormal observation or perform only time updates in order to output the fused positioning results.

[0124] The inertial measurement unit (IMU) involved in this invention provides vehicle angular velocity and linear acceleration for navigation state prediction. The satellite positioning module provides absolute position observations for navigation state correction. The lane detection module outputs absolute position when a high-precision map is available, and outputs the ratio of left and right lane lines when no map is available. The microcontroller is responsible for acquiring, synchronizing, and preprocessing data from each sensor, and executing the extended Kalman filter (EKF) process to achieve multi-source information fusion and state estimation. With a high-precision map, lane centerlines can be further provided to improve the accuracy of lateral deviation quantification. However, this invention's method can operate independently without relying on a map, ensuring feasibility and convenience for engineering implementation. A hardware connection diagram is shown in Figure 2, where all sensor signals are uniformly transmitted to the microcontroller, which performs data fusion and state estimation.

[0125] It should be noted that the information interaction and execution process between the above-mentioned devices are based on the same concept as the method embodiments of this application. They are systems corresponding to the adaptive multi-source information fusion positioning method based on lane line constraints. All implementation methods in the above-mentioned method embodiments are applicable to the embodiments of this system. For details on their specific functions and the resulting technical effects, please refer to the method embodiment section, which will not be repeated here.

[0126] Furthermore, in a specific embodiment of this invention, actual road tests are conducted on an open road, using satellite information as a reference, but not directly for navigation. On the same test route, the positional accuracy is compared under various fusion schemes: inertial / wheel speed measurement, inertial / wheel speed measurement / lane lines (without high-precision map assistance), and inertial / wheel speed measurement / lane lines (with high-precision map assistance). Specifically, as follows... Figure 3 As shown in the figure, actual tests demonstrate that when satellite signals are unavailable, introducing lane line observations can effectively suppress position error divergence. Specifically,

[0127] Curve 1 represents the time-varying error of fusion positioning using only inertial measurement units and wheel speedometers; Curve 2 represents the fusion positioning error without high-precision map assistance, relying solely on visual lane line lateral deviation observation; Curve 3 represents the fusion positioning error with high-precision map assistance, using lane centerline matching to obtain position error observation; Curve 4 indicates the valid state of lane line observation information, with high level indicating validity and low level indicating invalidity, used to analyze the timing of lane line constraint intervention or withdrawal. The results show that:

[0128] With the assistance of high-precision maps, the position error in 600 seconds decreased from 0.57m to 0.24m, improving accuracy by more than 50%. By using the absolute position reference provided by the lane centerline of the high-precision map and combining it with historical lateral deviation compensation values, inertial integral drift was eliminated, thus achieving high-precision absolute positioning.

[0129] Without the assistance of high-precision maps, the position error in 600 seconds decreased from 0.57m to 0.42m, which is more than 30% higher than the pure inertial / wheel speed meter solution. This shows that even in general scenarios without high-precision maps, this invention can still effectively constrain the lateral position error by converting the visual lane line detection results into lateral deviation observations, thus significantly improving the positioning accuracy.

[0130] During the brief invalid period observed in curve 4, curves 2 and 3 did not exhibit any jumps or divergences caused by erroneous observations. When lane line observations are abnormal or briefly lost, the system automatically masks the abnormal observation and reverts to the inertial / wheel speed sensor fusion mode, automatically reconnecting once observations are restored. Simultaneously, the adaptive observation noise adjustment mechanism dynamically adjusts the observation noise covariance based on lane line detection quality, ensuring a smooth transition of the filter when observation quality fluctuates.

[0131] The above experimental results fully demonstrate the effectiveness and superiority of the adaptive multi-source information fusion positioning method based on lane line constraints proposed in this invention. In environments with limited satellite signals, this invention can significantly suppress position error divergence, especially improving lateral positioning accuracy. Simultaneously, the adaptive noise adjustment and degradation processing mechanism ensures the system's robustness and continuous navigation capability in complex and dynamic road environments, achieving low-cost, high-precision, and reliable positioning.

[0132] The present invention also provides an embodiment of a computer-readable storage medium having a computer program stored thereon, characterized in that the program, when executed by a processor, implements the above-described adaptive multi-source information fusion positioning method based on lane line constraints.

[0133] In summary, this invention constructs an extended Kalman filter state model incorporating position, velocity, attitude errors, and sensor zero bias, deeply fusing data from the inertial measurement unit, satellite positioning, wheel speedometer, and lane detection. Depending on the availability of high-precision maps, a lane-constrained observation model for position errors or lateral deviations is flexibly constructed, and an adaptive noise adjustment mechanism based on lane quality and a chi-square detection degradation processing mechanism are introduced. This technical solution effectively addresses positioning drift issues in environments with satellite signal obstruction or weak signals, significantly improving lateral positioning accuracy through lane constraints. Simultaneously, the adaptive noise adjustment and anomaly detection mechanisms enhance the system's robustness and reliability under complex road conditions, achieving low-cost, high-precision continuous positioning.

[0134] The above description of the structure, features, and effects of the present invention is based on the embodiments shown in the figures. However, the above are only preferred embodiments of the present invention. It should be noted that the technical features involved in the above embodiments and their preferred methods can be reasonably combined and matched by those skilled in the art to form a variety of equivalent solutions without departing from or changing the design concept and technical effects of the present invention. Therefore, the present invention is not limited to the scope of implementation shown in the figures. Any changes made in accordance with the concept of the present invention, or modifications to equivalent embodiments, that do not exceed the spirit covered by the specification and figures, should be within the protection scope of the present invention.

Claims

1. An adaptive multi-source information fusion localization method based on lane line constraints, characterized in that, The method includes the following steps: Acquire inertial data output by the inertial measurement unit, absolute position data output by the satellite positioning module, and lane line observation data output by the lane line detection module; Based on the availability of high-precision map data, construct a corresponding lane line constraint model, and convert the lane line observation data into a position error observation vector or a lateral deviation observation vector. An extended Kalman filter is constructed, and the state is predicted using the inertial data. The state is updated using the absolute position data and the position error observation vector or the lateral deviation observation vector. In the state update process, the observation noise covariance matrix is ​​dynamically adjusted according to the lane line detection quality, and k-square detection is performed based on the observation residuals. When an anomaly is detected in the observation, a degradation processing mechanism is activated to either block the anomaly observation or perform only a time update in order to output the fused positioning results. Specifically, the observation noise covariance matrix is ​​dynamically adjusted based on the lane detection quality, including: Acquiring lane line image quality metrics Consistency index of left and right lane lines and the continuous effective duration index of lane line information The lane line ratio confidence level α is obtained by weighting each of the indicators according to the preset weights, where α ∈ [0.2, 1]. The confidence level of the lane line proportion Introducing the observation noise covariance matrix R k The formula for the calculation is as follows: ； in, For the reference noise covariance, when When approaching the upper limit, increase the weight of the observation; when When the value approaches 0, the weight of the observation is reduced.

2. The adaptive multi-source information fusion localization method based on lane line constraints according to claim 1, characterized in that, The step of constructing a corresponding lane constraint model based on the availability of high-precision map data specifically includes: If a high-precision map is available, the reference position of the lane centerline in the high-precision map can be obtained. Combined with the ratio of the left and right lane lines obtained by the visual sensor, the planar position error vector of the vehicle relative to the lane centerline can be calculated and projected onto the vehicle coordinate system to obtain the lateral error and longitudinal error, and the position error observation equation can be constructed. Without the assistance of high-precision maps, the lateral deviation of the vehicle relative to the center of the lane is calculated based on the geometric relationship between the left and right lane lines measured by visual sensors, and a lateral deviation observation equation is constructed.

3. The adaptive multi-source information fusion positioning method based on lane line constraints according to claim 2, characterized in that, If a high-precision map is available, the planar position error vector of the vehicle relative to the lane centerline is calculated, specifically including: Vehicle position Lane reference position Mapping to a geographic coordinate system yields a relative position vector: ,in, It is the planar position error vector of the vehicle relative to the lane reference point; Combined with the vehicle's current heading angle Construct the vehicle coordinate system orientation basis vectors and project the error vectors onto the vehicle coordinate system to obtain the lateral error. With longitudinal error : ,in, This represents the forward unit vector of the vehicle. This represents the lateral unit vector of the vehicle. This is the accumulated historical lateral offset compensation value, used to eliminate inertial integration errors or historical measurement biases, when satellite signal is good. The location is obtained through satellite and lane reference positions and deducted when calculating lateral deviation.

4. The adaptive multi-source information fusion localization method based on lane line constraints according to claim 2, characterized in that, Without the assistance of high-precision maps, the calculation of the vehicle's lateral deviation relative to the lane center includes: Acquire information on the left and right lane lines of the vehicle as measured by a vision sensor. , lane width ; Calculate the lateral deviation of the vehicle relative to the lane center : ,in, This indicates that the vehicle is veering to the right. This indicates that the vehicle is veering to the left.

5. The adaptive multi-source information fusion localization method based on lane line constraints according to claim 1, characterized in that, Constructing an extended Kalman filter and using the inertial data for state prediction specifically includes: A 15-dimensional EKF state vector is constructed, which includes the position error, velocity error, attitude misalignment angle, accelerometer bias, and gyroscope drift of the carrier. Perform EKF state prediction, and calculate the prior state estimate at time k based on the measurement value of the inertial measurement unit (IMU) through the state transition function; Perform EKF state update, sequentially fuse satellite position observation, wheel speed meter speed observation and lane line lateral deviation observation, calculate Kalman gain and correct prior state estimate to obtain posterior state estimate at time k; The original navigation solution of the vehicle is corrected based on the posterior state estimate, and the final positioning result is output.

6. The adaptive multi-source information fusion localization method based on lane line constraints according to claim 1, characterized in that, The k-square test is performed based on the observed residuals, specifically including: The k-square values ​​λ1 for satellite position, λ2 for wheel speed meter, and λ3 for lane line position are calculated separately using the following formulas: Satellite location item Kafang ; Wheel speed meter speed square ; Lane line position (click) ; In the formula, , Indicates the first To the One element, express The To the Line and number To the A matrix block composed of columns; because obey The distribution and fault determination criteria are as follows: ;in It is a pre-set threshold.

7. The adaptive multi-source information fusion localization method based on lane line constraints according to claim 6, characterized in that, The degradation process mechanism is activated, specifically including: When one or both of the satellite, wheel speedometer, or lane line detection results are abnormal, the abnormal observation is stopped from participating in the filter update, and only the normal observation is used for the status update. When the detection results of satellite, wheel speedometer, and lane line are all abnormal, the state update step is stopped, and only the time update step is executed to maintain the inertial prediction mode until the observation results are detected to return to normal.

8. An adaptive multi-source information fusion positioning system based on lane line constraints, characterized in that, The adaptive multi-source information fusion positioning method based on lane line constraints according to any one of claims 1 to 7, wherein the system includes a sensor module and a computing unit; The sensor module includes an inertial measurement unit, a satellite positioning module, a wheel speed sensor, and a lane detection module. The inertial measurement unit and the lane detection module are both connected to the computing unit via an RS422 interface. The satellite positioning module is connected to the computing unit via an RS232 interface. The wheel speed sensor is connected to the computing unit via a CAN interface. The computing unit includes a microcontroller configured to perform the following controls: Acquire inertial data output by the inertial measurement unit, absolute position data output by the satellite positioning module, wheel speed data output by the wheel speed meter, and lane line observation data output by the lane line detection module; Based on the availability of high-precision map data, construct a corresponding lane line constraint model and convert lane line observation data into position error observation vectors or lateral deviation observation vectors. An extended Kalman filter is constructed to predict the state using inertial data and wheel speed data, and to update the state using absolute position data and position error observation vectors or lateral deviation observation vectors. During the state update process, the observation noise covariance matrix is ​​dynamically adjusted based on the lane line detection quality, and chi-square detection is performed based on the observation residuals. When an observation anomaly is detected, a degradation processing mechanism is activated to shield the abnormal observation or perform only time updates in order to output the fused positioning results.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the adaptive multi-source information fusion localization method based on lane line constraints as described in any one of claims 1 to 7.

Images