A multi-sensor fusion positioning method and system in complex scenarios of driverless

Abstract

The present invention relates to a multi-sensor fusion positioning method and system under complex scenarios of driverless vehicles. The multi-sensor fusion positioning method uses a factor graph optimization method, that is, after converting the factor graph model constructed by positioning information into a non-linear problem by using maximum a posteriori probability estimation, the non-linear problem is solved to obtain the positioning result of the driverless vehicle, so as to fuse the observation data of the sensor positioning model, which can enhance the robustness of multi-sensor fusion positioning under complex scenarios of driverless vehicles while effectively improving the positioning accuracy and speed estimation accuracy of driverless vehicles.

Description

Technical Field

[0001] The present invention relates to the field of driverless positioning, and particularly to a multi-sensor fusion positioning method and system in complex driverless scenarios. Background Art

[0002] With the rapid development of the economic society, the global vehicle ownership has been increasing year by year, and people have put forward higher requirements for traffic safety and travel comfort. Driverless technology endows traditional vehicles with life and wisdom, enabling them to intelligently identify road conditions like humans and safely and autonomously drive to the destination, bringing safety and comfort to passengers. The intelligent and networked development of vehicles is also a new driving force for the development of the international automotive industry and the economic and social development.

[0003] The autonomous positioning of vehicles is an important link in realizing driverless driving, and it plays an important role in the environmental perception, dynamic planning, intelligent decision-making, and motion control of driverless driving. Only after the vehicle accurately determines its own position, attitude, and motion state can a series of driverless functions be realized. However, in complex driverless scenarios, sensors are interfered by the environment, their own accuracy is reduced, and the positioning information error output by the sensor positioning model also increases, resulting in the difficulty for the driverless vehicle positioning system to continuously output high-precision positioning information, attitude information, and speed information.

[0004] In the prior art, most of the multi-sensor fusion localization methods for driverless vehicles perform multi-sensor fusion using filtering methods. For example, Chinese Patent Publication No. CN110631593A, with the publication date of December 31, 2019, discloses a multi-sensor fusion localization method for autonomous driving scenarios. This method uses low-cost sensors combined with a vector map to achieve lane-level localization through an improved particle filter algorithm, which can ensure the localization accuracy and at the same time output high-frequency localization information with adjustable frequency, providing reference data for environmental perception and vehicle body control. Another example is Chinese Patent Publication No. CN108692701A, with the publication date of October 23, 2018, which discloses a multi-sensor fusion localization method for mobile robots based on a particle filter. This method updates the weights of particles in the particle filter using the observation model of each sensor and the sensor measurement data at the current moment, and finally uses the pose of the particle with the largest weight as the localization data of the robot at the current moment, greatly improving the accuracy of localization. Yet another example is Chinese Patent Publication No. CN107289948A, with the publication date of October 24, 2017, which discloses a multi-sensor fusion-based unmanned aerial vehicle navigation system and method. The invention discloses a multi-sensor fusion-based unmanned aerial vehicle navigation system and also discloses the corresponding multi-sensor fusion-based unmanned aerial vehicle navigation method. It uses an extended Kalman filter to fuse the data information obtained from the corresponding modules to obtain the current attitude and position of the unmanned aerial vehicle, which is suitable for the precise navigation of unmanned aerial vehicles. However, the above methods all use filtering methods for multi-sensor fusion and do not consider the problems of sensor failures and sensor localization errors in the multi-sensor localization system in complex driverless scenarios. In real driverless working conditions, due to environmental interference, the localization accuracy and attitude estimation accuracy of the filtering-based multi-sensor fusion localization method are limited, and the robustness is poor.

[0005] Therefore, there is an urgent need in the art to provide a multi-sensor fusion localization method in complex driverless scenarios with strong robustness to effectively improve the localization accuracy and speed estimation accuracy of driverless vehicles. Summary of the Invention

[0006] The object of the present invention is to provide a multi-sensor fusion localization method in complex driverless scenarios with strong robustness, which can effectively improve the localization accuracy and speed estimation accuracy of driverless vehicles.

[0007] To achieve the above object, the present invention provides the following solution:

[0008] A multi-sensor fusion localization method in complex driverless scenarios, comprising:

[0009] Constructing a factor graph model according to the localization information; the localization information includes: the localization information obtained based on the pre-integration model of the inertial measurement unit, the localization information obtained based on the global navigation satellite system localization model, and / or the localization information obtained based on the lidar odometer.

[0010] Convert the factor graph model into a non - linear problem using maximum a posteriori probability estimation;

[0011] Solve the non - linear problem to obtain the positioning result of the driverless vehicle; the positioning result includes: positioning information, attitude information, and speed information.

[0012] Preferably, constructing the factor graph model according to the positioning information specifically includes:

[0013] Obtain a fault detection index using a fault detection method;

[0014] Judge whether the global navigation satellite system has a fault according to the fault detection index to obtain a judgment result;

[0015] When the judgment result is that the global navigation satellite system has a fault, construct a factor graph model according to the positioning information obtained from the pre - integrated model based on the inertial measurement unit and the positioning information obtained from the lidar odometer;

[0016] When the judgment result is that the global navigation satellite system has no fault, construct a factor graph model according to the positioning information obtained from the pre - integrated model based on the inertial measurement unit, the positioning information obtained from the global navigation satellite system positioning model, and the positioning information obtained from the lidar odometer.

[0017] Preferably, the obtaining a fault detection index using a fault detection method specifically includes:

[0018] Construct a fault detection function using the residual sequence of the observation data and the residual covariance matrix of the observation data; both the residual sequence of the observation data and the residual covariance matrix of the observation data are determined based on the preset observation data of the driverless vehicle;

[0019] Perform normalization processing on the fault detection function to obtain a normalization processing result;

[0020] Perform weighted processing on the normalization processing result to obtain the fault detection index.

[0021] Preferably, the judging whether the global navigation satellite system has a fault according to the fault detection index to obtain a judgment result specifically includes:

[0022] When the fault detection index is less than 1, it is determined that the global navigation satellite system is working normally;

[0023] When the fault detection index is greater than or equal to 1, it is determined that the global navigation satellite system has a fault. At this time, the fault detection index is set to 1.

[0024] Preferably, it further includes:

[0025] The observation noise and the error covariance matrix of the observation noise of the lidar odometer at the current moment are corrected by using the noise mean and the noise covariance matrix of the lidar odometer at the previous moment.

[0026] The mean of the observation noise and the covariance matrix of the observation noise of the corrected lidar odometer are determined by using adaptive filtering.

[0027] According to the specific embodiments provided by the present invention, the following technical effects are disclosed by the present invention:

[0028] The multi-sensor fusion positioning method for unmanned driving in complex scenarios provided by the present invention, by adopting the factor graph optimization method, that is, converting the factor graph model constructed by positioning information into a non-linear problem by using maximum a posteriori probability estimation and then solving the non-linear problem to obtain the positioning result of the unmanned vehicle, fuses the observation data of the sensor positioning model, and can enhance the robustness of multi-sensor fusion positioning in complex scenarios of unmanned driving while effectively improving the positioning accuracy and speed estimation accuracy of unmanned vehicles.

[0029] In addition, corresponding to the multi-sensor fusion positioning method for unmanned driving in complex scenarios provided above, the present invention also provides the following implementation systems:

[0030] A multi-sensor fusion positioning system for unmanned driving in complex scenarios, comprising:

[0031] A factor graph model construction module, configured to construct a factor graph model according to positioning information; the positioning information includes: positioning information obtained based on an inertial measurement unit pre-integration model, positioning information obtained based on a global navigation satellite system positioning model, and / or positioning information obtained based on a lidar odometer.

[0032] A non-linear problem conversion module, configured to convert the factor graph model into a non-linear problem by using maximum a posteriori probability estimation.

[0033] A positioning result determination module, configured to solve the non-linear problem to obtain the positioning result of the unmanned vehicle; the positioning result includes: positioning information, attitude information, and speed information.

[0034] Preferably, the factor graph model construction module includes:

[0035] A fault detection index acquisition unit, configured to acquire a fault detection index by using a fault detection method.

[0036] A judgment unit, configured to judge whether a global navigation satellite system fails according to the fault detection index to obtain a judgment result.

[0037] A first factor graph model construction unit, configured to construct a factor graph model according to the positioning information obtained based on the inertial measurement unit pre-integration model and the positioning information obtained based on the lidar odometer when the judgment result is that the global navigation satellite system fails;

[0038] A second factor graph model construction unit, configured to construct a factor graph model according to the positioning information obtained based on the inertial measurement unit pre-integration model, the positioning information obtained based on the global navigation satellite system positioning model, and the positioning information obtained based on the lidar odometer when the judgment result is that the global navigation satellite system does not fail.

[0039] Preferably, the fault detection index acquisition unit includes:

[0040] A fault detection function construction subunit, configured to construct a fault detection function by using the residual sequence of the observation data and the residual covariance matrix of the observation data; both the residual sequence of the observation data and the residual covariance matrix of the observation data are determined based on the preset observation data of the unmanned vehicle;

[0041] A normalization processing subunit, configured to perform normalization processing on the fault detection function to obtain a normalization processing result;

[0042] A fault detection index determination subunit, configured to perform weighted processing on the normalization processing result to obtain the fault detection index.

[0043] Preferably, the judgment unit includes:

[0044] A first judgment subunit, configured to determine that the global navigation satellite system is working properly when the fault detection index is less than 1;

[0045] A second judgment subunit, configured to determine that the global navigation satellite system fails when the fault detection index is greater than or equal to 1. At this time, the fault detection index is set to 1.

[0046] Preferably, it further includes:

[0047] A correction module, configured to correct the observation noise and the error covariance matrix of the observation noise of the current lidar odometer by using the noise mean and the noise covariance matrix of the previous lidar odometer;

[0048] A noise mean-covariance matrix determination module, configured to determine the mean and the covariance matrix of the observation noise of the corrected lidar odometer by using adaptive filtering.

[0049] Since the technical effects achieved by the multi-sensor fusion positioning system in complex scenarios of driverless vehicles provided by the present invention are the same as those achieved by the multi-sensor fusion positioning method in complex scenarios of driverless vehicles provided above, they will not be elaborated here. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required in the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0051] Figure 1 is a flowchart of the multi-sensor fusion positioning method in complex scenarios of driverless vehicles provided by the present invention;

[0052] Figure 2 is a local factor graph provided by an embodiment of the present invention;

[0053] Figure 3 is a flowchart for correcting the error of lidar odometry provided by an embodiment of the present invention;

[0054] Figure 4 is an overall workflow framework diagram of the multi-sensor fusion positioning method in complex scenarios of driverless vehicles provided by an embodiment of the present invention;

[0055] Figure 5 is a schematic structural diagram of the multi-sensor fusion positioning system in complex scenarios of driverless vehicles provided by the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0056] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the protection scope of the present invention.

[0057] The purpose of the present invention is to provide a multi-sensor fusion positioning method in complex scenarios of driverless vehicles with strong robustness, which can effectively improve the positioning accuracy of driverless vehicles.

[0058] In order to make the above objects, features, and advantages of the present invention more obvious and understandable, the present invention will be further described in detail below with reference to the drawings and specific embodiments.

[0059] As Figure 1 shown, the multi-sensor fusion positioning method in complex scenarios of driverless vehicles provided by the present invention includes:

[0060] Step 100: Construct a factor graph model according to the positioning information. The positioning information includes: the positioning information obtained based on the inertial measurement unit pre-integration model, the positioning information obtained based on the global navigation satellite system positioning model, and / or the positioning information obtained based on the lidar odometer. For example, use the positioning information obtained by three sensors (inertial measurement unit, global navigation satellite system, and lidar odometer) to build a corresponding factor graph. As Figure 2 shown, the factor graph includes factor nodes and state variable nodes. Among them, the state variable nodes include the unmanned vehicle pose, the motion speed, and the zero bias of the inertial measurement unit, denoted as [p, v, b] T . Figure 2 only considers the state variable nodes and factor nodes at four moments. Among them, the imu factor node is a binary factor converted from the inertial measurement unit sub-positioning system and its error model, the gnss_p factor node is a unary factor converted from the global navigation satellite system sub-positioning system and its error model, the gnss_v factor node is a unary factor converted from the speed information output by the global navigation satellite system, the lo factor node is a unary factor converted from the lidar sub-positioning system and its error model, and the bias factor node is a binary factor converted from the zero bias and zero bias noise of the inertial measurement unit.

[0061] Step 101: Use the maximum a posteriori probability estimation to transform the factor graph model into a non-linear problem. For example, use the formula to obtain the global function X of the maximum a posteriori probability MAP , and obtain the sensor observation data z through the formula . i The non-standard posterior probability of the corresponding state variable X i is Substitute the formula into the formula and take the negative logarithm to obtain a non-linear least squares problem. The non-linear least squares problem is: where h i (X i ) is the observation function between the state variable X i and the sensor observation quantity z i , and is the squared Mahalanobis norm of the observation quantity z i .

[0062] Step 102: Solve the non-linear problem to obtain the positioning result of the unmanned vehicle. The positioning result includes: positioning information, attitude information, and speed information.

[0063] Based on the above process, the present invention can use the unmanned vehicle pose information and its error covariance matrix output by the factor graph optimization process as the data basis for adaptive filtering noise estimation and fault detection, so as to be used for the current frame lidar odometry noise estimation and the next frame global navigation satellite system fault detection. Based on this, during the construction process of the factor graph model using the positioning information, it is also necessary to detect the faults of the global navigation satellite system. This detection process includes:

[0064] Step 1000: Obtain the fault detection index by using the fault detection method. The implementation process of this step is preferably:

[0065] ① Construct a fault detection function using the residual sequence of the observation data and the residual covariance matrix of the observation data. Both the residual sequence of the observation data and the residual covariance matrix of the observation data are determined based on the preset observation data of the unmanned vehicle. Among them, the constructed fault detection function follows the chi-square distribution with 3 degrees of freedom as:

[0066] e k =Z k -H k X k,k-1

[0067]

[0068]

[0069] In the formula, Z k is the observation value of the global navigation satellite system positioning model. H k is the observation matrix. e k is the predicted residual sequence. X k,k-1 is the state prediction value at the current moment. P k,k-1 is the prediction covariance matrix. R k is the measurement noise covariance matrix. P e,k is the covariance matrix of the residual sequence.

[0070] ② Normalize the fault detection function to obtain the normalization result. The specific implementation process is: Set two levels of thresholds T D1 and T D2 for hierarchical fault detection. T D1 is the threshold determined according to the significance level in the χ 2 test method, that is, the false alarm rate. When the fault detection index satisfies λ k < T D1 , the state of the global navigation satellite system can be judged as fault-free. When λ k ≥ T D1 , in order to reduce the impact of the false alarm phenomenon on the multi-sensor fusion positioning system, a secondary fault detection threshold T D2. In secondary fault detection, when λ k > T D2 , it is directly determined that the global navigation satellite system has a fault. When T D1 ≤λ k ≤T D2 , the probability of false alarm in the fault detection method is relatively high. Map λ k to the interval [0, 1] and perform normalization processing to obtain

[0071] ③ Perform weighted processing on the normalization result to obtain a fault detection index. Considering more rapid identification of faults in the global navigation satellite system, perform weighted processing on the current target value to be detected and the previous detection target value. Only when the continuous fault detection index λ k is less than the fault release threshold can it be determined that the fault of the global navigation satellite system is released. The weighted processing method is as shown in the following formula:

[0072]

[0073] where, α1 and α2 are weighting coefficients, representing the proportions of the current detection index and the previous detection index in the fault detection function.

[0074] Step 1001: Judge whether the global navigation satellite system has a fault according to the fault detection index to obtain a judgment result.

[0075] Step 1002: When the judgment result is that the global navigation satellite system has a fault, construct a factor graph model according to the positioning information obtained from the inertial measurement unit pre-integration model and the positioning information obtained from the lidar odometer.

[0076] Step 1003: When the judgment result is that the global navigation satellite system has no fault, construct a factor graph model according to the positioning information obtained from the inertial measurement unit pre-integration model, the positioning information obtained from the global navigation satellite system positioning model, and the positioning information obtained from the lidar odometer.

[0077] For example, when the fault detection index is less than 1, it is determined that the global navigation satellite system is working normally. When the fault detection index is greater than or equal to 1, it is determined that the global navigation satellite system has a fault. At this time, the fault detection index is set to 1. After the global navigation satellite system subsystem is determined to have a fault, only when the test value is less than the fault release threshold ε can the fault alarm be released, that is, λ k < ε.

[0078] Further, in order to improve robustness, the multi-sensor fusion positioning method in the complex scene of driverless provided by the present invention further includes:

[0079] Use the noise mean and noise covariance matrix of the lidar odometry at the previous moment to correct the observation noise and the error covariance matrix of the observation noise of the lidar odometry at the current moment.

[0080] Use adaptive filtering to determine the mean of the observation noise and the covariance matrix of the observation noise of the corrected lidar odometry.

[0081] In the present invention, mainly an adaptive filtering method is used to estimate the noise of the observation values of the lidar sub-positioning system in the multi-sensor positioning system. The adaptive filtering noise estimation method will output the noise mean and noise covariance of the observation values of the lidar sub-positioning system at the current moment, where the noise mean reflects the error of the lidar odometry. As Figure 3 shown, the adaptive filtering noise estimation process is as follows:

[0082] ① Use the observation value z k of the lidar sub-positioning system at the current moment and the current moment unmanned vehicle pose information output by factor graph optimization to calculate the noise mean of the lidar sub-positioning system.

[0083] ② Construct the residual e k of the observation value of the lidar sub-positioning system.

[0084] ③ Use the residual e k and the error covariance matrix of the current moment unmanned vehicle pose output by factor graph optimization to calculate the noise covariance matrix of the lidar sub-positioning system.

[0085] Among them, the formula used in the noise estimation process is:

[0086]

[0087]

[0088]

[0089] In the formula, e k is the residual of the lidar odometry observation value. H k is the measurement matrix of the lidar odometry. is the estimation result of the unmanned vehicle pose at the current moment. is the error covariance matrix of the unmanned vehicle pose estimation at the current moment. is the estimation result of the observation noise mean of the lidar odometry at the previous moment. is the estimation result of the observation noise covariance matrix of the lidar odometry at the previous moment.

[0090] In summary, the overall workflow framework of the multi-sensor fusion positioning method in complex scenarios for driverless vehicles provided by the present invention is as Figure 4 shown.

[0091] In addition, corresponding to the multi-sensor fusion positioning method in complex scenarios for driverless vehicles provided above, the present invention also provides a multi-sensor fusion positioning system in complex scenarios for driverless vehicles, as Figure 5 shown. The multi-sensor fusion positioning system includes: a factor graph model construction module 1, a non-linear problem transformation module 2, and a positioning result determination module 3.

[0092] The factor graph model construction module 1 is used to construct a factor graph model according to the positioning information. The positioning information includes: the positioning information obtained based on the inertial measurement unit pre-integration model, the positioning information obtained based on the global navigation satellite system positioning model, and / or the positioning information obtained based on the lidar odometer.

[0093] The non-linear problem transformation module 2 is used to transform the factor graph model into a non-linear problem by using the maximum a posteriori probability estimation.

[0094] The positioning result determination module 3 is used to solve the non-linear problem to obtain the positioning result of the driverless vehicle. The positioning result includes: positioning information, attitude information, and speed information.

[0095] Furthermore, in order to improve the positioning accuracy of the driverless vehicle, the factor graph model construction module 1 provided above in the present invention may further be provided with: a fault detection index acquisition unit, a judgment unit, a first factor graph model construction unit, and a second factor graph model construction unit.

[0096] The fault detection index acquisition unit is used to obtain the fault detection index by using a fault detection method.

[0097] The judgment unit is used to judge whether the global navigation satellite system has a fault according to the fault detection index to obtain a judgment result.

[0098] The first factor graph model construction unit is used to construct a factor graph model according to the positioning information obtained based on the inertial measurement unit pre-integration model and the positioning information obtained based on the lidar odometer when the judgment result is that the global navigation satellite system has a fault.

[0099] The second factor graph model construction unit is used to construct a factor graph model according to the positioning information obtained based on the inertial measurement unit pre-integration model, the positioning information obtained based on the global navigation satellite system positioning model, and the positioning information obtained based on the lidar odometer when the judgment result is that the global navigation satellite system has no fault.

[0100] Among them, in order to improve robustness, the above-mentioned fault detection index acquisition unit may further include: a fault detection function construction subunit, a normalization processing subunit, and a fault detection index determination subunit.

[0101] The fault detection function construction subunit is used to construct a fault detection function by using the residual sequence of the observation data and the residual covariance matrix of the observation data. Both the residual sequence of the observation data and the residual covariance matrix of the observation data are determined based on the preset observation data of the unmanned vehicle.

[0102] The normalization processing subunit is used to perform normalization processing on the fault detection function to obtain a normalization processing result.

[0103] The fault detection index determination subunit is used to perform weighted processing on the normalization processing result to obtain a fault detection index.

[0104] Furthermore, also in order to improve robustness, the above-mentioned judgment unit adopted in the present invention may include: a first judgment subunit and a second judgment subunit.

[0105] Among them, the first judgment subunit is used to determine that the global navigation satellite system is working normally when the fault detection index is less than 1.

[0106] The second judgment subunit is used to determine that the global navigation satellite system has a fault when the fault detection index is greater than or equal to 1. At this time, the fault detection index is set to 1.

[0107] Furthermore, in order to improve the positioning accuracy while improving robustness, the above-mentioned multi-sensor fusion positioning system in complex scenarios of unmanned driving provided by the present invention may further include: a correction module and a noise mean-covariance matrix determination module.

[0108] Among them, the correction module is used to correct the observation noise and the error covariance matrix of the observation noise of the lidar odometer at the current moment by using the noise mean and the noise covariance matrix of the lidar odometer at the previous moment.

[0109] The noise mean-covariance matrix determination module is used to determine the mean of the observation noise and the covariance matrix of the observation noise of the corrected lidar odometer by using adaptive filtering.

[0110] The various embodiments in this specification are described in a progressive manner. Each embodiment focuses on the differences from other embodiments. The same or similar parts among the embodiments can be referred to each other. For the system disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the description of the method part.

[0111] In this article, specific examples are used to elaborate on the principles and implementation manners of the present invention. The description of the above embodiments is only used to help understand the method and its core idea of the present invention; at the same time, for those of ordinary skill in the art, according to the idea of the present invention, there will be changes in the specific implementation manners and application scopes. In summary, the content of this specification should not be construed as a limitation to the present invention.

Claims

1. A multi-sensor fusion localization method in complex scenarios for driverless vehicles, characterized in that, Including: Constructing a factor graph model according to the positioning information; The positioning information includes: positioning information obtained based on an inertial measurement unit pre-integration model, positioning information obtained based on a global navigation satellite system positioning model, and / or positioning information obtained based on a lidar odometer; Converting the factor graph model into a non-linear problem by using maximum a posteriori probability estimation; Solving the non-linear problem to obtain the positioning result of the driverless vehicle; the positioning result includes: positioning information, attitude information, and speed information; Using the pose information of the driverless vehicle and its error covariance matrix output by the factor graph optimization process as the data basis for adaptive filtering noise estimation and fault detection, for the current frame lidar odometer noise estimation and the next frame global navigation satellite system fault detection; The constructing a factor graph model according to the positioning information specifically includes: Adopting a fault detection method to obtain a fault detection index; Judging whether the global navigation satellite system has a fault according to the fault detection index to obtain a judgment result; When the judgment result is that the global navigation satellite system has a fault, constructing a factor graph model according to the positioning information obtained based on the inertial measurement unit pre-integration model and the positioning information obtained based on the lidar odometer; When the judgment result is that the global navigation satellite system has no fault, constructing a factor graph model according to the positioning information obtained based on the inertial measurement unit pre-integration model, the positioning information obtained based on the global navigation satellite system positioning model, and the positioning information obtained based on the lidar odometer; The adopting a fault detection method to obtain a fault detection index specifically includes: Constructing a fault detection function by using the residual sequence of the observation data and the residual covariance matrix of the observation data; both the residual sequence of the observation data and the residual covariance matrix of the observation data are determined based on the preset observation data of the driverless vehicle; Performing normalization processing on the fault detection function to obtain a normalization processing result; Performing weighted processing on the normalization processing result to obtain the fault detection index.

2. The multi-sensor fusion positioning method in an unmanned complex scenario according to claim 1, wherein, The judging whether the global navigation satellite system has a fault according to the fault detection index to obtain a judgment result specifically includes: When the fault detection index is less than 1, it is determined that the global navigation satellite system is working normally; When the fault detection index is greater than or equal to 1, it is determined that the global navigation satellite system has a fault. At this time, the fault detection index is set to 1.

3. The multi-sensor fusion positioning method in an unmanned complex scenario according to claim 1, wherein, Also including: Using the noise mean and noise covariance matrix of the previous moment lidar odometer to correct the observation noise and the error covariance matrix of the observation noise of the current moment lidar odometer; Using adaptive filtering to determine the mean and covariance matrix of the observation noise of the corrected lidar odometer.

4. A multi-sensor fusion positioning system in a complex unmanned driving scenario, characterized in that, Including: A factor graph model construction module for constructing a factor graph model according to the positioning information; The positioning information includes: positioning information obtained based on an inertial measurement unit pre-integration model, positioning information obtained based on a global navigation satellite system positioning model, and / or positioning information obtained based on a lidar odometer; A non - linear problem transformation module, which is used to transform the factor graph model into a non - linear problem by using maximum a posteriori probability estimation; A positioning result determination module, which is used to solve the non - linear problem to obtain the positioning result of the driverless vehicle; the positioning result includes: positioning information, attitude information and speed information; Using the pose information of the driverless vehicle and its error covariance matrix output by the factor graph optimization process as the data basis for adaptive filtering noise estimation and fault detection, for the current - frame lidar odometry noise estimation and the next - frame global navigation satellite system fault detection; The factor graph model construction module includes: A fault detection index acquisition unit, which is used to obtain a fault detection index by using a fault detection method; A judgment unit, which is used to judge whether the global navigation satellite system has a fault according to the fault detection index, and obtain a judgment result; A first factor graph model construction unit, which is used to construct a factor graph model according to the positioning information obtained from the inertial measurement unit pre - integration model and the positioning information obtained from the lidar odometry when the judgment result is that the global navigation satellite system has a fault; A second factor graph model construction unit, which is used to construct a factor graph model according to the positioning information obtained from the inertial measurement unit pre - integration model, the positioning information obtained from the global navigation satellite system positioning model and the positioning information obtained from the lidar odometry when the judgment result is that the global navigation satellite system has no fault; The fault detection index acquisition unit includes: A fault detection function construction sub - unit, which is used to construct a fault detection function by using the residual sequence of the observation data and the residual covariance matrix of the observation data; both the residual sequence of the observation data and the residual covariance matrix of the observation data are determined based on the preset observation data of the driverless vehicle; A normalization processing sub - unit, which is used to perform normalization processing on the fault detection function to obtain a normalization processing result; A fault detection index determination sub - unit, which is used to perform weighted processing on the normalization processing result to obtain the fault detection index.

5. The multi-sensor fusion positioning system in an unmanned complex scenario according to claim 4, characterized in that The judgment unit includes: A first judgment sub - unit, which is used to determine that the global navigation satellite system is working properly when the fault detection index is less than 1; A second judgment sub - unit, which is used to determine that the global navigation satellite system has a fault when the fault detection index is greater than or equal to 1. At this time, the fault detection index is set to 1.

6. The multi-sensor fusion positioning system in an unmanned complex scenario according to claim 4, wherein It also includes: A correction module, which is used to correct the observation noise and the error covariance matrix of the observation noise of the current - moment lidar odometry by using the noise mean and noise covariance matrix of the previous - moment lidar odometry; A noise mean - covariance matrix determination module, which is used to determine the mean and covariance matrix of the observation noise of the corrected lidar odometry by using adaptive filtering.

Images