Vehicle state estimation method and device, computer equipment and storage medium

Abstract

The invention provides a vehicle state estimation method and device, computer equipment and a storage medium, and belongs to the field of vehicle state estimation.The method comprises the steps that point cloud information of a vehicle in a driving state is obtained; compressing the point cloud information, and extracting distance information of each point cloud in the point cloud information to obtain a distance image; performing feature point matching according to the distance image, and performing pose estimation on the vehicle according to a matching result to obtain a pose transformation matrix of the vehicle; performing time sequence analysis on the historical acceleration of the vehicle through an ARIMA model to obtain a predicted acceleration; and determining the current state of the vehicle according to the pose transformation matrix and the predicted acceleration. In this way, the accuracy of vehicle state estimation can be improved.

Description

Technical Field

[0001] The present invention belongs to the field of vehicle state estimation, and in particular relates to a vehicle state estimation method, device, computer equipment and storage medium. Background Art

[0002] Vehicle control is based on vehicle state parameters as a reference, so real-time and accurate acquisition of vehicle state parameters is a prerequisite for stability control. The estimation accuracy of vehicle state parameters will directly affect active and passive safety systems such as anti-lock braking systems, anti-skid control systems, and traction control systems, and thus indirectly affect the safety and stability control requirements of the vehicle.

[0003] Because accurate estimation of vehicle status is the premise and basis for achieving vehicle stability control, and the GPS-based status estimation method may fail in a small space due to lack of signal, therefore, low-frequency GPS signals are difficult to support vehicle stability control under high-speed conditions.

[0004] In recent years, it is more popular to use the entropy based on the higher-order moments of the error (minimum error entropy MEE, maximum correlation entropy MCC, etc.) as the cost function for filtering. Compared with the Kalman filter based on the MSE indicator, the Kalman filter based on the entropy indicator has greatly improved accuracy and robustness. Since the calculation complexity of Kalman filtering based on the MEE indicator is much more complicated than that based on the MCC, the calculation amount is large, so the Kalman based on the MCC is more widely used.

[0005] In the MCC-based Kalman filter, the kernel width is the only free parameter, which plays a decisive role in the existence of local optimal values, convergence speed, and robustness to non-Gaussian noise. However, most of the existing technical solutions determine the selection of a fixed-size kernel width for a specific non-Gaussian noise based on experience or trial and error. On the one hand, the non-Gaussian noise of the actual system is unknown, and the fixed-size kernel width determined based on a specific noise may have poor estimation performance under the actual non-Gaussian noise condition; on the other hand, the noise is not stable. For example, the initial noise is very large, and the noise tends to stabilize over time, so it is not easy to achieve the optimal using a fixed kernel width, which leads to inaccurate estimation of the vehicle state in the end. Summary of the invention

[0006] In order to solve the above-mentioned problem of inaccurate vehicle state estimation, the present invention provides a vehicle state estimation method, apparatus, computer equipment and storage medium.

[0007] In order to achieve the above object, the present invention provides the following technical solutions:

[0008] First, a vehicle state estimation method is provided, the method comprising:

[0009] Obtain the historical acceleration of the vehicle and the point cloud information of the vehicle in driving state;

[0010] The point cloud information is compressed, and the distance information of each point cloud in the compressed point cloud information is extracted to obtain a distance image;

[0011] Performing feature point matching according to the distance image, estimating the position and posture of the vehicle according to the matching result, and obtaining a position and posture transformation matrix of the vehicle;

[0012] Performing a time series analysis on the historical acceleration of the vehicle through an ARIMA model to obtain a predicted acceleration;

[0013] The current state of the vehicle is determined based on the posture transformation matrix and the predicted acceleration.

[0014] Optionally, compressing the point cloud information includes:

[0015] Performing a posture transformation on the point cloud information according to the driving data of the vehicle, and converting the point cloud information into a same origin coordinate system;

[0016] Determine the angle between adjacent laser beam scanning points of the point cloud information in the same coordinate system, and perform clustering processing on the point cloud information according to the angle;

[0017] The point cloud information after clustering processing is compressed from three-dimensional space to two-dimensional plane to obtain compressed two-dimensional point cloud data.

[0018] Optionally, edge point matching and plane point matching are performed according to the distance image, and the posture of the vehicle is estimated according to the matching results to obtain the posture transformation matrix of the vehicle, including:

[0019] Determine the smoothness of any point in each frame of the range image, and obtain the feature points of each frame of the range image through the smoothness;

[0020] Perform frame matching between the feature points at the current moment and the previous moment to obtain a comprehensive distance from any feature point to the image frame at the previous moment;

[0021] The comprehensive distance is iteratively optimized to obtain the posture transformation matrix.

[0022] Optionally, the feature points include edge points and plane points, and the feature points are matched between frames at the current moment and the previous moment, and the comprehensive distance from any feature point to the image frame at the previous moment is obtained, including:

[0023] Get the current time t k+1 The edge point set A point in And at the previous moment t kSelect from The nearest point and points The closest point in adjacent scan lines right Match points to lines and determine To the straight line The first distance;

[0024] Get the current time t k+1 The set of plane points A point in And at the previous moment t k Select from Closest point Find the same vertical resolution laser point set as Closest point and points The closest point between adjacent scanning laser lines right Perform point-to-surface matching to determine To The second distance;

[0025] The comprehensive distance from any feature point to the image frame at the previous moment is determined according to the first distance and the second distance.

[0026] Optionally, performing a time series analysis on the historical acceleration of the vehicle by using an ARIMA model to obtain a predicted acceleration includes:

[0027] An ARIMA model is constructed by using the historical acceleration of the vehicle at multiple consecutive moments and the corresponding continuous historical random disturbances;

[0028] The acceleration at the current moment is predicted through the ARIMA model to obtain the predicted acceleration.

[0029] Optionally, before determining the current state of the vehicle, an adaptive unscented Kalman filter may be used to filter the posture transformation matrix of the vehicle through the covariance of the dynamic adjustment process noise and the observation noise of the posture transformation matrix.

[0030] Secondly, a vehicle state estimation device is provided, the device comprising:

[0031] Point cloud acquisition module, used to obtain the historical acceleration of the vehicle and the point cloud information of the vehicle in driving state;

[0032] An information extraction module is used to compress the point cloud information, extract the distance information of each point cloud in the point cloud information, and obtain a distance image; perform feature point matching according to the distance image, estimate the vehicle posture according to the matching result, and obtain the vehicle posture transformation matrix; perform time series analysis on the historical acceleration of the vehicle through the ARIMA model to obtain the predicted acceleration;

[0033] The state determination module is used to determine the current state of the vehicle based on the posture transformation matrix and the predicted acceleration.

[0034] In addition, a computer-readable storage medium is provided, wherein the storage medium stores a computer program, and when the computer program is executed by a processor, the above-mentioned vehicle state estimation method is implemented.

[0035] Finally, a computer device is provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned vehicle state estimation method when executing the program.

[0036] The vehicle state estimation method provided by the present invention has the following beneficial effects:

[0037] First, by compressing the point cloud information to obtain the distance image, the amount of point cloud data can be significantly reduced, the speed of data transmission and processing can be improved, and because the amount of data to be processed is greatly reduced, it can respond quickly to environmental changes. According to the distance image, feature point matching and determining the pose transformation matrix can accurately determine the relative position and posture between objects, and the matching of feature points has a certain robustness to interference factors such as lighting changes, occlusion, and noise. In complex environments, feature point matching can ensure the stability and reliability of the system. Finally, time series analysis is introduced to predict the acceleration of the vehicle, which further reduces the error in the vehicle state estimation, which is conducive to improving the accuracy of the vehicle state estimation. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the embodiment of the present invention and its design scheme, the following briefly introduces the drawings required for this embodiment. The drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0039] Figure 1 The figure is a schematic diagram of the structural framework of a vehicle state estimation system provided according to an exemplary embodiment of the present invention.

[0040] Figure 2 The present invention is a flow chart of a vehicle state estimation method provided according to an exemplary embodiment of the present invention.

[0041] Figure 3 The figure is a schematic diagram of a point cloud clustering principle provided by the present invention according to an exemplary embodiment.

[0042] Figure 4 The present invention is a block diagram of a vehicle state estimation device according to an exemplary embodiment. DETAILED DESCRIPTION

[0043] In order to enable those skilled in the art to better understand the technical solution of the present invention and implement it, the present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and cannot be used to limit the scope of protection of the present invention.

[0044] The technical solutions provided by various embodiments of the present invention are described in detail below in conjunction with the accompanying drawings.

[0045] The present invention proposes a method for accurately estimating the driving state of a vehicle by fusing laser radar and IMU information. The structural framework of the vehicle state estimation system is as follows: Figure 1 shown.

[0046] In order to solve the problem of point cloud distortion caused by the laser radar moving with the vehicle, the IMU (Inertial Measurement Unit) is used to predict the position change of the laser radar to reduce the impact of point cloud motion distortion on the odometer accuracy; after the feature extraction process of the point cloud, the position and pose of the feature points between frames are estimated to obtain the positioning information of the vehicle. In the vehicle state estimation module, in order to reduce the model error and suppress the filter divergence, a noise adaptive module is designed in the unscented Kalman filter, and time series analysis is introduced to predict the longitudinal acceleration and lateral acceleration of the vehicle, further reducing the estimation error of the vehicle's longitudinal velocity, lateral velocity and yaw angular velocity;

[0047] First, the present invention provides a vehicle state estimation method, specifically as follows Figure 2 As shown, the following steps are included:

[0048] S201. Obtain historical acceleration of the vehicle and point cloud information of the vehicle in a driving state.

[0049] In this step, the historical acceleration of the vehicle can be obtained through the on-board accelerometer, and the point cloud information of the vehicle in the driving state can be obtained through the on-board laser radar.

[0050] S202: compress the point cloud information, extract the distance information of each point cloud in the point cloud information, and obtain a distance image.

[0051] Specifically, the point cloud information can be transformed in position and posture according to the vehicle's driving data, and the point cloud information can be converted to the same origin coordinate system; the angle between adjacent laser beam scanning points of the point cloud information in the same coordinate system is determined, and the point cloud information is clustered according to the angle; the point cloud information after clustering is compressed from three-dimensional space to two-dimensional plane. Among them, the vehicle's driving data can be obtained through IUM.

[0052] In one embodiment, the usual point cloud motion distortion correction method assumes that the vehicle motion in each frame is uniform, but this will lead to cumulative errors. In order to improve the accuracy of motion distortion correction of the corrected point cloud, it is necessary to obtain the position and posture changes of the vehicle within a frame, that is, the data output frequency of the sensor must be higher than that of the lidar. The IMU sampling frequency can reach 100 Hz or even higher, which can accurately reflect the movement of the vehicle within a frame of the lidar and perform high-precision local position estimation. By integrating the data collected by the IMU and linearly interpolating the IMU data, the position and posture of the vehicle at any time t∈(t k-1 ,t k The origin of the laser point cloud coordinates relative to time t k-1 The pose transformation matrix of the origin of the laser point cloud coordinates is calculated as follows:

[0053]

[0054] Among them, T i is the pose transformation matrix; t k ,t i ,t k-1 They are different time nodes respectively.

[0055] Through the pose transformation matrix, t i The laser point coordinates at the moment are converted to the first laser point coordinate system, that is:

[0056] P′ i =T i P i .

[0057] Among them, P i Yes i Laser point position at time; P′ i YesP i The position in the laser point cloud coordinate system at the next moment.

[0058] Using IMU to eliminate motion distortion of laser point clouds can effectively avoid problems such as map ghosting caused by distortion.

[0059] In another embodiment, when the laser radar is collecting information, small objects around the vehicle body may not appear repeatedly in adjacent frames, which will cause errors in the inter-frame matching process. To address this problem, the angles between adjacent laser beam scanning points are calculated and point clouds are clustered according to the angles. If the clustering result is less than 30 data points, they are removed as noise points to improve the accuracy of feature extraction and the efficiency of inter-frame matching. Figure 3 As shown, the red point is used as the center to traverse the surrounding areas, and the angle α between the center point and the adjacent points is calculated, and then the flatness Fangle is calculated to determine whether they belong to the same object, until the clustering condition is not met. In the figure, d1 and d2 are the lengths of the blue and orange laser beams, respectively. Alpha and alpha' are the angular resolutions in the horizontal and vertical directions, respectively. When the adjacent laser beam scanning points are on the same horizontal line, α is equal to the horizontal angular resolution alpha. When the adjacent laser beam scanning points are on the same vertical line, α is equal to the vertical angular resolution alpha'. During the iteration process, points with flatness greater than the threshold are put into the set When the iteration is finished the collection When the number is less than 30, The points in are removed as noise points.

[0060] The calculation process is:

[0061] h=d2sinα.

[0062] l2=d2cosα.

[0063] l1=d1-l2=d1-d2cosα.

[0064] The flatness Fangle is:

[0065] Fangle=atan 2 (d2 sinα,d1-d2 cosα).

[0066] After the original laser radar point cloud is clustered, the clustered point cloud is compressed from three-dimensional space to a two-dimensional plane, and the distance information of each point cloud is extracted and stored separately to form a distance image.

[0067] S203 , performing feature point matching according to the distance image, and estimating the position and posture of the vehicle according to the matching result to obtain a position and posture transformation matrix of the vehicle.

[0068] Specifically, the smoothness of any point in each frame of the distance image is determined, and the feature points of each frame of the distance image are obtained through the smoothness; the feature points are matched between the current moment and the previous moment to obtain the comprehensive distance from any feature point to the image frame of the previous moment; the comprehensive distance is iteratively optimized to obtain the posture transformation matrix.

[0069] For example, the distance image is divided into several sub-images in the horizontal direction, and each sub-image is traversed and feature extraction is performed in turn. i , at point p i Select multiple continuous points in the vertical direction to construct a point set S, and calculate the smoothness of each point in the point set:

[0070]

[0071] Set the threshold c th To distinguish the features of each point in the point set, the smoothness is greater than c th The points with the largest roughness are selected from each row in the distance image. Then select the plane feature point with the minimum roughness in the same way

[0072] By extracting the feature points of each laser radar frame and matching the feature points of adjacent frames, the laser radar posture transformation matrix can be solved. k Time and t k+1 The point clouds scanned by the lidar at each moment are and The edge feature points extracted by features are and The plane feature point sets are and The purpose of inter-frame matching is to obtain and The pose transformation matrix between them is also obtained and The pose transformation matrix between and Since the laser radar motion process produces distortion, t k+1 The edge feature point set at the moment And the plane feature point set Projection to t k time, respectively and Edge points are points in three-dimensional space that produce mutations in adjacent areas, so edge feature points use point-to-line matching; while plane points are points in three-dimensional space that are relatively flat in adjacent areas, so plane feature points use point-to-plane matching.

[0073] Specifically, get the current time t k+1 The edge point set A point in And at the previous moment tk Select from The nearest point and points The closest point in adjacent scan lines right Match points to lines and determine To the straight line The first distance; Get the current time t k+1 The set of plane points A point in And at the previous moment t k Select from Closest point Find the same laser point set with the same vertical resolution Closest point and points The closest point between adjacent scanning laser lines right Perform point-to-surface matching to determine To According to the first distance and the second distance, determine the comprehensive distance from any feature point to the image frame at the previous moment.

[0074] In one embodiment, t is selected k+1 The edge feature point set at the moment A point in In t k Select the time and Closest point and points The closest point in adjacent scan lines The coordinates are and point To the straight line The distance d E for:

[0075]

[0076] In another embodiment, select t k+1 The plane feature point set at time A point in In t k Select the time and Closest point Find the same laser point set with the same vertical resolution Closest point and points The closest point between adjacent scanning laser lines The coordinates of the four points are and point To The distance d P for:

[0077]

[0078] After matching the edge points and plane points, the point-to-line distance d is obtained E and the distance from point to surface d P , in order to obtain the optimal posture, combining the above point-to-line distance formula and point-to-plane distance formula, a cost function can be established:

[0079]

[0080] Using the Levenberg-Marquardt method to optimize the solution and establish the error function, we can get:

[0081]

[0082] Where: J is the Jacobian matrix; μ is the trust radius; D is the coefficient matrix, is the pose transformation matrix at time k+1; f is the cost function.

[0083] Construct the Lagrangian function:

[0084]

[0085] Where: λ is the coefficient.

[0086] After simplifying and derivatizing the above cost function, we can get:

[0087] (JJ T +λD T D)ΔT=-Jf.

[0088] Substituting the simplified derivative function into the cost function, we get:

[0089] ΔT=-(JJ T +λD T D) -1 (Jf).

[0090] Where: The coefficient matrix D can be approximately regarded as JJ T , so under the premise that the initial Jacobian matrix J is known, the residual ΔT can be solved. Substituting into the gradient descent formula is:

[0091]

[0092] Use two LM optimization iterations to solve until convergence and obtain the posture transformation matrix

[0093] S204: Perform time series analysis on the historical acceleration of the vehicle through an ARIMA model to obtain a predicted acceleration.

[0094] Specifically, an ARIMA model is constructed through the historical accelerations of the vehicle at multiple consecutive moments and the corresponding continuous historical random disturbances; the acceleration at the current moment is predicted through the ARIMA model to obtain the predicted acceleration.

[0095] The longitudinal acceleration and lateral acceleration of the vehicle are related to the driver's habits and the current road conditions, which are difficult to describe using mathematical language. The ARIMA model in time series analysis is introduced to predict the longitudinal acceleration and lateral acceleration of the vehicle, thereby improving the estimation accuracy. The ARIMA model achieves short-term prediction through p continuous historical observation data and q continuous historical random disturbances. Its mathematical expression is:

[0096] x k =φ0+φ1x k-1 +φ2x k-2 +…+φ p x k-p +ε k -θ1ε k-1 -…-θ q ε k-q .

[0097] Where: p and q represent the model order; x k Represents the predicted data; φ0, φ1, φ2 and φ p represents the model coefficient; ε k , ε k-1 and ε k-q represents continuous historical noise; θ1, θ2 and θ q Represents the model coefficients.

[0098] The 10 longitudinal and lateral acceleration data are respectively used as a set of longitudinal and lateral acceleration historical data, and the longitudinal and lateral acceleration values ​​at the next moment are predicted by ARIMA.

[0099] S205. Determine the current state of the vehicle according to the posture transformation matrix and the predicted acceleration.

[0100] The obtained pose transformation matrix Its essence is the transformation relationship between the vehicle coordinate system at the current moment and the previous moment, that is, the relative coordinates and relative angles of the vehicle at the current moment and the previous moment. If it is assumed that the vehicle is moving at a constant speed between each frame, then we can take The differential value of is used as the rough observation value Z of the vehicle longitudinal velocity, lateral velocity and yaw angular velocity k However, since there is no pose transformation matrix in the laser inertial odometer module Further optimization makes the observation quantity Z k It has large noise and cannot be directly used for vehicle longitudinal stability control, so the observation quantity Z k The estimation accuracy is improved by using unscented Kalman filtering with adaptive noise; at the same time, time series analysis is introduced in combination with IMU information to further improve the estimation accuracy.

[0101] In addition, before determining the current state of the vehicle, an adaptive unscented Kalman filter may be used to filter the posture transformation matrix of the vehicle through the covariance of the dynamic adjustment process noise and the observation noise of the posture transformation matrix.

[0102] Specifically, the nonlinear system of vehicle motion is:

[0103]

[0104] Where: X k and X k-1 They represent the motion state at time k and time k-1 respectively; F represents the motion model transfer matrix; Q k represents system noise; Z k represents the observed value at time k; H represents the observation model transfer matrix; R k represents the observation noise.

[0105] First, initialize and set the initial value of the motion state Error covariance P0, that is:

[0106]

[0107] At the same time, the motion state quantity Sigma sampling point is calculated, that is:

[0108]

[0109] Where: n represents the dimension of X; λ represents the scale adjustment factor; λ=1;

[0110]

[0111] Next, the prediction update is performed, and the Sigma sampling points are mapped by the motion model transfer matrix F, and then:

[0112]

[0113] Through weighted calculation, the predicted value of the motion state quantity at time k and the error covariance matrix are obtained as follows:

[0114]

[0115] Where: w i Represents the weight, that is:

[0116]

[0117] Then the observation is updated and the observation value Sigma sampling point is calculated:

[0118]

[0119] Where: By mapping the Sigma sampling points through the motion model transfer matrix H, we have:

[0120]

[0121] Then, a weighted calculation is performed to obtain the prediction and error covariance matrix of the observation value at time k and the cross-covariance matrix of the observation prediction error:

[0122]

[0123] Based on this, the Kalman gain can be calculated:

[0124]

[0125] Therefore, the state of the vehicle at time k can be obtained:

[0126]

[0127] in, is the prediction of the observed value at time k; is the Sigma sampling point; P Y is the error covariance matrix; P XY is the cross-covariance matrix of the observation errors.

[0128] Finally, the noise covariance is updated, and the covariance matrix of the system noise is corrected according to the difference between the observed value and the predicted value, thereby improving the estimation accuracy and suppressing the filter divergence. The difference between the observed value and the predicted value is defined as:

[0129]

[0130] Since d is affected by the state equation error and the initial estimate, the covariance matrix of the system noise can be estimated based on d. Therefore, the theoretical covariance matrix of d is:

[0131]

[0132] Due to the existence of observation noise, the actual value of the covariance matrix of d may be biased. Therefore, it is also necessary to correct the covariance matrix of d, that is:

[0133]

[0134] Where: M is the length of the sliding window.

[0135] By comparing the theoretical covariance matrix C of d k and the actual covariance matrix Size, to achieve the system noise covariance matrix Q k When Reduce the system noise covariance matrix Q k ;when In order to prevent the filter from diverging, the system noise covariance matrix Q k Intervention. Define the system noise covariance matrix Q k The adjustment factor is:

[0136]

[0137] System noise covariance matrix Q k and the observation noise covariance matrix R k Generally, the reverse correction is adopted. Therefore, the system noise covariance matrix Q k and the observation noise covariance matrix R k The correction processes are:

[0138]

[0139] R k =α k R k .

[0140] By adopting the above method, the point cloud information is first compressed to obtain a range image, which can significantly reduce the amount of point cloud data and improve the speed of data transmission and processing. Moreover, because the amount of data to be processed is greatly reduced, it can respond quickly to environmental changes. According to the range image, feature point matching is performed and the pose transformation matrix is ​​determined, the relative position and posture between objects can be accurately determined. The matching of feature points has a certain robustness to interference factors such as illumination changes, occlusion, and noise. In complex environments, feature point matching can ensure the stability and reliability of the system. Finally, time series analysis is introduced to predict the acceleration of the vehicle, which further reduces the error in the vehicle state estimation, which is conducive to improving the accuracy of the vehicle state estimation.

[0141] Secondly, the present invention also provides a vehicle state estimation device, such as Figure 4 As shown, including:

[0142] The point cloud acquisition module 401 is used to acquire the historical acceleration of the vehicle and the point cloud information of the vehicle in a driving state.

[0143] The information extraction module 402 is used to compress the point cloud information, extract the distance information of each point cloud in the point cloud information, and obtain a distance image; perform feature point matching based on the distance image, estimate the vehicle's posture based on the matching results, and obtain the vehicle's posture transformation matrix; perform time series analysis on the vehicle's historical acceleration through the ARIMA model to obtain a predicted acceleration.

[0144] The state determination module 403 is used to determine the current state of the vehicle according to the posture transformation matrix and the predicted acceleration.

[0145] By using the above device, the point cloud information is first compressed to obtain a distance image, which can significantly reduce the amount of point cloud data, improve the speed of data transmission and processing, and because the amount of data to be processed is greatly reduced, it can respond quickly to environmental changes. According to the distance image, feature point matching is performed and the pose transformation matrix is ​​determined, the relative position and posture between objects can be accurately determined, and the matching using feature points has a certain robustness against interference factors such as lighting changes, occlusion, and noise. In complex environments, the stability and reliability of the system can be ensured by matching feature points. Finally, time series analysis is introduced to predict the acceleration of the vehicle, which further reduces the error in the vehicle state estimation, which is conducive to improving the accuracy of the vehicle state estimation.

[0146] The present invention also provides a computer-readable storage medium, which stores a computer program, which can be used to execute the above Figure 1 The steps of a vehicle state estimation method are provided.

[0147] The present invention also provides a computer device. At the hardware level, the computer device includes a processor, an internal bus, a network interface, a memory, and a non-volatile memory. Of course, it may also include hardware required for other services. The processor reads the corresponding computer program from the non-volatile memory into the memory and then runs it to achieve the above Figure 1 The steps of a vehicle state estimation method are provided.

[0148] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, systems or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0149] The present invention is described with reference to flowcharts and / or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as a combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0150] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0151] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0152] It should be noted that the above-described specific implementations can enable those skilled in the art to more fully understand the invention, but do not limit the invention in any way. Therefore, although the invention has been described in detail in this specification, those skilled in the art should understand that the invention can still be modified or replaced by equivalents; and all technical solutions and improvements that do not deviate from the spirit and scope of the invention are included in the protection scope of the patent for the invention. Any figure mark in the claims should not be regarded as limiting the claims involved.

Claims

1. A vehicle state estimation method, characterized in that: The method comprises: Obtain the historical acceleration of the vehicle and the point cloud information of the vehicle in driving state; The point cloud information is compressed, and the distance information of each point cloud in the compressed point cloud information is extracted to obtain a distance image; Matching feature points of adjacent frames according to the range image, estimating the position and posture of the vehicle according to the matching result, and obtaining a position and posture transformation matrix of the vehicle; Performing a time series analysis on the historical acceleration of the vehicle through an ARIMA model to obtain a predicted acceleration; The current state of the vehicle is determined based on the posture transformation matrix and the predicted acceleration.

2. The vehicle state estimation method according to claim 1, characterized in that: The compressing the point cloud information comprises: Performing a posture transformation on the point cloud information according to the driving data of the vehicle, and converting the point cloud information into a same origin coordinate system; Determine the angle between adjacent laser beam scanning points of the point cloud information in the same coordinate system, and perform clustering processing on the point cloud information according to the angle; The point cloud information after clustering processing is compressed from three-dimensional space to two-dimensional plane to obtain compressed two-dimensional point cloud data.

3. The vehicle state estimation method according to claim 1, characterized in that: Matching feature points of adjacent frames is performed according to the distance image, and the posture of the vehicle is estimated according to the matching result to obtain the posture transformation matrix of the vehicle, including: Determine the smoothness of any point in each frame of the range image, and obtain the feature points of each frame of the range image through the smoothness; Perform frame matching between the feature points at the current moment and the previous moment to obtain a comprehensive distance from any feature point to the image frame at the previous moment; The comprehensive distance is iteratively optimized to obtain the posture transformation matrix.

4. The vehicle state estimation method according to claim 3, characterized in that: The feature points include edge points and plane points. The feature points are matched between frames at the current moment and the previous moment, and the comprehensive distance from any feature point to the image frame at the previous moment is obtained, including: Get the current time t k+1 The edge point set A point in And at the previous moment t k Select from The nearest point and points The closest point in adjacent scan lines right Match points to lines and determine To the straight line The first distance; Get the current time t k+1 The set of plane points A point in And at the previous moment t k Select from Closest point Find the same vertical resolution laser point set as Closest point and points The closest point between adjacent scanning laser lines right Perform point-to-surface matching to determine To The second distance; The comprehensive distance from any feature point to the image frame at the previous moment is determined according to the first distance and the second distance.

5. The vehicle state estimation method according to claim 1, characterized in that: The time series analysis of the historical acceleration of the vehicle by using the ARIMA model to obtain the predicted acceleration includes: An ARIMA model is constructed by using the historical acceleration of the vehicle at multiple consecutive moments and the corresponding continuous historical random disturbances; The acceleration at the current moment is predicted through the ARIMA model to obtain the predicted acceleration.

6. The vehicle state estimation method according to claim 1, characterized in that: Before determining the current state of the vehicle, an adaptive unscented Kalman filter may be used to filter the posture transformation matrix of the vehicle through the covariance of the dynamic adjustment process noise and the observation noise of the posture transformation matrix.

7. A vehicle state estimation device, characterized in that: The device comprises: Point cloud acquisition module, used to obtain the historical acceleration of the vehicle and the point cloud information of the vehicle in driving state; An information extraction module is used to compress the point cloud information, extract the distance information of each point cloud in the point cloud information, and obtain a distance image; perform feature point matching according to the distance image, estimate the vehicle posture according to the matching result, and obtain the vehicle posture transformation matrix; perform time series analysis on the historical acceleration of the vehicle through the ARIMA model to obtain the predicted acceleration; The state determination module is used to determine the current state of the vehicle based on the posture transformation matrix and the predicted acceleration.

8. A computer-readable storage medium, characterized in that: The storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

9. A computer device, characterized in that: The method comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the method according to any one of claims 1 to 7 when executing the program.

Images