Method and system for determining alignment parameters of a radar sensor
Through the combination of computer hardware components and speed sensors, the misalignment parameters of the radar sensor are determined, which solves the error problem caused by chassis misalignment in automotive applications, and improves detection accuracy and alignment stability.
Patent Information
- Application Number
- CN202110488715.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Priority Date
- 2021-03-31
- Filing Date
- 2021-05-06
- Publication Date
- 2024-11-12
- Estimated Expiration
- 2041-05-06
AI Technical Summary
In automotive applications, existing radar sensors cannot accurately compensate for pitch angle or elevation errors caused by heavy objects and yaw angle, azimuth angle or roll angle errors caused by vehicle chassis misalignment, affecting the detection accuracy during driving.
Methods performed by computer hardware components, using radar sensors to measure distance change rate, azimuth angle and elevation measurements, combined with the speed information determined by the speed sensor, calculate misalignment parameters, including azimuth angle, elevation angle and rolling angle misalignment, and through optimization Methods and nonlinear least squares regression methods are corrected to determine system parameters and adjust the alignment of radar sensors.
Improve the detection quality of radar sensors, improve the overall execution of alignment, ensure the accuracy and stability of radar processing, and enhance the ability to observe the environment.
Smart Images

Figure CN113608182B_ABST
Abstract
Description
Technical Field
[0001] The present disclosure relates to methods and systems for determining alignment parameters of a radar sensor. Background Art
[0002] For example, radar sensors in automotive applications need to be aligned with the chassis of the host vehicle in order to accurately know the location of detected objects. The alignment procedure performed when assembling the host vehicle cannot compensate for pitch or elevation errors caused by heavy objects, and yaw, azimuth or roll errors caused by misalignment of the host vehicle's wheels or chassis, which may affect the "acquisition" or "dog-tracking" of the host vehicle while in motion.
[0003] Therefore, there is a need to provide a reliable determination of alignment parameters. Summary of the invention
[0004] The present disclosure provides a computer-implemented method, a radar system, a computer system, a vehicle and a non-transitory computer-readable medium. Implementations are given in the dependent claims, the description and the drawings.
[0005] In one aspect, the present disclosure relates to a computer-implemented method for determining alignment parameters of a radar sensor, the method comprising the following steps performed (in other words: implemented) by computer hardware components: determining measurement data using the radar sensor, the measurement data comprising range-rate measurements, azimuth measurements, and elevation measurements; determining a speed (in other words: velocity) of the radar sensor (e.g., using a speed sensor or together with misalignment parameters); and determining misalignment parameters based on the measurement data and the speed, the misalignment parameters comprising azimuth misalignment, elevation misalignment, and roll misalignment.
[0006] It should be appreciated that determining the velocity of the radar sensor may be performed explicitly (eg, using a velocity sensor) or implicitly (eg, by including the velocity in an optimization problem as described herein for the 2D case and 3D).
[0007] According to various embodiments, the misalignment parameter may be determined by comparing a speed determined by a speed sensor (which may be independent of the radar sensor) with speed information determined based on radar measurement data.
[0008] The measurement data may relate to a radar sensor, for example a radar sensor mounted on a vehicle (e.g. an at least partially autonomous vehicle, such as a car) (e.g. a chassis of the vehicle). The speed sensor may be external to the radar sensor; for example, the speed sensor may be a speed sensor of the vehicle, for example based on odometer measurements, gyroscope measurements and / or acceleration measurements.
[0009] According to another aspect, the misalignment parameter further includes a speed-scaling error.
[0010] It has been found that providing three-axis automatic alignment with velocity adjustment error estimation can improve the detection quality of radar sensors.
[0011] According to another aspect, the computer-implemented method further comprises the following steps performed by the computer hardware component: determining system parameters, the system parameters comprising at least one of a desired azimuth installation angle, a desired elevation installation angle, and a desired roll installation angle; wherein the misalignment parameters are also determined based on the system parameters.
[0012] System parameters may be set by a manufacturer of a system (eg, a vehicle) in which the radar sensor is provided. The system parameters may describe or define how the radar sensor is provided (eg, installed) in a vehicle.
[0013] According to another aspect, the measurement data may relate to a plurality of objects external to the radar sensor. For example, measurements relating to at least three objects may be provided. For example, measurements relating to at least four objects may be provided. For example, measurements relating to at least five objects may be provided. The number of objects may be selected such that the system for determining the misalignment parameter is at least not under-determined, for example, such that the system is over-determined (i.e., more measurements may be available than the unknown misalignment parameter to be determined; this may allow implicit corrections for measurement inaccuracies or measurement errors, for example by using regression methods to determine the misalignment parameter).
[0014] The object may be disposed outside of a vehicle provided with the radar sensor.
[0015] According to another aspect, the plurality of objects are stationary. Therefore, the speed determined by the speed sensor may be considered to be the same as the (relative) speed of the stationary objects (relative to the radar sensor). This may be used to determine the misalignment parameter by determining the misalignment parameter to bring the speed determined based on the radar measurements into line with the speed determined by the speed sensor (e.g., to minimize the difference between the speed determined based on the radar measurements and the speed determined by the speed sensor).
[0016] According to another aspect, the misalignment parameter is determined based on an iterative method.
[0017] A closed-form solution for determining the misalignment parameters may not be available, or may be numerically unstable. An iterative method may be used starting with an initial estimate (or guess) (e.g., setting all misalignment parameters to default values (e.g., 0)). The estimates may be updated iteratively. After one iteration, the misalignment parameters may already have a reasonable accuracy. Depending on resources such as time and computing power, more than one iteration may be performed. Iterations may continue until a predetermined stopping criterion is reached, such as a maximum time that can be used to determine the misalignment parameters, or a predetermined accuracy of the misalignment parameters (which may be determined, for example, based on how many (e.g., absolute or relative) estimates of the misalignment parameters change from one iteration to the next).
[0018] According to another aspect, the misalignment parameter is determined based on an optimization method.
[0019] The optimization method can optimize the misalignment parameter to minimize the difference between the speed measured by the speed sensor and the speed determined based on the measurement data. The optimization method can determine the misalignment parameter from an overdetermined system (where more measurements (e.g., related to multiple objects) are available than required for a closed-form determination of the misalignment parameter). The optimization method can use a Jacobian matrix, including partial derivatives of the modeled output with respect to the misalignment parameter.
[0020] According to another aspect, the misalignment parameter is determined based on a nonlinear least squares regression method. According to another aspect, the misalignment parameter is determined based on a nonlinear total least squares regression method.
[0021] According to another aspect, the misalignment parameter is determined based on filtering. For example, the misalignment parameter that has been determined based on an optimization or regression method may be filtered, such as a Kalman filter, to further improve the misalignment parameter.
[0022] According to another aspect, the measurement data is determined using a plurality of radar sensors (which may include the radar sensor and at least one other radar sensor).
[0023] According to another aspect, velocity and misalignment parameters of the radar sensor are simultaneously determined by solving an optimization problem.
[0024] Using methods according to various embodiments, the overall performance of alignment may be improved, which may have a positive impact on the performance of further radar processing.
[0025] In another aspect, the present disclosure relates to a computer-implemented method for determining corrected measurement data, the method comprising the following steps performed by a computer hardware component: determining measurement data using a radar sensor; correcting the measurement data based on a misalignment parameter determined according to the computer-implemented method described herein to obtain corrected measurement data.
[0026] With the corrected measurement data, improved inspection results can be achieved.
[0027] Based on the misalignment parameter, control of the radar sensor (eg, the area in which the radar sensor is to be used to determine the measurements) may be adjusted (eg, to a desired area or a desired field of view).
[0028] In another aspect, the present disclosure relates to a computer system comprising a plurality of computer hardware components configured to perform several or all of the steps of the computer-implemented method described herein. The computer system may be part of a vehicle.
[0029] The computer system may include multiple computer hardware components (e.g., a processor (e.g., a processing unit or a processing network), at least one memory (e.g., a storage unit or a storage network), and at least one non-transitory data storage). It should be understood that other computer hardware components may be provided and used to perform the steps of the computer-implemented method in the computer system. The non-transitory data storage and / or storage unit may include a computer program for instructing a computer, for example, using a processing unit and at least one storage unit to perform several or all steps or aspects of the computer-implemented method described herein.
[0030] In another aspect, the present disclosure relates to a radar system comprising a radar sensor having a misalignment parameter determined according to the computer-implemented method described herein.
[0031] In another aspect, the present disclosure is directed to a vehicle comprising the radar system described herein.
[0032] In another aspect, the present disclosure relates to a non-transitory computer-readable medium containing instructions for performing several or all steps or aspects of the computer-implemented methods described herein. The computer-readable medium can be configured as: an optical medium such as a compact disk (CD) or a digital versatile disk (DVD); a magnetic medium such as a hard disk drive (HDD); a solid-state drive (SSD); a read-only memory (ROM) such as a flash memory; etc. In addition, the computer-readable medium can be configured as a data storage unit accessible via a data connection such as an Internet connection. The computer-readable medium can be, for example, an online data repository or cloud storage.
[0033] The present disclosure also relates to a computer program for instructing a computer to execute several or all steps or aspects of the computer-implemented method described herein. BRIEF DESCRIPTION OF THE DRAWINGS
[0034] Example implementations and functions of the present disclosure are described herein in conjunction with the following drawings which are schematically shown:
[0035] Figure 1A is an illustration of the radar field of view when misaligned in azimuth and elevation;
[0036] Figure 1B is an illustration of three-dimensional misalignment in azimuth, elevation, and roll;
[0037] Figure 2 is a flow chart illustrating an overview of a method of determining alignment parameters according to various embodiments;
[0038] Figure 3 is an illustration of the radar field of view after the introduction of roll angle misalignment;
[0039] Figure 4 is an illustration of a radar field of view and a compensation field of view according to various embodiments;
[0040] Figure 5 is a diagram illustrating azimuth alignment error distribution according to various embodiments;
[0041] Figure 6 is a diagram illustrating elevation alignment error distribution according to various embodiments;
[0042] Figure 7 is a diagram illustrating a roll angle alignment error distribution according to various embodiments (Instantaneous roll angle alignment error distribution (500000 samples));
[0043] Figure 8 is a flow chart illustrating a method of determining alignment parameters of a radar sensor according to various embodiments;
[0044] Fig. 9 is an alignment parameter estimation system according to various embodiments; and
[0045] Fig.10 is a computer system having a plurality of computer hardware components, the computer system being configured to perform the steps of a computer-implemented method of determining alignment parameters of a radar sensor according to various embodiments;
[0046] Fig.11 is an example of the vehicle coordinate system position;
[0047] Fig.12 is an illustration of a sensor coordinate system relative to a vehicle coordinate system (VCS); and
[0048] Fig.13 is an example of a detection definition in the sensor coordinate system.
[0049] Reference numerals list
[0050] 100 Example of radar field of view when misaligned in azimuth and elevation
[0051] 102 Horizontal Line
[0052] 104 vertical dividing line
[0053] 106 Ideal orientation angle
[0054] 108 True heading angle
[0055] 110 Radar Field of View
[0056] 112 Inaccuracy
[0057] 150 Illustration of 3D misalignment in azimuth, elevation and roll
[0058] 152 x-axis
[0059] 154 y-axis
[0060] 156 z axis
[0061] 158 Ground Plane
[0062] 160 Actual visual axis direction
[0063] 162 Expected Field of View
[0064] 164 Actual Field of View
[0065] 200 is a flowchart illustrating an overview of a method for determining alignment parameters according to various embodiments.
[0066] 202 Data Acquisition
[0067] 204 Return
[0068] 206 Filter
[0069] 300 Example of radar field of view after the introduction of roll angle misalignment
[0070] 302 Illustration from the ground perspective
[0071] 304 Example from the Sensor's Perspective
[0072] 400 Example of radar field of view and compensation field of view according to various embodiments
[0073] 402 Compensated Field of View
[0074] 500 illustrates a diagram of an azimuth alignment error distribution according to various embodiments.
[0075] 502 Horizontal axis
[0076] 504 Vertical axis
[0077] 506 is a graph showing the result of 2-axis alignment.
[0078] 508 is a graph illustrating the results of a 3-axis aligned NL-LS method according to various embodiments.
[0079] 510 is a graph illustrating the results of the 3-axis alignment NL-TLS method according to various embodiments.
[0080] 600 illustrates a diagram of elevation alignment error distribution according to various embodiments.
[0081] 602 Horizontal axis
[0082] 604 Vertical Axis
[0083] 606 is a graph showing the result of 2-axis alignment.
[0084] 608 illustrates a graph of the results of a 3-axis aligned NL-LS method according to various embodiments.
[0085] 610 is a graph illustrating the results of the 3-axis alignment NL-TLS method according to various embodiments.
[0086] 700 illustrates a graph of roll angle alignment error distribution according to various embodiments.
[0087] 702 Horizontal axis
[0088] 704 Vertical Axis
[0089] 706 illustrates a graph of the results of a 3-axis aligned NL-LS method according to various embodiments.
[0090] 708 illustrates a graph of the results of a 3-axis alignment NL-TLS method according to various embodiments.
[0091] 800 is a flowchart illustrating a method for determining an alignment parameter of a radar sensor according to various embodiments.
[0092] 802 Steps for determining measurement data using a radar sensor
[0093] 804 Steps to determine the speed of a radar sensor
[0094] 806 Steps for determining misalignment parameters based on measurement data and velocity
[0095] 900 Alignment parameter estimation system according to various embodiments
[0096] 902 Measurement data determination circuit
[0097] 904 Speed determination circuit
[0098] 906 Misalignment Parameter Determination Circuit
[0099] 908 connections
[0100] 1000 Computer system according to various embodiments
[0101] 1002 Processor
[0102] 1004 Memory
[0103] 1006 Non-transitory data storage unit
[0104] 1008 Radar Sensor
[0105] 1010 Speed Sensor
[0106] 1012 connections
[0107] 1100 Example of vehicle coordinate system position
[0108] 1102 Host Vehicle
[0109] 1104 Ground Plane
[0110] 1106 Rear suspension
[0111] 1108 front bumper
[0112] 1110 Sprung mass centre line
[0113] 1200 Example of sensor coordinate system relative to vehicle coordinate system (VCS)
[0114] 1202 Translation
[0115] 1204 Rotation
[0116] 1300 Example of detection definition in sensor coordinate system
[0117] 1302 Sensor origin
[0118] 1304 X-axis
[0119] 1306 Y-axis
[0120] 1308 Z axis
[0121] 1310 Detection azimuth
[0122] 1312 Detecting elevation angle
[0123] 1314 Detection distance
[0124] 1316 Detection Target
[0125] 1318 Projection of the line connecting the origin and the detection target on the (X; Y) plane DETAILED DESCRIPTION
[0126] Radar alignment is the process of determining the angular misalignment of the radar's heading direction. Correction of these installation errors can be critical to the proper functioning of radar-based tracking methods and most of the features that operate on the detection or tracking of objects provided by the radar sensor. The two main categories of methods / approaches designed to address misalignment are:
[0127] 1) Static alignment methods, which can also be called factory alignment because they require dedicated equipment to perform the calibration process; the most commonly used static methods use corner reflectors, Doppler generators or steel plates;
[0128] 2) Dynamic alignment methods, which may also be referred to as automatic alignment since they may not require any external (additional) equipment (devices). The radar sensor corrects its heading angle based on its observation of the environment while the car is moving (ie driving).
[0129] Figure 1A and Figure 1B Illustrations 100 and 150 of misalignment and radar field of view relationships are shown.
[0130] Figure 1A An illustration 100 is shown when the radar field of view 110 is misaligned in azimuth (along horizontal line 102) and elevation (along vertical dividing line 104 at the front of the vehicle, which corresponds to 0 degrees). The ideal (or desired or assumed) heading angle 106 is different from the true (or actual) heading angle 108, so that the misalignment 112 exists.
[0131] Figure 1B An illustration 150 of three-dimensional misalignment in azimuth (α), elevation (β), and roll (γ) is shown. The coordinate axes, namely X VCS 152. Y VCS 154 and Z VCS 156. A ground plane 158 is illustrated. The radar sensor can be mounted above the ground plane 158 at the origin of the vehicle coordinate system. The desired boresight direction for perfect alignment would be the same as the X VCS However, the actual visual axis direction 160 is different from the X axis direction 152 due to misalignment. VCS 152 is different, and the actual visual axis direction 160 is different from X VCSThe difference between 152 can be described by a three-dimensional misalignment in azimuth (α), elevation (β) and roll (γ). Due to the misalignment, the desired FOV 162 (field of view) may be different from the actual FOV 164. However, the desired FOV 162 can be obtained by correction based on the misalignment parameters based on the actual FOV 164.
[0132] To calculate the misalignment in the case of misalignment in azimuth and elevation (not taking into account possible misalignment in roll angle), the range rate equation for stationary detection can be used. It can be derived from the projection of the negative vehicle velocity vector onto the range vector, forming the following equation:
[0133]
[0134] in:
[0135] · is the rate of change of range (Doppler velocity for stationary detection);
[0136] ·V X is the longitudinal velocity of the vehicle (which may be a vector relative to the vehicle's forward direction, parallel to the ground; V X can be the length of the vector);
[0137] ·V Y is the lateral velocity of the vehicle (which may be a vector towards the left side of the vehicle, parallel to the ground; V Y can be the length of the vector);
[0138] ·V Z is the vertical velocity of the vehicle (which can be a vector pointing toward the ground, perpendicular to the ground; V Z can be the length of the vector);
[0139] α VCS is an azimuth measurement of the detection aligned with the vehicle coordinate system (where an angle of 0 [degrees] may be defined as facing in the same direction as the front of the vehicle on which the radar sensor is mounted); and
[0140] β VCS is the elevation measurement of the detection aligned with the vehicle coordinate system.
[0141] The measurement angle of VCS alignment can be decomposed into the sensor installation angle, sensor detection measurement angle and sensor installation misalignment angle:
[0142] α VCS =α SCS +α B +α M (2)
[0143] β VCS =βSCS +β B +β M (3) Among which:
[0144] α SCS is the azimuth measurement detected in the sensor coordinate system (SCS);
[0145] α B is the desired azimuth installation angle;
[0146] α M is the expected azimuth misalignment;
[0147] β SCS is the elevation angle measurement detected in the sensor coordinate system;
[0148] β B is the desired elevation mounting angle; and
[0149] β M The elevation angle is inaccurate.
[0150] Using the speed correction factor V c (Also referred to as the (vehicle) speed adjustment error or speed compensation factor) Further expanding the equation yields:
[0151]
[0152]
[0153]
[0154] in:
[0155] ·V Xm is the measured longitudinal velocity of the vehicle;
[0156] ·V Ym is the measured lateral velocity of the vehicle;
[0157] ·V Zm is the measured vertical velocity of the vehicle; and
[0158] ·V c is the speed adjustment error.
[0159] You can create an equation based on the measured values (X), known constants (C), and the equation results (Y) to find the unknown parameters (B), as shown below:
[0160]
[0161] Various methods, such as iterative nonlinear least squares regression or the method of errors in variables, can be used to determine the unknown parameters. The parameters found by these methods can then be further improved by time filtering methods such as the Kalman filter.
[0162] Figure 2 A flow chart 200 is shown illustrating an overview of a method of determining alignment parameters according to various embodiments (either without considering misalignment in roll angle, or considering misalignment in roll angle, as will be described in more detail below). At 202, data acquisition can be performed, which can include stationary detection, vehicle speed, and radar information. At 204, regression can be performed, for example including an estimate of azimuth misalignment, an estimate of elevation misalignment, an estimate of roll angle misalignment (described in more detail below), and an estimate of velocity adjustment error. At 206, filtering can be provided (e.g., using a Kalman filter).
[0163] A significant roll misalignment of the sensor (desired sensor roll angle is typically zero) can significantly affect azimuth and elevation measurements because it reduces the radar field of view, as shown below.
[0164] Roll misalignment may affect the azimuth-elevation misalignment calculation itself.
[0165] Figure 3 An example 300 of the radar field of view after the introduction of roll angle misalignment is shown, which is consistent with Figure 1A is similar to that shown in Figure 1A Roll angle misalignment is not illustrated in the example). Figure 3 The illustration 302 in the left part of FIG. 1 shows an illustration from the perspective of the ground, while Figure 3 Illustration 304 in the right portion of shows the illustration from the sensor's perspective.
[0166] As from Figure 3 As can be seen, if the roll misalignment is not considered, the azimuth misalignment can be considered to be 0 (zero) or close to 0, and the elevation misalignment can be estimated to be greater than the elevation misalignment correction value, since the ideal heading angle is just below the reported heading angle seen by the radar sensor. This may result in finding incorrect misalignment parameters, such as zero azimuth misalignment and too high elevation misalignment values.
[0167] The effect of roll misalignment is barely observable for small angles (e.g., less than 1 [degree]) and may grow exponentially with the roll angle, introducing the largest error at the field of view angle. Figure 3 and Figure 4As shown, at a radar roll angle of 30 [degrees] (i.e., misalignment of the radar sensor at a roll angle of 30 degrees), the error reduces the radar's field of view by half (the sensor may observe the ground near the front of the vehicle, and the sky above the opposite side of the FOV; however, from the perspective of active safety methods, neither the ground nor the sky area is relevant). At the same time, a significant portion of important information about objects located approximately 0.5m above the ground is missed. In addition, undetected roll misalignment may render the azimuth and elevation misalignment angles unreliable. For example, an object at 60 degrees in azimuth and 5 degrees in elevation under the influence of a 5-degree sensor roll misalignment will be reported as being at 60.20 degrees in azimuth and -0.2484 degrees in elevation.
[0168] Radar roll misalignment can be determined by extending the range rate equation, which makes it possible to estimate the roll angle. This equation can further improve the estimation of yaw and pitch misalignment. After estimating the 3 angular misalignments (yaw, pitch, roll), the measured detection results can be corrected to represent the detection results in the vehicle coordinate system, or in the case of extreme misalignment, the estimated value may trigger an alarm, which will stop the execution of other radar functions.
[0169] The range rate equation (1) can be used together with equations (2) and (3), which can be expanded to take advantage of the roll angle effect, which yields the following equation:
[0170] α VCS =α SCS ′+α B +α M (8)
[0171] β VCS =β SCS ′+β B +β M (9)
[0172] α′ SCS =α SCS cos(γ M +γ B )-β SCS sin(γ M +γ B ) (10)β′ SCS =α SCS sin(γ M +γ B )+β SCS cos(γ M +γ B ) (11) Among them:
[0173] α' SCSis the azimuth angle measurement corrected for roll misalignment in the sensor coordinate system;
[0174] β' SCS is the elevation angle measurement corrected for roll misalignment in the sensor coordinate system;
[0175] γ M is the roll angle misalignment; and
[0176] γ B is the ideal mounting position for the roll angle (it can usually be 0 degrees or 180 degrees).
[0177] Therefore, α' SCS and β' SCS can be visualized as detections at the same distance from the center of the radar field of view (0, 0) but rotated (aligned) relative to the horizontal line (assuming the desired roll angle installation angle is 0 [degrees]), resulting in Figure 4 The “compensated field of view” is shown.
[0178] Figure 4 An illustration 400 of the radar field of view 110 and the compensation field of view 402 is shown in accordance with various embodiments. Figure 4 The various parts and Figure 1A The parts are similar or identical, so the same figure marks may be used and repeated descriptions may be omitted.
[0179] In equations (8) and (9)(α' SCS and β' SCS ) into equations (2) and (3) for α SCS and β SCS ), we can get the following equations respectively:
[0180] α VCS =α SCS cos(γ M +γ B )-β SCS sin(γ M +γ B )+α B +α M (12)
[0181] β VCS =α SCS sin(γ M +γ B )+β SCS cos(γ M +γ B )+β B +β M (13)
[0182] This change modifies the general model in equation (7) to:
[0183]
[0184] The general form of the model can be derived from the distance rate equation (1):
[0185]
[0186] Assume that the vehicle is only moving in the forward direction (V Ym and V Zm = 0) and the desired installation angles of the elevation and roll angles are 0 [degrees], then equation (13) can be simplified to the following form:
[0187]
[0188] Equation (16) can be used to compose (create) an overdetermined system of equations that can be solved by various methods to find the parameters of interest, namely the misalignment angle and the velocity compensation factor. In one example, equation (14) can be used as a model for nonlinear least squares regression. For example, the model may be iterative, meaning that the method will iterate on the same data set containing data from one time instance to converge to a solution to the nonlinear equation in multiple steps of linearizing the model around its operating point:
[0189] (J T ·J)ΔB=J T ·ΔY (17)
[0190] The matrix equation (17) can be transformed into:
[0191] ΔB=(J T ·J) -1 ·J T ·ΔY (18)
[0192] in:
[0193] ΔB is the update of the parameter value in a single iteration (a single step in the solution space);
[0194] J is the partial derivatives of the model function with respect to the parameters (which may be called the Jacobian matrix); and
[0195] ΔY is the difference between the measured output of the system and the modeled output:
[0196]
[0197] The Jacobian matrix can be of the following form:
[0198]
[0199] in:
[0200]
[0201]
[0202]
[0203]
[0204] After assuming the initial values of the parameters, the nonlinear regression model described by equations (18) to (24) can be solved:
[0205]
[0206] Where m is the number of successful alignment cycles (radar cycles). In other words, a misalignment of 0 in azimuth, elevation and roll angles and velocity correction factors can be assumed as the starting point for the first iteration, while the results of the previous iteration can be used as the starting point for the next iteration.
[0207] Therefore, equation (26) can use the values of the i-th iteration to update the model parameter values:
[0208] B i =B i-1 +ΔB i (26)
[0209] B i After countless iterations (or when the number of iterations approaches infinity), it may converge to the actual speed adjustment error and misalignment angle:
[0210]
[0211] The method according to various embodiments also provides alignment parameters for the roll axis that accurately describe the physical model.
[0212] Figure 5 , Figure 6 and Figure 7 The execution of the method according to various embodiments is illustrated on simulated data for random (non-zero) misalignment for all mounting angles (i.e., azimuth, elevation, and roll). The following simulation parameters were used:
[0213] Number of test cases (samples): 500,000 (the input of each method is the same);
[0214] Misalignment (azimuth, elevation and roll): pseudo-random values drawn from a uniform distribution over the interval [-5:5] degrees;
[0215] Azimuth boresight pseudo-random values drawn from a uniform distribution over the interval [-180:180] degrees;
[0216] The vehicle moves along a straight line (X VCS direction) movement;
[0217] Minimum vehicle speed: 11m / s;
[0218] Number of tests: 64
[0219] Azimuth Gaussian zero-mean noise: standard deviation 0.3 degrees;
[0220] Gaussian zero-mean noise at elevation: standard deviation 1 degree;
[0221] Distance change rate Gaussian zero mean noise: standard deviation 0.03m / s;
[0222] Vehicle speed Gaussian zero mean noise on 3 axes: standard deviation 0.2 m / s.
[0223] Figure 5 , Figure 6 and Figure 7 The results of a single iteration (first iteration) of each compared method are shown respectively. NL-LS in the method name may stand for nonlinear least squares regression for finding the solution, while NL-TLS may stand for nonlinear total least squares regression. For comparison purposes, a 2-axis alignment without roll angle estimation is shown.
[0224] Figure 5 A graph 500 illustrating the distribution of azimuth alignment errors according to various embodiments is shown. The deviation angle is illustrated on the horizontal axis 502. The number of samples is illustrated on the vertical axis 504. Curve 506 illustrates the results of a 2-axis alignment. Curve 508 illustrates the results of a 3-axis alignment NL-LS method according to various embodiments. Curve 510 illustrates the results of a 3-axis alignment NL-TLS method according to various embodiments.
[0225] Figure 6 An illustration 600 illustrating an elevation alignment error distribution according to various embodiments is shown. The deviation angle is illustrated on the horizontal axis 602. The number of samples is illustrated on the vertical axis 604. Curve 606 illustrates the results of a 2-axis alignment. Curve 608 illustrates the results of a 3-axis alignment NL-LS method according to various embodiments. Curve 610 illustrates the results of a 3-axis alignment NL-TLS method according to various embodiments.
[0226] Figure 7A graph 700 illustrating a roll angle alignment error distribution according to various embodiments is shown. Deviation angle is illustrated on horizontal axis 702. Number of samples is illustrated on vertical axis 704. Curve 706 illustrates the results of a 3-axis alignment NL-LS method according to various embodiments. Curve 708 illustrates the results of a 3-axis alignment NL-TLS method according to various embodiments.
[0227] like Figure 5 and Figure 6 As shown, in the case of azimuth and elevation misalignment estimation for 3-axis alignment, even the (non-iterative) results based on one time instance are much better compared to 2-axis alignment.
[0228] Figure 7 The error distribution for roll angle alignment is shown, in this case 2-axis alignment is not available as it does not return a roll angle estimate.
[0229] Figure 8 A flow chart 800 illustrating a method of determining alignment parameters of a radar sensor according to various embodiments is shown. At 802, measurement data can be determined using the radar sensor, where the measurement data can include range rate measurements, azimuth measurements, and elevation measurements. At 804, a velocity of the radar sensor can be determined. At 806, a misalignment parameter can be determined based on the measurement data and the velocity, where the misalignment parameter can include azimuth misalignment, elevation misalignment, and roll misalignment.
[0230] According to various embodiments, the misalignment parameter may also include a speed adjustment error.
[0231] According to various embodiments, a system parameter may be determined, wherein the system parameter may include at least one of a desired azimuth mounting angle, a desired elevation mounting angle, and a desired roll mounting angle. A misalignment parameter may also be determined based on the system parameter.
[0232] According to various specific embodiments, the measurement data may relate to a plurality of objects external to the radar sensor.
[0233] According to various implementations, the plurality of objects may be stationary.
[0234] According to various embodiments, the misalignment parameter may be determined based on an iterative method.
[0235] According to various embodiments, the misalignment parameter may be determined based on an optimization method.
[0236] According to various embodiments, the misalignment parameter may be determined based on a non-linear least squares regression method.
[0237] According to various embodiments, the misalignment parameter may be determined based on a nonlinear total least squares regression method.
[0238] According to various embodiments, the misalignment parameter may be determined based on filtering.
[0239] According to various embodiments, the measurement data may be determined using a plurality of radar sensors.
[0240] According to various implementations, velocity and misalignment parameters of a radar sensor may be determined simultaneously by solving an optimization problem.
[0241] Each of the above steps 802, 804, 806 and other steps may be performed by computer hardware components.
[0242] Fig. 9 An alignment parameter estimation system 900 is shown according to various embodiments. The alignment parameter estimation system 900 may include a measurement data determination circuit 902, a velocity determination circuit 904, and a misalignment parameter determination circuit 906.
[0243] The measurement data determination circuit 902 may be configured to use a radar sensor to determine measurement data, the measurement data including a range rate measurement value, an azimuth measurement value, and an elevation measurement value. The radar sensor may be part of the measurement data determination circuit 902 or may be provided outside the measurement data determination circuit 902.
[0244] The speed determination circuit 904 may be configured to determine the speed of the radar sensor. The speed sensor may be part of the speed determination circuit 904 or may be provided external to the speed determination circuit 904.
[0245] The misalignment parameter determination circuit 906 may be configured to determine misalignment parameters of the radar sensor based on the measurement data and the speed, the misalignment parameters including azimuth misalignment, elevation misalignment, and roll misalignment.
[0246] The measurement data determination circuit 902 , the speed determination circuit 904 , and the misalignment parameter determination circuit 906 may be coupled to one another, for example, by exchanging electrical signals via an electrical connection 908 such as a cable or a computer bus, or via any other suitable electrical connection.
[0247] A "circuit" may be understood as any kind of logic-implemented entity, which may be a dedicated circuit or a processor executing a program stored in a memory, firmware or any combination thereof.
[0248] Fig.10A computer system 1000 having a plurality of computer hardware components is shown, the computer system being configured to perform the steps of a computer-implemented method for determining alignment parameters of a radar sensor according to various embodiments. The computer system 1000 may include a processor 1002, a memory 1004, and a non-transitory data storage unit 1006. A radar sensor 1008 and / or a speed sensor 1010 may be provided as part of the computer system 1000 (e.g., Fig.10 as shown), or may be disposed outside the computer system 1000.
[0249] Processor 1002 may execute instructions provided in memory 1004. Non-transitory data storage 1006 may store computer programs including instructions that may be transferred to memory 1004 and then executed by processor 1002. Radar sensor 1008 may be used to determine measurement data as described above. Speed sensor 1010 may be used to determine speed as described above.
[0250] The processor 1002, the memory 1004 and the non-transitory data storage 1006 can be connected to each other, for example, via an electrical connection 1012 such as a cable or a computer bus or via any other suitable electrical connection to exchange electrical signals. The radar sensor 1008 and / or the speed sensor 1010 can be connected to the computer system 1000, for example, via an external interface, or can be arranged as part of the computer system (in other words: for example, inside the computer system, for example, connected via the electrical connection 1012).
[0251] The terms “coupled” or “connected” are intended to include a direct “coupled” (eg, via a physical link) or direct “connected” as well as an indirect “coupled” or indirect “connected” (eg, via a logical link), respectively.
[0252] It should be understood that what is described with respect to one of the above methods may be similarly applicable to the alignment parameter estimation system 900 and / or the computer system 1000 .
[0253] In the following, other embodiments will be described. These embodiments may provide simultaneous estimation of host vehicle speed and radar misalignment based on multiple Doppler radar detections.
[0254] As used herein, various coordinate systems may be used.
[0255] The ISO-Earth-fixed coordinate system can be based on an axis system that remains fixed in an inertial reference frame. The origin of the coordinate system can be fixed to the ground plane. The position and orientation can be chosen in an arbitrary manner (e.g., based on the desired application). The system may be named in the ISO standard as (xE; yE; zE).
[0256] Another coordinate system may be the vehicle coordinate system.
[0257] Fig.11 An illustration 1100 of the position of a vehicle coordinate system is shown. The vehicle coordinate system may include three right-hand axes: (XV; YV; ZV). Values in the coordinate system may be represented as (xV; yV; zV). The vehicle coordinate system may be centered on the rear axle of the vehicle in a stable state. The coordinate system may remain fixed to the sprung mass of the vehicle, meaning that it may not be aligned with the rear axle during motion or when leaning. The position of the VCS on a stationary vehicle body relative to the dimensions of the vehicle may be as shown in FIG. Fig.11 As shown, in Fig.11 In the top side view of FIG. 1 , a stationary host vehicle 1102 is shown on a ground plane 1104 with a rear overhang 1106 and a distance 1108 between the rear axle and the front bumper. Fig.11 The top view shown at the bottom of the drawing shows the sprung mass centerline 1110. The axes (XV; YV; ZV) are also Fig.11 Shown in.
[0258] Another coordinate system may be the sensor coordinate system.
[0259] Sensors such as LiDAR and radar report detections in polar coordinates. Each detection can be described by three terms: range; azimuth; elevation.
[0260] Fig.12 An illustration 1200 of a sensor coordinate system with respect to a vehicle coordinate system (VCS) is shown. The origin of the sensor coordinate system may be defined relative to the VCS by a translation 1202 and a rotation 1204.
[0261] Fig.13 An illustration 1300 of a detection definition in a sensor coordinate system is shown. A sensor origin 1302, an X-axis 1304, a Y-axis 1306, and a Z-axis 1308 are shown. A detection azimuth 1310 may be defined as the angle between the X-axis 1304 and a projection 1318 of a line connecting the origin 1302 and a detection target 1316 onto the (X;Y) plane. A detection elevation 1312 may be defined as the angle between the (X;Z) plane and the line 1318 connecting the origin and the detection target 1316. A detection distance 1314 may be defined as the Cartesian distance between the sensor origin 1302 and the detection target 1316.
[0262] Each radar scan produces a set of (radar) detections, each of which may have the following properties:
[0263] Distance (which can be the distance from the sensor origin to the detection target, such as Fig.13 shown);
[0264] The rate of change of range (which can be the radial velocity of the detected target; in other words: the time derivative of the range);
[0265] Azimuth (e.g. Fig.13 shown); and
[0266] Elevation angle (e.g. Fig.13 shown).
[0267] The target plane motion can be described by:
[0268] V t =[ω t x t y t ] T
[0269] The equation for the rate of change of distance for a single raw detection relative to the target can be expressed as:
[0270]
[0271] in:
[0272] · - the i-th detection distance change rate,
[0273] ·θ i -The azimuth of the i-th detection.
[0274] In order to simplify the representation, the concept of compensation distance change rate can be introduced and defined as:
[0275]
[0276] Then, the distance rate equation can be simplified to:
[0277]
[0278] In vector form it can be described as follows:
[0279]
[0280] coefficient and It can be called velocity distribution.
[0281] If at least 2 detections can be made relative to a target object, the velocity distribution can be successfully estimated. This estimation can be done by applying the least squares method or the method proposed in US 2019 / 0369228. Both methods can provide an estimate of the velocity distribution and its covariance matrix:
[0282]
[0283] Gauss–Newton optimization can be used to minimize a function in quadratic form:
[0284] Q(μ)=f(μ) T f(μ)
[0285] The Gauss-Newton optimization method may include the following steps:
[0286] 1. You can first give the point μ 0 , function f(μ) and its Jacobian matrix Set i=0.
[0287] 2. Set up γ i =1.
[0288] 3. Calculation:
[0289] μ i+1 =μ i +γ i (J(μ i ) T J(μ i )) -1 J(μ i ) T f(μ i )
[0290] 4. If Q(μ i+1 )>Q(μ i ), then set And repeat from 3.
[0291] 5. If the stopping criteria are met, terminate.
[0292] 6. Otherwise set i=i+1 and repeat from 2.
[0293] According to various embodiments, an apparatus and method for simultaneously estimating vehicle speed and radar misalignment angle may be provided. For a single radar, such an approach may not be possible, because only with multiple radars can the measurements from one sensor be corrected based on the measurements of other sensors.
[0294] Hereinafter, a method for a 2D (two-dimensional) case will be described.
[0295] In the 2D case, the distance rate equation can have the following form:
[0296]
[0297] in:
[0298] · It can be the rate of change of the detected distance,
[0299] ·θ can be the azimuth angle of detection,
[0300] · Can be the sensor velocity reported in the sensor coordinate system (SCS).
[0301] If α is the sensor yaw angle, then the rate of change of the detected range can be a function of the sensor velocity in the vehicle coordinate system (VCS) and can have the following form:
[0302]
[0303] in
[0304] · can be the sensor velocity reported in the vehicle coordinate system (VCS), and
[0305] The sensor yaw angle α can be the calibration angle α 0 and misalignment angle α m The sum of:
[0306] α=α 0 +α m
[0307] Sensor speed can also be expressed as a function of host speed in VCS:
[0308]
[0309] in
[0310] · Can be the host velocity reported in the vehicle coordinate system (VCS).
[0311] ω can be the host yaw rate,
[0312] Δx, Δy can be the position of the radar in the VCS.
[0313] Based on these equations, the distance rate equation can be written as:
[0314]
[0315] Taking into account all static detections from all radars, the following system of equations can be provided:
[0316]
[0317] In short, this can be written as:
[0318]
[0319] in
[0320]
[0321] Based on this, the host velocity vector can be found by optimizing the following function Align with the radar:
[0322]
[0323] This optimization can be done by the Gauss-Newton method as described above, where the Jacobian matrix has the following form:
[0324]
[0325] in:
[0326] ·
[0327] ·
[0328] ·
[0329] ·
[0330] Based on this Jacobian matrix, the Hessian matrix can be approximated in the following form:
[0331]
[0332] in:
[0333]
[0334] In addition, for optimization methods, the matrix
[0335]
[0336] and it can be calculated as:
[0337]
[0338] Hereinafter, the 3D case will be described.
[0339] In the 3D case, the distance rate equation can have the following form:
[0340]
[0341] in:
[0342] · It can be the sensor velocity in the sensor coordinate system (SCS).
[0343] ·θ can be the azimuth angle of detection,
[0344] · It can be the elevation angle of the detection.
[0345] In this case, the radar orientation can be assumed to be 3D, so it can have yaw, pitch, and roll angles:
[0346]
[0347] Again, in this case, the actual orientation can be the sum of the calibrated orientation and the misalignment:
[0348]
[0349] Taking this orientation into account, the rate of change of distance can be expressed by the sensor velocity in the VCS by the following equation:
[0350]
[0351] in
[0352] · It can be the sensor velocity in the vehicle coordinate system (VCS).
[0353] · It can be a rotation 3D matrix obtained by rotating around the axes in the following order: OZ, OY, OX. To simplify the calculation, the following assumptions may be reasonable:
[0354]
[0355] α 0,r =0,
[0356]
[0357] It should be understood that the points in the first hypothesis may represent r and / or p and / or y.
[0358] Then the rotation matrix can have the following form:
[0359]
[0360] in:
[0361] ·r 11 =cosα 0,y -α m,y sinα 0,y ,
[0362] ·r 21 =-(sinα 0,y +α m,y cosα 0,y )(cosα p,0-a m,p Sinai 0,p ),
[0363] ·r 31 =sina 0,p +a m,p thing 0,p ,
[0364] ·r 12 =(sina) 0,y +a m,y thing 0,y )+a m,r (things) 0,y -a m,y Sinai 0,y (sina) 0,p +a m,p thing 0,p ),
[0365] ·r 22 =(thing) 0,y -a m,y Sinai 0,y )-a m,r (sina) 0,y -a m,y thing 0,y (sina) 0,p +a m,p thing 0,p ),
[0366] ·r 32 =-a m,r (things) 0,p -a m,p Sinai 0,p ),
[0367] ·r 13 =(sina) 0,y +a m,y thing 0,y )-a m,r (things) 0,y -a m,y Sinai 0,y (sina) 0,p +a m,p thing 0,p ),
[0368] ·r 23 =-a m,r (things) 0,y -a m,y Sinai 0,y )+(sina0,y +α m,y cosα 0,y )(sinα 0,p +α 0,p cosα 0,p ),
[0369] ·r 33 =cosα 0,p -α m,p sinα 0,p .
[0370] Summing all these equations, the distance rate equation can be written as:
[0371]
[0372] Given all detections from all radars, one can pose an optimization problem to find the host velocity and radar misalignment angle (this may be similar to the 2D case above):
[0373]
[0374] in:
[0375]
[0376] As mentioned above, this can be solved by nonlinear least squares using the Gauss-Newton method, for which the Jacobian matrix can be provided in the following form:
[0377]
[0378] in:
[0379] ·
[0380] ·
[0381] ·
[0382] ·
[0383] ·
[0384] · .
Claims
1. A computer-implemented method for determining alignment parameters of a radar sensor, the computer-implemented method comprising the following steps performed by computer hardware components: determining measurement data using the radar sensor, the measurement data comprising range rate measurements, azimuth measurements, and elevation measurements; determining a speed of the radar sensor using a speed sensor; as well as determining misalignment parameters based on the measurement data and the velocity, the misalignment parameters comprising azimuth misalignment, elevation misalignment and roll misalignment, wherein the azimuth misalignment and the roll misalignment are used to correct the azimuth measurement value, and The elevation misalignment and the roll misalignment are used to correct the elevation angle measurement.
2. The computer-implemented method according to claim 1, in, The misalignment parameters also include actual speed adjustment errors.
3. The computer-implemented method of claim 1 or 2, further comprising the following steps performed by the computer hardware components: determining system parameters, the system parameters comprising at least one of a desired azimuth mounting angle, a desired elevation mounting angle, and a desired roll mounting angle of the radar sensor on the vehicle; in, The misalignment parameter is also determined based on the system parameter.
4. The computer-implemented method of claim 1 , wherein: The measurement data relate to a plurality of objects external to the radar sensor.
5. The computer-implemented method of claim 4, in, The plurality of objects are stationary.
6. The computer-implemented method of claim 1, in, The misalignment parameter is determined based on an iterative method.
7. The computer-implemented method of claim 1, in, The misalignment parameters are determined based on an optimization method.
8. The computer-implemented method of claim 1, in, The misalignment parameter is determined based on a nonlinear least squares regression method; and / or Wherein, the misalignment parameter is determined based on filtering.
9. The computer-implemented method of claim 1, in, The misalignment parameter is determined based on a nonlinear total least squares regression method.
10. The computer-implemented method of claim 1, in, The measurement data are determined using a plurality of radar sensors.
11. A computer-implemented method for determining corrected measurement data, The computer-implemented method comprises the following steps performed by computer hardware components: Determine measurement data using radar sensors; Based on the misalignment parameter determined by the computer-implemented method according to any one of claims 1 to 10, the measurement data is corrected to obtain the corrected measurement data, wherein: The azimuth angle measurement is corrected based on the azimuth angle misalignment and the roll angle misalignment, and the elevation angle measurement is corrected based on the elevation angle misalignment and the roll angle misalignment.
12. A computer system comprising a plurality of computer hardware components configured to perform the steps of the computer-implemented method according to any one of claims 1 to 11.
13. A radar system comprising a radar sensor, a misalignment parameter of the radar sensor being determined by the computer-implemented method according to any one of claims 1 to 11.
14. A vehicle comprising a radar system according to claim 13.
15. A non-transitory computer-readable medium comprising instructions for executing the computer-implemented method according to any one of claims 1 to 11.
