Vehicle-mounted sensor installation error calibration method, device, equipment and medium

Abstract

The invention provides a method, a device, equipment and a medium for calibrating the installation error of a vehicle-mounted sensor, and the method comprises the steps: measuring a plurality of targets with known accurate geographic coordinates through the vehicle-mounted sensor at different positions and postures, and obtaining a plurality of groups of actual measurement data; an optimization model with three installation angle error values of the vehicle-mounted sensor as optimization variables is constructed, and a fitness function of the optimization model is the sum of horizontal spherical distance errors between target geographic coordinates and real target geographic coordinates obtained according to current installation errors based on all multiple groups of actually measured data; improving the Harris eagle optimization algorithm, optimizing the optimization model based on the improved Harris eagle optimization algorithm, and searching to obtain a fitness function as a minimum installation error; and the vehicle-mounted sensor performs on-line compensation and correction on the original measurement value of the sensor based on the fitness function of the minimum installation error, and outputs an optimized data result.

Description

Technical Field

[0001] This invention belongs to the field of trajectory prediction technology for fixed-wing unmanned aerial vehicles (UAVs), specifically relating to a method, device, equipment, and medium for calibrating the installation error of vehicle-mounted sensors. Background Technology

[0002] Vehicle-mounted sensing sensors (such as lidar, millimeter-wave radar, and photoelectric detection equipment) are the core of modern intelligent vehicle environmental perception systems. These sensors are typically rigidly mounted on the vehicle body or gimbal platform. The raw data they measure (such as azimuth, pitch, and distance) needs to be transformed into the vehicle coordinate system based on their mounting matrix, and then combined with GNSS / IMU data to transform into the global geodetic coordinate system, ultimately obtaining the target's geographical location information. The sensor mounting matrix is ​​theoretically determined by three mounting angles (azimuth, pitch, and roll). However, in practical applications, due to machining tolerances, installation process limitations, and vibrations and shocks during vehicle operation, there is an unavoidable deviation between the actual mounting angle of the sensor and the theoretical design value, i.e., installation error. This error is amplified during coordinate transformation, leading to a significant decrease in the final target geolocation accuracy, directly affecting the reliability of target tracking and map construction.

[0003] Existing calibration methods rely on specialized equipment such as high-precision theodolites and optical calibration fields, and are performed manually by professionals. This method is cumbersome, time-consuming, labor-intensive, and costly, and cannot be easily recalibrated on-site after vehicle deployment. Other methods are based on optimization algorithms. These methods collect observation data from multiple sensors on a target with known geographic coordinates, as well as the vehicle's own pose data. They construct the installation error calibration problem as a nonlinear optimization problem, seeking an optimal set of installation error values ​​that minimizes the overall error between the calculated target position and its true position. Subsequently, optimization algorithms (such as particle swarm optimization and genetic algorithms) are used to solve this problem. However, standard swarm intelligent optimization algorithms generally suffer from problems such as getting trapped in local optima, limited convergence accuracy, slow convergence speed in later stages, and poor stability when solving such complex optimization problems involving high precision, nonlinearity, and multiple peaks.

[0004] The Harris Eagle Optimization Algorithm (Harris Eagle) is a novel metaheuristic algorithm that simulates the cooperative hunting behavior of Harris Eagles in nature. While it boasts advantages such as simple structure and few parameters, its standard version still suffers from shortcomings in handling such precise calibration problems, including insufficient population diversity and an imbalance between global exploration and local exploitation. This results in calibration accuracy and reliability that fail to meet the demands of high-precision applications. Therefore, this application aims to develop an automated calibration method that overcomes the shortcomings of existing algorithms and achieves high precision, robustness, and efficiency. Summary of the Invention

[0005] In order to overcome the technical problem that the calibration accuracy and reliability of existing technologies are difficult to meet the requirements of high-precision application scenarios, the present invention provides a method, device, equipment and medium for calibrating the installation error of vehicle-mounted sensors.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] In a first aspect, embodiments of this disclosure provide a method for calibrating the installation error of an on-board sensor, comprising the following steps:

[0008] Step S1: Measure multiple targets with known precise geographic coordinates at different positions and attitudes using vehicle-mounted sensors to obtain multiple sets of measured data;

[0009] Step S2: Construct an error optimization model with the error values ​​of the three installation angles of the vehicle sensor as optimization variables, and use the sum of the horizontal spherical distance errors between the target geographic coordinates and the real target geographic coordinates in all measured data as the fitness function of the error optimization model;

[0010] Step S3: Improve the Harris Eagle optimization algorithm, and optimize the error optimization model based on the improved Harris Eagle optimization algorithm. Based on the optimized error optimization model, use a local search algorithm to obtain the fitness function with minimum installation error.

[0011] Step S4: The vehicle-mounted sensor performs online compensation and correction on the sensor's measured data based on the fitness function with minimum installation error, and outputs optimized data results.

[0012] Furthermore, each set of data in step S1 includes:

[0013] The geographic coordinates of the vehicle-mounted sensors, including longitude, latitude, and altitude;

[0014] Attitude angles, including true north azimuth, pitch angle, and roll angle;

[0015] Sensor measurements, including relative azimuth, pitch, and distance;

[0016] In addition, the target's actual geographic coordinates, including longitude, latitude, and altitude.

[0017] Furthermore, in step S3, obtaining the fitness function with the minimum installation error includes:

[0018] For a specific set of measured data, optimize the variables based on the installation error assumption of the current iteration. Construct the installation error rotation matrix Among them, optimization variables include:

[0019] ;

[0020] In the formula, This refers to the installation error in orientation; For pitch installation error; For roll installation error; the , ,and All are errors at three installation angles of the vehicle-mounted sensor;

[0021] The installation error rotation matrix include:

[0022] ;

[0023] in,

[0024] ;

[0025] In the formula, Install an error rotation matrix for the azimuth angle; Install an error rotation matrix for the pitch angle; Install the roll angle error rotation matrix;

[0026] Based on the error rotation matrix The unit vectors measured by the vehicle-mounted sensors are corrected to obtain the unit vectors in the vehicle-mounted sensor coordinate system under ideal error-free installation conditions, and this is combined with the pose data of the vehicle-mounted sensors at that moment. and measuring distance Obtain the geographic coordinates of the target through coordinate transformation. ;

[0027] The geographic coordinates of the target are obtained using the Haversine horizontal spherical distance formula. relative to the target's true coordinates Horizontal spherical distance error between ;

[0028] The formula for the horizontal spherical distance of Haversvine is as follows:

[0029] ;

[0030] ;

[0031] ;

[0032] ;

[0033] ;

[0034] In the formula, rice; The difference in latitude between two points; Let be the latitude of the calculated or target location of the i-th point; Let i be the latitude of the i-th point; The difference in longitude between two points; Let be the longitude of the calculated or target location for the i-th point; Let be the longitude of the i-th point; Intermediate variables calculated based on the difference between latitude and longitude: The spherical angular distance between the two points;

[0035] Accumulated horizontal spherical distance error of all measured data The current installation error assumption is obtained. The sum of the horizontal spherical distance errors of all targets is used as the fitness function, expressed as:

[0036] ;

[0037] The goal is to minimize In the formula, M is the number of measured data sets. For the first Group of measured data.

[0038] Furthermore, the geographic coordinates of the target are obtained. The specific steps include:

[0039] Step A1: Transform the sensor measurements to the vehicle coordinate system and calculate the unit vector in the sensor coordinate system. :

[0040] ;

[0041] In the formula, The azimuth angle in the sensor coordinate system; The pitch angle in the sensor coordinate system;

[0042] Step A2: Combine installation error rotation matrix and unit vector in the sensor coordinate system The unit vector in the vehicle coordinate system is calculated. :

[0043] ;

[0044] In the formula, Let be the longitudinal distance of the i-th target in the vehicle coordinate system; Let be the lateral distance of the i-th target in the vehicle coordinate system; Let be the vertical height of the i-th target in the vehicle coordinate system; then , The azimuth angle in the sensor coordinate system;

[0045] Step A3: Combine the unit vector obtained in step A2 Based on counterclockwise angle atan2 The counterclockwise angle of the target on the horizontal plane of the vehicle coordinate system is obtained by formula, where, Range of values arrive ;

[0046] Define the azimuth angle in the vehicle coordinate system as clockwise from the X-axis, and convert counterclockwise angles to clockwise angles: Convert to positive values ​​to obtain the final vehicle coordinate system azimuth angle. ,like ,but ,in The value range is 0 to ; Obtain the pitch angle of the vehicle coordinate system , The range of values arrive ;

[0047] Step A4: and Substituting the values ​​into the solution function f, we obtain the calculated target coordinates. :

[0048] ;

[0049] In the formula, It is a known solution function that converts measurements in the vehicle coordinate system into global coordinates; y is the roll angle in the sensor coordinate system.

[0050] Furthermore, the process of improving the Harris Eagle optimization algorithm in step S3 includes:

[0051] An improved Harris Eagle optimization algorithm based on opposition learning, population initialization, Lévy flight exploration, adaptive boundary adjustment, and periodic local search;

[0052] The process of improving the Harris Eagle optimization algorithm for population initialization based on opposition learning includes:

[0053] In a given 3-dimensional search space, the range of each dimension is: ), N solutions are randomly generated to form the initial population P. Each solution is a 3D vector. ;

[0054] For each solution in the initial population P Calculate the alternative solution of the initial population P by dimension. For the first dimension, ,in This yields an opposing population P' containing N opposing solutions;

[0055] The initial population P is merged with the opposing population P' to form a merged population C containing 2N solutions;

[0056] Calculate the fitness value f(X) of all 2N solutions in the merged population C to measure the quality of each solution;

[0057] The N solutions with the smallest fitness values ​​are selected from the merged population C to form the initial population of the new generation, thus completing the population initialization process based on opposition learning.

[0058] The process of the improved Harris Eagle optimization algorithm based on Lévy flight exploration includes:

[0059] The uniform random step size in the standard exploration strategy of the Harris Eagle optimization algorithm is replaced with the Lévy flight step size, including:

[0060] ;

[0061] In the formula, This indicates element-wise multiplication; This indicates that during iteration, an individual is randomly selected from the population; Indicates the current iteration time; Step size generation for Lévy flight exploration includes:

[0062] Lévy ) u / | v (1 / );

[0063] in:

[0064] ;

[0065] ;

[0066] In the formula, It is the gamma function. It is an index;

[0067] The process of improving the Harris Eagle optimization algorithm based on the aforementioned adaptive boundary adjustment includes:

[0068] During the iteration process, the current global optimal solution is periodically used. Centered on the search space, the upper and lower bounds are dynamically shrunk. The search boundary gradually narrows as the number of iterations t increases, as shown below:

[0069] ;

[0070] ;

[0071] Where: δ is the contraction coefficient; (1-t / T) is a factor that linearly decays from 1 to 0;

[0072] The process of improving the Harris Eagle optimization algorithm based on periodic local search includes:

[0073] During the iteration process, every K generations, the global optimal solution at the current moment is used. Starting from a point, a local search algorithm is used to solve the problem. The result of the local search is compared with the fitness value of the current best solution. The solution with the smaller fitness value is selected as the new best solution, and the population is updated.

[0074] The updated population continues to execute the main loop steps of the Harris Eagle Optimization Algorithm until the maximum number of iterations is reached.

[0075] Furthermore, the main loop steps of the Harris Eagle optimization algorithm include:

[0076] Step B1: Calculate the fitness value for each individual based on the fitness function. Find the current fitness value. Minimum optimal solution , This represents the best individual found so far, and is the globally optimal solution updated after each iteration;

[0077] Step B2: For each individual, calculate the escape energy. The escape energy E is used to control the phase transition:

[0078] ;

[0079] In the formula, This represents the current iteration number; The total number of iterations is the preset maximum number of iterations; E is the escape energy. As initial energy, = rand(-1, 1), which is a random number uniformly distributed in the interval [-1, 1]; E is the updated energy, E = 2. (1 - t / T), which decreases linearly as iteration t increases;

[0080] Step B3: For each individual The location has been updated:

[0081] During the exploration phase, if ,and ,but:

[0082] ;

[0083] otherwise:

[0084] ;

[0085] During the development phase, if :

[0086] Then let

[0087] if and :

[0088] ;

[0089] if and :

[0090] ;

[0091] if and :

[0092] calculate ;

[0093] if , but Otherwise calculate In the formula, If it is a random vector, , but Otherwise ;

[0094] In the formula, For individuals The new position is calculated using the position update formula; rand() and rand(·) are random numbers uniformly distributed in the interval [0, 1]. This is the absolute value of energy E, used to determine the current stage; Let X be an individual randomly selected from the population; Lévy(β) is the Lévy flight step size, where β is a parameter of the Lévy distribution; mean(X) is the average position vector of population X; r is a random number, r = rand(), used to select the encirclement strategy; Y is a temporary position, Y = - E · | - |;Z is another temporary position, Z = Y + S · Lévy(β), where S is a random vector;

[0095] Step B4: Dynamically narrow the search range around the current optimal solution and construct an adaptive boundary:

[0096] Every The next iteration narrows the search scope around

[0097] ;

[0098] ;

[0099] In the formula, The interval for adaptive boundary updates; and The lower and upper bounds of the initial search space;

[0100] Step B5: Every K = 10K = 10 iterations, for Perform a Nelder-Mead simplex local search and obtain updates. ;

[0101] Step B6: After the maximum number of iterations is completed, output the optimal solution. .

[0102] In a second aspect, embodiments of this disclosure provide an apparatus for calibrating the installation error of an on-board sensor, comprising:

[0103] The data acquisition unit is configured to measure multiple targets with known precise geographic coordinates at different positions and attitudes using vehicle-mounted sensors, and obtain multiple sets of measured data.

[0104] The model building unit is configured to build an error optimization model with the error values ​​of the three installation angles of the vehicle sensor as optimization variables, and use the sum of the horizontal spherical distance errors between the target geographic coordinates and the real target geographic coordinates in all measured data as the fitness function of the error optimization model.

[0105] The model optimization unit is configured to improve the Harris Eagle optimization algorithm and optimize the error optimization model based on the improved Harris Eagle optimization algorithm. Based on the optimized error optimization model, a fitness function with minimum installation error is obtained by using a local search algorithm.

[0106] The output unit is configured to perform online compensation and correction on the measured data of the vehicle-mounted sensor based on a fitness function with minimum installation error, and output optimized data results.

[0107] In a third aspect, embodiments of this disclosure provide an electronic device, characterized in that the electronic device comprises:

[0108] At least one processor; and,

[0109] The memory is communicatively connected to the at least one processor; wherein,

[0110] The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method for calibrating the installation error of the vehicle-mounted sensor.

[0111] In a fourth aspect, embodiments of this disclosure provide a non-transitory computer-readable storage medium, characterized in that the non-transitory computer-readable storage medium stores computer instructions for causing the computer to perform the method for calibrating the installation error of the vehicle-mounted sensor.

[0112] The beneficial effects of this invention are:

[0113] This invention provides a method, apparatus, device, and medium for calibrating the installation error of an on-board sensor. The method includes: measuring multiple targets with known precise geographic coordinates using an on-board sensor at different positions and orientations, obtaining multiple sets of measured data; constructing an error optimization model with the installation angle error values ​​of the on-board sensor as optimization variables; using the sum of the horizontal spherical distance errors between the target geographic coordinates and the actual target geographic coordinates in all measured data as the fitness function of the error optimization model; improving the Harris Eagle optimization algorithm and optimizing the error optimization model based on the improved Harris Eagle algorithm; obtaining the fitness function with the minimum installation error based on the optimized error optimization model using a local search algorithm; and performing online compensation and correction of the sensor's measured data based on the fitness function with the minimum installation error, outputting the optimized data results. This application requires no additional dedicated equipment, utilizing only the on-board platform and the sensor's own measurement data, enabling rapid, automatic, and high-precision calibration of the sensor's installation angle error. Attached Figure Description

[0114] Figure 1A flowchart illustrating a method for calibrating the installation error of vehicle-mounted sensors according to an embodiment of this disclosure is shown;

[0115] Figure 2 The optimized results of the vehicle-mounted sensor installation error calibration method according to the embodiments of this disclosure are shown;

[0116] Figure 3 The comparison results between the estimated error and the actual error of the embodiments of this disclosure are shown;

[0117] Figure 4 A convergence curve diagram of an embodiment of this disclosure is shown;

[0118] Figure 5 A comparison diagram of the convergence curves of the vehicle-mounted sensor installation error calibration method according to an embodiment of the present disclosure with those of particle swarm optimization and genetic algorithms is shown.

[0119] Figure 6 A diagram of an apparatus for calibrating the installation error of an on-board sensor according to an embodiment of the present disclosure is shown. Detailed Implementation

[0120] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0121] To address the installation errors of vehicle-mounted sensing sensors, the first embodiment of this invention provides a method for calibrating the installation errors of vehicle-mounted sensors. Figure 1 A flowchart 100 illustrating a method for calibrating the installation error of an on-board sensor according to an embodiment of the present disclosure is shown, including the following steps:

[0122] In step S101, the vehicle-mounted sensors measure multiple targets with known precise geographic coordinates at different positions and attitudes, obtaining multiple sets of measured data.

[0123] In this embodiment of the disclosure, each set of data in step S1 includes:

[0124] The geographic coordinates of the vehicle-mounted sensors, including longitude, latitude, and altitude;

[0125] Attitude angles, including true north azimuth, pitch angle, and roll angle;

[0126] Sensor measurements, including relative azimuth, pitch, and distance;

[0127] In addition, the target's actual geographic coordinates, including longitude, latitude, and altitude.

[0128] Specifically, in this embodiment of the disclosure, in a calibration field containing multiple static target points with known precise geodetic coordinates, a vehicle-mounted platform equipped with sensors to be calibrated is manipulated to perform movements in various postures, including different orientations, pitch angles, and roll angles. This ensures that the sensors can observe multiple target points from multiple angles and distances, and synchronously record at least M sets of measured data in real time.

[0129] In this embodiment of the disclosure, the coordinate system is first defined, including:

[0130] Vehicle platform coordinate system (vehicle coordinate system): X-axis points towards the front of the vehicle, Y-axis points to the right, and Z-axis points downward.

[0131] Sensor coordinate system: Ideally aligned with the vehicle coordinate system, but installation errors exist.

[0132] Global coordinate system: WGS84 geographic coordinate system (longitude, latitude, altitude).

[0133] For each set of measurement data (common Group), known:

[0134] Sensor measurement: distance Azimuth angle in the sensor coordinate system (Clockwise 0° to 180°, counterclockwise) (Up to -180°), pitch angle (Upward is positive, downward is negative).

[0135] Vehicle platform status: geographic coordinates azimuth (True North 0° clockwise), pitch angle Roll angle ;

[0136] And, the target's actual geographic coordinates: .

[0137] Next, proceed to step S102.

[0138] In step S102, an error optimization model is constructed with the error values ​​of the three installation angles of the vehicle-mounted sensor as optimization variables. The sum of the horizontal spherical distance errors between the target geographic coordinates and the real target geographic coordinates in all measured data is used as the fitness function of the error optimization model.

[0139] In this embodiment of the disclosure, the sensor has three installation angle error values, including: azimuth installation error. Pitch installation error and roll installation error Defined as the decision variables to be optimized, and combined into optimization variables. The fitness function f(X) of the optimization model is constructed, which quantitatively describes the optimization variable assuming installation error is the optimization variable. Under the premise of [condition], the overall deviation between the positioning result obtained after target calculation of all M sets of measured data and the actual position.

[0140] Next, proceed to step S103.

[0141] In step S103, the improved Harris Eagle optimization algorithm is used to optimize the optimization model, and the fitness function is searched and obtained as the minimum installation error.

[0142] Specifically, the process of improving the Harris Eagle optimization algorithm in step S103 includes:

[0143] An improved Harris Eagle optimization algorithm based on opposition learning, population initialization, Lévy flight exploration, adaptive boundary adjustment, and periodic local search;

[0144] Specifically, the process of improving the Harris Eagle optimization algorithm based on opposition learning for population initialization includes:

[0145] In a given 3-dimensional search space, the range of each dimension is: ), N solutions are randomly generated to form the initial population P. Each solution is a 3D vector. ;

[0146] For each solution in the initial population P Calculate the alternative solution of the initial population P by dimension. For the first dimension, ,in This yields an opposing population P' containing N opposing solutions;

[0147] The initial population P is merged with the opposing population P' to form a merged population C containing 2N solutions;

[0148] Calculate the fitness value f(X) of all 2N solutions in the merged population C to measure the quality of each solution;

[0149] The N solutions with the smallest fitness values ​​are selected from the merged population C to form the initial population of the new generation, thus completing the population initialization process based on opposition learning.

[0150] It should be noted that the original Harris Eagle algorithm typically uses a random method to generate the initial population, which may result in insufficient population diversity or low quality of initial solutions. In contrast, the opposition learning initialization simultaneously generates the initial solution and its opposition solution, and selects the solution with better fitness to form the population, thus enhancing the diversity of the initial population and its global search capability, providing a better starting point for subsequent iterations of the algorithm.

[0151] Specifically, the process of improving the Harris Eagle optimization algorithm based on the Lévy flight stride includes:

[0152] The uniform random step size in the standard exploration strategy of the Harris Eagle optimization algorithm is replaced with the Lévy flight step size, including:

[0153] ;

[0154] In the formula, This indicates element-wise multiplication; This indicates that during iteration, an individual is randomly selected from the population; Indicates the current iteration time; Step size generation for Lévy flight exploration includes:

[0155] Lévy ) u / | v (1 / );

[0156] in:

[0157] ;

[0158] ;

[0159] In the formula, It is the gamma function. It is an index, usually 1.5.

[0160] It should be noted that the original Harris Eagle algorithm may have issues with local search bias or unreasonable search step size in its individual position update method during the exploration phase. By introducing the Lévy flight step size, a probability distribution model with random long jump characteristics, the algorithm can jump out of local areas with a higher probability during the exploration phase, enhancing global exploration capabilities and reducing the risk of getting trapped in local optima.

[0161] Specifically, the process of improving the Harris Eagle optimization algorithm based on adaptive boundary adjustment includes:

[0162] During the iteration process, the current global optimal solution is periodically used. Centered on the search space, the upper and lower bounds are dynamically shrunk. The search boundary gradually narrows as the number of iterations t increases, as shown below:

[0163] ;

[0164] ;

[0165] Where: δ is the contraction coefficient, such as 0.1; (1-t / T) is a factor that linearly decays from 1 to 0.

[0166] It should be noted that the original Harris Eagle algorithm typically has a fixed search boundary, which may lead to an excessively large (inefficient) or excessively small (missing the optimal solution) search range in later iterations. Adaptive boundary adjustment dynamically reduces the search range around the current optimal solution as the iteration progresses. It retains a larger search space in the early stages of iteration to ensure global exploration, and focuses on local regions in later stages to improve search accuracy, thus balancing the efficiency of global exploration and local development.

[0167] Specifically, improving the Harris Eagle optimization algorithm based on periodic local search includes:

[0168] During the iteration process, every K generations, the current globally optimal solution is used. Starting from the local search algorithm, a fine-grained search is performed. The results obtained from the local search are compared with the current optimal solution and the best solution is selected and updated. After the best solution is selected, the main loop of Harris Eagle continues.

[0169] It should be noted that the original Harris Eagle algorithm's local search relies primarily on its own iterative mechanism, which may be insufficient for fine-tuning the search for the current optimal solution. Periodically (e.g., every 10 iterations) by introducing local search methods such as the Nelder-Mead simplex allows for local fine-tuning and in-depth exploration of the current optimal solution, enhancing the algorithm's ability to explore local areas near the optimal solution and improving the accuracy of the final solution.

[0170] The process of optimizing the optimization model using the improved Harris Eagle algorithm in step S103 includes:

[0171] First, perform initialization:

[0172] Initialization based on opposition learning:

[0173] Randomly generate the initial population , in .

[0174] For each Calculate the complementary solution .

[0175] Choose the one with better adaptability The solutions form a new population.

[0176] In the formula, Represents the initial population, which includes Each individual (solution), each individual It is a D-dimensional vector representing a candidate solution in the search space; The first in the population Individual, ; and : These represent the lower and upper bounds of the search space, respectively; : A D-dimensional random vector, each element of which is uniformly distributed in the interval [0,1], used to randomly initialize the position of an individual; :individual The opposite solution is generated through mirroring to increase the diversity of the population; Population size, i.e., the number of individuals; The dimension of the problem is the number of variables contained in each individual, which represents the number of parameters.

[0177] Then, execute the main loop steps of the Harris Eagle optimization algorithm:

[0178] Step B1: Calculate the fitness value for each individual based on the fitness function. Find the current fitness value. Minimum optimal solution , This represents the best individual found so far, and is the globally optimal solution updated after each iteration;

[0179] Step B2: For each individual, calculate the escape energy. The escape energy E is used to control the phase transition:

[0180] ;

[0181] In the formula, This represents the current iteration number; The total number of iterations is the preset maximum number of iterations; E is the escape energy. As initial energy, = rand(-1, 1), which is a random number uniformly distributed in the interval [-1, 1]; E is the updated energy, E = 2. (1 - t / T), which decreases linearly as iteration t increases;

[0182] Step B3: For each individual The location has been updated:

[0183] During the exploration phase, if ,and ,but:

[0184] ;

[0185] otherwise:

[0186] ;

[0187] During the development phase, if :

[0188] Then let

[0189] if and :

[0190] ;

[0191] if and :

[0192] ;

[0193] if and :

[0194] calculate ;

[0195] if , but Otherwise calculate In the formula, If it is a random vector, , but Otherwise ;

[0196] In the formula, For individuals The new position is calculated using the position update formula; rand() and rand(·) are random numbers uniformly distributed in the interval [0, 1]. This is the absolute value of energy E, used to determine the current stage; Let X be an individual randomly selected from the population; Lévy(β) is the Lévy flight step size, where β is a parameter of the Lévy distribution; mean(X) is the average position vector of population X; r is a random number, r = rand(), used to select the encirclement strategy; Y is a temporary position, Y = - E · | - |;Z is another temporary position, Z = Y + S · Lévy(β), where S is a random vector;

[0197] Step B4: Dynamically narrow the search range around the current optimal solution and construct an adaptive boundary:

[0198] Every The next iteration narrows the search scope around

[0199] ;

[0200] ;

[0201] In the formula, The interval for adaptive boundary updates; and The lower and upper bounds of the initial search space;

[0202] Step B5: Every K = 10K = 10 iterations, for Perform a Nelder-Mead simplex local search and obtain updates. ;

[0203] Step B6: After the maximum number of iterations is completed, output the optimal solution. Optimal solution It is a vector.

[0204] In this embodiment of the disclosure, obtaining the fitness function with the minimum installation error in step S103 includes:

[0205] For the i-th set of data in M ​​sets of measured data, optimize the variables using the installation error assumption of the current iteration. Construct the installation error rotation matrix Among them, optimization variables include:

[0206] ;

[0207] In the formula, This refers to the installation error in orientation; For pitch installation error; For roll installation error; the , ,and All are errors at three installation angles of the vehicle-mounted sensor;

[0208] The installation error rotation matrix Typically, the Euler angle rotation follows the ZYX sequence, which is yaw-pitch-roll, including:

[0209] ;

[0210] in,

[0211] ;

[0212] In the formula, Install an error rotation matrix for the azimuth angle; Install an error rotation matrix for the pitch angle; The roll angle is set to the rotation matrix for error.

[0213] Based on the error rotation matrix The unit vectors measured by the vehicle-mounted sensors are corrected to obtain the unit vectors in the vehicle-mounted sensor coordinate system under ideal error-free installation conditions, and this is combined with the pose data of the vehicle-mounted sensors at that moment. and measuring distance Obtain the geographic coordinates of the target through coordinate transformation. ;

[0214] The geographic coordinates of the target are obtained using the Haversine horizontal spherical distance formula. relative to the target's true coordinates Horizontal spherical distance error between ;

[0215] The formula for the horizontal spherical distance of Haversvine is as follows:

[0216] ;

[0217] ;

[0218] ;

[0219] ;

[0220] ;

[0221] In the formula, rice; The difference in latitude between two points; Let be the latitude of the calculated or target location of the i-th point; Let i be the latitude of the i-th point; The difference in longitude between two points; Let be the longitude of the calculated or target location for the i-th point; Let be the longitude of the i-th point; Intermediate variables calculated based on the difference between latitude and longitude: y is the spherical angular distance between the two points.

[0222] Accumulated horizontal spherical distance error of all measured data The current installation error assumption optimization variables are obtained. The sum of the horizontal spherical distance errors of all targets is used as the fitness function, expressed as:

[0223] ;

[0224] The goal is to minimize In the formula, M is the number of measured data sets. For the first Group of measured data.

[0225] Furthermore, the geographic coordinates of the target are obtained. The specific steps include:

[0226] Step A1: Transform the sensor measurements to the vehicle coordinate system and calculate the unit vector in the sensor coordinate system. :

[0227] ;

[0228] In the formula, The azimuth angle in the sensor coordinate system; The pitch angle in the sensor coordinate system;

[0229] Step A2: Combine installation error rotation matrix The unit vector in the vehicle coordinate system is calculated. The unit vector in the vehicle coordinate system is calculated. :

[0230] ;

[0231] In the formula, Let be the longitudinal distance of the i-th target in the vehicle coordinate system; Let be the lateral distance of the i-th target in the vehicle coordinate system; Let be the vertical height of the i-th target in the vehicle coordinate system, then , The azimuth angle in the sensor coordinate system;

[0232] Step A3: Combine the unit vector obtained in step A2 Based on counterclockwise angle atan2 The counterclockwise angle of the target on the horizontal plane of the vehicle coordinate system is obtained by formula, where, Range of values arrive ;

[0233] Define the azimuth angle in the vehicle coordinate system as clockwise from the X-axis, and convert counterclockwise angles to clockwise angles: Convert to positive values ​​to obtain the final vehicle coordinate system azimuth angle. ,like ,but ,in The value range is 0 to ; Obtain the pitch angle of the vehicle coordinate system , The range of values arrive ;

[0234] Step A4: and Substituting the values ​​into the solution function f, we obtain the calculated target coordinates. :

[0235] ;

[0236] In the formula, It is a known solution function that converts measurements in the vehicle coordinate system into global coordinates; y is the roll angle in the sensor coordinate system.

[0237] Next, proceed to step S104.

[0238] In step S104, the vehicle-mounted sensor performs online compensation and correction on the original sensor measurement values ​​based on the fitness function with minimum installation error, and outputs optimized data results.

[0239] This disclosure also discloses a simulation experiment of a vehicle-mounted sensor installation error calibration method based on an improved Harris Eagle optimization algorithm.

[0240] The algorithm parameters for the simulation experiment include:

[0241] Population size .

[0242] Maximum number of iterations .

[0243] Optimization Dimensions .

[0244] LB=[-0.1745,-0.1745,-0.1745]: Search for the lower bound (in radians).

[0245] UB=[0.1745,0.1745,0.1745]: Search for the upper bound (in radians).

[0246] Lévy distribution parameters.

[0247] Ten targets with known precise coordinates were deployed at a test range. At each point, after the vehicle came to a standstill, its latitude, longitude, altitude, and attitude were recorded via an onboard integrated navigation system. A lidar scanner scanned and recorded the azimuth, elevation, and slant range of each target in its sensor coordinate system. The true coordinates T of the reflector were also recorded. A total of ten sets of valid observation data were collected.

[0248] Define the optimization variable as the installation error vector. Based on prior knowledge, the initial search range for each error component is set to [-0.1745 rad, +0.1745 rad].

[0249] Construct a fitness function f(X). For a given optimization variable X, perform the following for each of the 10 datasets:

[0250] The sensor measurements are corrected using a rotation matrix composed of optimized variables X;

[0251] Using coordinate calculation functions, including transformations from the sensor system to the vehicle system, from the vehicle system to the navigation system, and from the navigation system to the geodetic system, the target's calculated coordinates are determined. ;

[0252] Calculated using the Haversine formula The horizontal error d_horiz between the true value and the actual value T;

[0253] Finally, f(X) represents all 100... The sum of f(X). Our goal is to find the X that minimizes f(X).

[0254] Configure algorithm parameters: population size N=50, maximum number of iterations T=200, local search interval K=20, boundary shrinkage coefficient δ=0.1.

[0255] Run the improved Harris Eagle optimization algorithm. Figure 2 The optimized results of the vehicle-mounted sensor installation error calibration method according to embodiments of this disclosure are shown, such as... Figure 2 As shown, ΔAz: true value 1.5, estimated value 1.4995, error 0.0005; ΔEl: true value -0.8, estimated value -0.7999, error 0.0002; ΔRoll: true value 0.6, estimated value 0.6001, error 0.0001. A comparison of the true and estimated errors is shown below. Figure 3 As shown. During the optimization process, the convergence curve of the algorithm is as follows. Figure 4 As shown. From Figure 5It is evident that, compared to the Genetic Algorithm (GA) and the Particle Swarm Optimization (PSO) algorithm, the improved Harris Eagle optimization algorithm provided in this application achieves the fastest convergence speed for HHO, and the final converged fitness value (total error) is significantly lower than that of the comparative genetic algorithm and particle swarm optimization algorithm, demonstrating the superior performance of this application. The algorithm ultimately outputs the optimal installation error estimate. .

[0256] The second embodiment of the present invention provides an apparatus for calibrating the installation error of vehicle-mounted sensors, comprising:

[0257] The data acquisition unit is configured to measure multiple targets with known precise geographic coordinates at different positions and attitudes using vehicle-mounted sensors, and obtain multiple sets of measured data.

[0258] The model building unit is configured to build an error optimization model with the error values ​​of the three installation angles of the vehicle sensor as optimization variables, and use the sum of the horizontal spherical distance errors between the target geographic coordinates and the real target geographic coordinates in all measured data as the fitness function of the error optimization model.

[0259] The model optimization unit is configured to improve the Harris Eagle optimization algorithm and optimize the error optimization model based on the improved Harris Eagle optimization algorithm. Based on the optimized error optimization model, a fitness function with minimum installation error is obtained by using a local search algorithm.

[0260] The output unit is configured to perform online compensation and correction on the measured data of the vehicle-mounted sensor based on a fitness function with minimum installation error, and output optimized data results.

[0261] The third embodiment of the present invention also provides an electronic device, the electronic device comprising:

[0262] At least one processor; and,

[0263] The memory is communicatively connected to the at least one processor; wherein,

[0264] The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the vehicle sensor installation error calibration method of any of the foregoing embodiments.

[0265] The fourth embodiment of the present invention also provides a non-transitory computer-readable storage medium storing computer instructions for causing the computer to perform the vehicle sensor installation error calibration method described in any of the foregoing embodiments.

[0266] The fifth embodiment of the present invention also provides a computer program product, which includes a computing program stored on a non-transitory computer-readable storage medium. The computer program includes program instructions that, when executed by a computer, cause the computer to perform the vehicle sensor installation error calibration method of any of the foregoing embodiments.

[0267] Figure 6 The illustration shows a method or device 1000 implementing an embodiment of the present invention. In some embodiments, more or fewer devices may be included than illustrated. In some embodiments, it may be implemented using a single or multiple devices. In some embodiments, it may be implemented using cloud-based or distributed devices.

[0268] like Figure 6 As shown, device 1000 includes a processor 1001, which can perform various appropriate operations and processes based on programs and / or data stored in read-only memory (ROM) 1002 or programs and / or data loaded from storage portion 1008 into random access memory (RAM) 1003. Processor 1001 may be a multi-core processor or may contain multiple processors. In some embodiments, processor 1001 may include a general-purpose main processor and one or more special coprocessors, such as a central processing unit (CPU), graphics processing unit (GPU), neural network processor (NPU), digital signal processor (DSP), etc. Various programs and data required for the operation of device 1000 are also stored in RAM 1003. Processor 1001, ROM 1002, and RAM 1003 are interconnected via bus 1004. Input / output (I / O) interface 1005 is also connected to bus 1004.

[0269] The processor and memory described above are used together to execute programs stored in the memory. When the program is executed by a computer, it can implement the methods, steps, or functions described in the above embodiments.

[0270] The following components are connected to I / O interface 1005: an input section 1006 including a keyboard, mouse, touchscreen, etc.; an output section 1007 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and speakers, etc.; a storage section 1008 including a hard disk, etc.; and a communication section 1009 including a network interface card such as a LAN card, modem, etc. The communication section 1009 performs communication processing via a network such as the Internet. A drive 1010 is also connected to I / O interface 1005 as needed. A removable medium 1011, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed on drive 1010 as needed so that computer programs read from it can be installed into storage section 1008 as needed. Figure 6 The diagram only shows a portion of the components and does not imply that the device 1000 only includes... Figure 6 The components shown.

[0271] The systems, devices, modules, or units described in the above embodiments can be implemented by a computer or its associated components. The computer may be, for example, a mobile terminal, smartphone, personal computer, laptop computer, in-vehicle human-machine interface device, personal digital assistant, media player, navigation device, game console, tablet computer, wearable device, smart TV, Internet of Things system, smart home, industrial computer, server, or a combination thereof.

[0272] Although not shown, in this embodiment of the invention, a computer-readable storage medium is provided having a computer program / instructions stored thereon, which, when executed by a processor, implements the method for calibrating the installation error of the vehicle-mounted sensor as described in the embodiment.

[0273] Storage media in embodiments of the present invention include articles that are permanent and non-permanent, removable and non-removable, capable of storing information by any method or technology. Examples of storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transfer medium that can be used to store information accessible by a computing device.

[0274] Although not shown, embodiments of the present invention also provide a computer program product, including: a computer program / instructions that, when executed by a processor, implement the method for calibrating the installation error of an on-board sensor as described in the embodiments.

[0275] The methods, programs, systems, apparatuses, etc., in embodiments of the present invention can be executed or implemented in one or more networked computers, or practiced in a distributed computing environment. In the embodiments of this specification, in these distributed computing environments, tasks can be performed by remote processing devices connected via a communication network.

Claims

1. A method for calibrating the installation error of an on-board sensor, characterized in that, Includes the following steps: Step S1: Measure multiple targets with known precise geographic coordinates at different positions and attitudes using vehicle-mounted sensors to obtain multiple sets of measured data; Step S2: Construct an error optimization model with the error values ​​of the three installation angles of the vehicle sensor as optimization variables, and use the sum of the horizontal spherical distance errors between the target geographic coordinates and the real target geographic coordinates in all measured data as the fitness function of the error optimization model; Step S3: Improve the Harris Eagle optimization algorithm, and optimize the error optimization model based on the improved Harris Eagle optimization algorithm. Based on the optimized error optimization model, use a local search algorithm to obtain the fitness function with minimum installation error. Step S4: The vehicle-mounted sensor performs online compensation and correction on the sensor's measured data based on the fitness function with minimum installation error, and outputs optimized data results.

2. The method for calibrating the installation error of vehicle-mounted sensors according to claim 1, characterized in that, Each set of measured data in step S1 includes: The geographic coordinates of the vehicle-mounted sensors, including longitude, latitude, and altitude; Attitude angles, including true north azimuth, pitch angle, and roll angle; Sensor measurements, including relative azimuth, pitch, and distance; In addition, the target's actual geographic coordinates, including longitude, latitude, and altitude.

3. The method for calibrating the installation error of vehicle-mounted sensors according to claim 1, characterized in that, When obtaining the fitness function with minimum installation error in step S3, the following steps are included: For a specific set of measured data, optimize the variables based on the installation error assumption of the current iteration. Construct the installation error rotation matrix Among them, optimization variables include: ； In the formula, This refers to the installation error in orientation; For pitch installation error; For roll installation error; the , ,and All are errors at three installation angles of the vehicle-mounted sensor; The installation error rotation matrix include: ； in, ； In the formula, Install an error rotation matrix for the azimuth angle; Install an error rotation matrix for the pitch angle; Install the roll angle error rotation matrix; Based on the error rotation matrix The unit vectors measured by the vehicle-mounted sensors are corrected to obtain the unit vectors in the vehicle-mounted sensor coordinate system under ideal error-free installation conditions, and this is combined with the pose data of the vehicle-mounted sensors at that moment. and measuring distance Obtain the geographic coordinates of the target through coordinate transformation. ; The geographic coordinates of the target are obtained using the Haversine horizontal spherical distance formula. relative to the target's true coordinates Horizontal spherical distance error between ; The formula for the horizontal spherical distance of Haversvine is as follows: ； ； ； ； ； In the formula, rice; The difference in latitude between two points; Let be the latitude of the calculated or target location of the i-th point; Let i be the latitude of the i-th point; The difference in longitude between two points; Let be the longitude of the calculated or target location for the i-th point; Let be the longitude of the i-th point; Intermediate variables calculated based on the difference between latitude and longitude: The spherical angular distance between the two points; Accumulated horizontal spherical distance error of all measured data The current installation error assumption is obtained. The sum of the horizontal spherical distance errors of all targets is used as the fitness function, expressed as: ； The goal is to minimize In the formula, M is the number of measured data sets. For the first Group of measured data.

4. The method for calibrating the installation error of vehicle-mounted sensors according to claim 3, characterized in that, Obtain the geographic coordinates of the target. The specific steps include: Step A1: Transform the sensor measurements to the vehicle coordinate system and calculate the unit vector in the sensor coordinate system. : ； In the formula, The azimuth angle in the sensor coordinate system; The pitch angle in the sensor coordinate system; Step A2: Combine installation error rotation matrix and unit vector in the sensor coordinate system The unit vector in the vehicle coordinate system is calculated. : ； In the formula, Let be the longitudinal distance of the i-th target in the vehicle coordinate system; Let be the lateral distance of the i-th target in the vehicle coordinate system; Let be the vertical height of the i-th target in the vehicle coordinate system; then , The azimuth angle in the sensor coordinate system; Step A3: Combine the unit vector obtained in step A2 Based on counterclockwise angle atan2 The counterclockwise angle of the target on the horizontal plane of the vehicle coordinate system is obtained by formula, where, Range of values arrive ; Define the azimuth angle in the vehicle coordinate system as clockwise from the X-axis, and convert counterclockwise angles to clockwise angles: Convert to positive values ​​to obtain the final vehicle coordinate system azimuth angle. ,like ,but ,in The value range is 0 to ; Obtain the pitch angle of the vehicle coordinate system , The range of values arrive ; Step A4: and Substituting the values ​​into the solution function f, we obtain the calculated target coordinates. : ； In the formula, It is a known solution function that converts measurements in the vehicle coordinate system into global coordinates; y is the roll angle in the sensor coordinate system.

5. The method for calibrating the installation error of vehicle-mounted sensors according to claim 1, characterized in that, The process of improving the Harris Eagle optimization algorithm in step S3 includes: An improved Harris Eagle optimization algorithm based on opposition learning, population initialization, Lévy flight exploration, adaptive boundary adjustment, and periodic local search; The process of improving the Harris Eagle optimization algorithm for population initialization based on opposition learning includes: In a given 3-dimensional search space, the range of each dimension is: ), N solutions are randomly generated to form the initial population P. Each solution is a 3D vector. ; For each solution in the initial population P Calculate the alternative solution of the initial population P by dimension. For the first dimension, ,in This yields an opposing population P' containing N opposing solutions; The initial population P is merged with the opposing population P' to form a merged population C containing 2N solutions; Calculate the fitness value f(X) of all 2N solutions in the merged population C to measure the quality of each solution; The N solutions with the smallest fitness values ​​are selected from the merged population C to form the initial population of the new generation, thus completing the population initialization process based on opposition learning. The process of the improved Harris Eagle optimization algorithm based on Lévy flight exploration includes: The uniform random step size in the standard exploration strategy of the Harris Eagle optimization algorithm is replaced with the Lévy flight step size, including: ； In the formula, This indicates element-wise multiplication; This indicates that during iteration, an individual is randomly selected from the population; Indicates the current iteration time; Step size generation for Lévy flight exploration includes: Left( ) u / | in ( 1 / ); in: ； ； In the formula, It is the gamma function. It is an index; The process of improving the Harris Eagle optimization algorithm based on the aforementioned adaptive boundary adjustment includes: During the iteration process, the current global optimal solution is periodically used. Centered on the search space, the upper and lower bounds are dynamically shrunk. The search boundary gradually narrows as the number of iterations t increases, as shown below: ； ； Where: δ is the contraction coefficient; (1-t / T) is a factor that linearly decays from 1 to 0; The process of improving the Harris Eagle optimization algorithm based on periodic local search includes: During the iteration process, every K generations, the global optimal solution at the current moment is used. Starting from a point, a local search algorithm is used to solve the problem. The result of the local search is compared with the fitness value of the current best solution. The solution with the smaller fitness value is selected as the new best solution, and the population is updated. The updated population continues to execute the main loop steps of the Harris Eagle Optimization Algorithm until the maximum number of iterations is reached.

6. The method for calibrating the installation error of vehicle-mounted sensors according to claim 5, characterized in that, The main loop steps of the Harris Eagle optimization algorithm include: Step B1: Calculate the fitness value for each individual based on the fitness function. Find the current fitness value. Minimum optimal solution , This represents the best individual found so far, and is the globally optimal solution updated after each iteration; Step B2: For each individual, calculate the escape energy. The escape energy E is used to control the phase transition: ； In the formula, This represents the current iteration number; The total number of iterations is the preset maximum number of iterations; E is the escape energy. As initial energy, = rand(-1, 1), which is a random number uniformly distributed in the interval [-1, 1]; E is the updated energy, E = 2. (1 - t / T), which decreases linearly as iteration t increases; Step B3: For each individual The location has been updated: During the exploration phase, if ,and ,but: ； otherwise: ； During the development phase, if : Then let ; if and : ； if and : ； if and : calculate ; if , but Otherwise calculate In the formula, If it is a random vector, , but Otherwise ; In the formula, For individuals The new position is calculated using the position update formula; rand() and rand(·) are random numbers uniformly distributed in the interval [0, 1]. This is the absolute value of energy E, used to determine the current stage; Let X be an individual randomly selected from the population; Lévy(β) is the Lévy flight step size, where β is a parameter of the Lévy distribution; mean(X) is the average position vector of population X; r is a random number, r = rand(), used to select the encirclement strategy; Y is a temporary position, Y = - E · | - |;Z is another temporary position, Z = Y + S · Lévy(β), where S is a random vector; Step B4: Dynamically narrow the search range around the current optimal solution and construct an adaptive boundary: Every The next iteration narrows the search scope around : ； ； In the formula, The interval for adaptive boundary updates; and The lower and upper bounds of the initial search space; Step B5: Every K = 10K = 10 iterations, for Perform a Nelder-Mead simplex local search and obtain updates. ; Step B6: After the maximum number of iterations is completed, output the optimal solution. .

7. A device for calibrating the installation error of an on-board sensor, characterized in that, The method for calibrating the installation error of vehicle-mounted sensors according to any one of claims 1-6 includes: The data acquisition unit is configured to measure multiple targets with known precise geographic coordinates at different positions and attitudes using vehicle-mounted sensors, and obtain multiple sets of measured data. The model building unit is configured to build an error optimization model with the error values ​​of the three installation angles of the vehicle sensor as optimization variables, and use the sum of the horizontal spherical distance errors between the target geographic coordinates and the real target geographic coordinates in all measured data as the fitness function of the error optimization model. The model optimization unit is configured to improve the Harris Eagle optimization algorithm and optimize the error optimization model based on the improved Harris Eagle optimization algorithm. Based on the optimized error optimization model, a fitness function with minimum installation error is obtained by using a local search algorithm. The output unit is configured to perform online compensation and correction on the measured data of the vehicle-mounted sensor based on a fitness function with minimum installation error, and output optimized data results.

8. An electronic device, characterized in that, The electronic device includes: At least one processor; and, The memory is communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method for calibrating the installation error of the vehicle sensor as described in any one of claims 1-6.

9. A non-transitory computer-readable storage medium, characterized in that, The non-transitory computer-readable storage medium stores computer instructions for causing the computer to perform the method for calibrating the installation error of vehicle sensors as described in any one of claims 1-6.

Images