Methods for calibrating Doppler sensor systems, methods for estimating vehicle motion vectors, data processing equipment, computer programs, computer-readable storage media, and Doppler sensor systems.
By estimating the vehicle motion vector and calibrating the Doppler sensor system, the problems of angle detection error and signal interference after Doppler sensor installation were solved, enabling accurate vehicle position and speed estimation under high-dynamic driving conditions and unreliable satellite signals.
Patent Information
- Application Number
- CN202610647091.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Priority Date
- 2025-05-14
- Filing Date
- 2026-05-12
- Publication Date
- 2026-11-17
AI Technical Summary
After a Doppler sensor system is installed on a vehicle, angle detection errors and signal interference may occur, leading to inaccurate vehicle odometer readings, which are particularly difficult to correct under conditions of high-dynamic driving and unreliable global navigation satellite system signals.
By obtaining environmental detection data from Doppler sensors, vehicle motion vectors can be estimated and angle detection errors can be determined. Data models or artificial neural networks can be used to calibrate the Doppler sensor system, reducing reliance on global navigation satellite systems.
It improves the calibration accuracy and reliability of the Doppler sensor system, especially in areas where global navigation satellite system signals are not readily available, and enhances the estimation accuracy of vehicle motion vectors, making it suitable for high-dynamic driving scenarios.
Smart Images

Figure CN123112840A_ABST
Abstract
Description
Technical Field
[0001] This disclosure relates to a method for calibrating a Doppler sensor system. The Doppler sensor system can be installed in a vehicle.
[0002] Furthermore, this disclosure relates to a method for estimating the vehicle motion vector. This method estimates the vehicle motion vector while the vehicle is in motion.
[0003] Furthermore, this disclosure relates to data processing equipment, computer programs, and computer-readable storage media.
[0004] Additionally, this disclosure relates to a Doppler sensor system that can be installed in a vehicle and a calibration system for the Doppler sensor system. Background Technology
[0005] Vehicle odometers are methods for estimating the position and orientation of an ego-vehicle, and how that position and / or orientation changes over time (i.e., the vehicle's speed). For this purpose, data from wheel speed sensors can be used. This wheel speed sensor data can be corrected using GPS position data to compensate for potential offset errors due to the accumulation of wheel speed over time. However, GPS connectivity may not be reliably available at every location. Furthermore, in highly dynamic driving maneuvers, wheel slippage can further increase the inaccuracy of the vehicle odometer, which may not be compensated for in a timely manner using GPS position data.
[0006] Doppler sensor systems have the potential to provide accurate vehicle odometer readings during highly dynamic driving maneuvers without relying on global navigation satellite systems. However, when Doppler sensors are mounted on a vehicle, their actual position and orientation on the vehicle may be misaligned compared to their desired position and orientation. Furthermore, for aesthetic reasons, Doppler sensors are typically concealed behind the vehicle's exterior housing. This housing can interfere with the signals from the Doppler sensors. Because the vehicle's exterior housing may not possess uniform Doppler signal transmission properties, the Doppler sensors behind the housing may additionally experience detection errors due to variations in the Doppler sensor's field of view. Such angular detection errors can lead to inaccurate vehicle odometer estimates. Summary of the Invention
[0007] Therefore, one object of this disclosure is to provide an improved method for calibrating a Doppler sensor system that can be mounted on a vehicle, such that the angle detection error of the Doppler sensor system and / or the vehicle motion vector can be estimated with increased accuracy.
[0008] This problem is at least partially solved or mitigated by the subject matter of this disclosure.
[0009] According to a first aspect, a method for calibrating a Doppler sensor system that can be mounted on a vehicle is provided. The method includes: - Obtain first data indicating the environment of the vehicle derived from multiple Doppler sensors that can be mounted on the vehicle, wherein the Doppler sensor detection includes angle information. - Estimate the vehicle motion vector based on the first data, and determine the angle detection error based on the first data and the estimated vehicle motion vector, or The trained data model is used to determine the angle detection error and / or vehicle motion vector based on the first data, and - Provide second data indicating the angle detection error and / or the vehicle motion vector.
[0010] Doppler sensor systems can include any type of Doppler sensor, such as radar sensors, ultrasonic sensors, and / or laser detection and ranging (LiDAR, also known as lidar) sensors. A common characteristic of all Doppler sensors is that a Doppler signal is emitted by the Doppler sensor. The Doppler signal can be reflected by an object in the Doppler sensor's field of view. Depending on the object's velocity relative to the Doppler sensor, the wavelength of the Doppler signal can change during reflection. The Doppler sensor can receive at least a portion of the reflected Doppler signal. Based on the difference between the wavelength of the emitted and received Doppler signals, the velocity of the object relative to the Doppler sensor can be obtained. More precisely, the velocity of the object relative to the Doppler sensor along the direction of the direct connection between the object and the Doppler sensor can be determined. Such a velocity can also be referred to as the radial rate or radial velocity of the object relative to the Doppler sensor. Furthermore, such a velocity can be referred to as the measured Doppler rate or measured Doppler velocity. Additionally, the angular position of the object relative to the Doppler sensor can be determined based on the fact that the angular position of the Doppler sensor on the vehicle is known. The measured Doppler velocity and the object's angular position are included in the acquired first data. Note that acquiring the first data may include receiving or determining the first data. In spherical coordinates, the object's angular position may include the object's azimuth and elevation angles relative to the Doppler sensor. By comparing the object's angular positions relative to the Doppler sensor from different time points, the object's angular velocity vector relative to the Doppler sensor can be obtained.
[0011] In this context, it should be understood that the object is located outside the vehicle. In other words, the object is located in the environment, i.e., in the environment surrounding the vehicle. The first data includes Doppler sensor detections of the vehicle environment from at least two Doppler sensors that can be mounted on the vehicle. The at least two Doppler sensors may include overlapping fields of view, meaning they can sense the same object within their respective fields of view but from different perspectives. The different perspectives are due to the fact that the at least two Doppler sensors can be mounted at different locations on the vehicle. Alternatively, the at least two Doppler sensors may include non-overlapping fields of view. Thus, the at least two Doppler sensors can sense different objects or different portions of the same object within their respective fields of view. In either case, the vehicle motion vector, i.e., the vector indicating the vehicle's position, orientation, and velocity, can be estimated based on the first data. More precisely, the vehicle motion vector can be estimated from the radial velocity vector of the object in the respective fields of view of the at least two Doppler sensors. This is possible under the assumption that the relative positions and orientations of the Doppler sensors in the Doppler sensor system remain constant while obtaining the first data indicating the Doppler sensor detections of the vehicle's environment.
[0012] According to a first alternative of this disclosure, the angle detection error of each of the multiple Doppler sensors can be obtained by comparing the estimated vehicle motion vector with the radial velocity vector of the object detected by a separate Doppler sensor. Therefore, the angle detection error of each of the multiple Doppler sensors is obtained based on the first data and the estimated vehicle motion vector. In other words, the angle detection error of each of the multiple Doppler sensors is obtained based on the first data and the estimated vehicle motion vector. Again, in other words, when the first data and the estimated vehicle motion vector are provided as input to the comparison function, the angle detection error of each of the multiple Doppler sensors is obtained as the output.
[0013] According to a second alternative of this disclosure, a trained data model can be used to determine angle detection errors and / or vehicle motion vectors. For this purpose, the data model is designed and trained to take first data, indicating the vehicle environment from multiple Doppler sensors that can be mounted to the vehicle, as input. The data model can be trained by providing the data model with signals indicating the wheel speeds of the vehicle's wheels from wheel speed sensors and / or vehicle odometer data based on position signals from a Global Navigation Satellite System (GNSS). During training, the vehicle odometer data based on the signals from the wheel speed sensors and / or the position signals from the GNSS can be used as ground reality. Still during training, the trainable data model can be configured to minimize the prediction error between the vehicle motion vector prediction based on the first data and the vehicle motion vector based on the ground reality. Additionally or alternatively, during training, the trainable data model can be configured to minimize the prediction error between the angle detection error of each of the multiple Doppler sensors based on the first data and the angle detection error based on the ground reality. Therefore, after training and during the application of the trained data model, the trained data model can be configured to provide the angle detection error of each of the multiple Doppler sensors and / or the vehicle motion vector as output when the first data is provided as input. It should be understood that during training and application, different datasets of the first data are provided to the data model, for example, from a training drive sequence and from a testing drive sequence.
[0014] The determined vehicle motion vector and / or the determined angle detection error of each of the multiple Doppler sensors are provided as outputs of a method for calibrating a vehicle-mountable Doppler sensor system. In other words, the method for calibrating a vehicle-mountable Doppler sensor system takes first data indicating the vehicle's environment detected by the Doppler sensors as input data and provides second data indicating the vehicle motion vector and / or the angle detection error of each of the multiple Doppler sensors as output data. Therefore, the angle detection error can be used to correct the obtained first data. This has the effect that the vehicle's position, orientation, and speed relative to its environment can be determined from the first data without requiring data corresponding to the data items from the Global Navigation Satellite System (GNSS). In other words, the provided second data does not require correction from data from the GNSS. This increases the reliability of the method for calibrating a vehicle-mountable Doppler sensor system, especially in areas where signals from the GNSS are not easily obtained, such as in underground parking lots, under bridges, or in tunnels. Furthermore, particularly for high-dynamic driving maneuvers, the accuracy of the second data provided can be improved compared to the method of using wheel speed sensors to obtain the rotational speed of the vehicle's wheels, which may slip during high-dynamic driving maneuvers.
[0015] In one example, the first data is associated with the vehicle traveling on a straight path and / or the vehicle's yaw rate being below a predetermined yaw rate threshold. If the vehicle is traveling on a straight path, the vehicle's yaw rate is zero or at least substantially zero. This means that the vehicle does not turn or rotate around its height axis. The vehicle's height axis is defined as an axis perpendicular to the road surface on which the vehicle is traveling. If the first data corresponds to a driving scenario where the vehicle is traveling on a straight path, the provided angle detection error and / or the provided vehicle motion vector may have particularly high accuracy. If the first data corresponds to a driving scenario where the vehicle is traveling at a yaw rate below the predetermined yaw rate threshold, acceptable accuracy of the provided angle detection error and / or the provided vehicle motion vector can still be achieved. Furthermore, if the vehicle's yaw rate is zero, substantially zero, or at least below the predetermined yaw rate threshold, the orientation of the vehicle motion vector in the estimation of the vehicle motion vector based on the first data can be preset to a direction parallel to the vehicle's standard forward travel direction. The vehicle's standard forward travel direction is the direction in which the vehicle travels without turning (i.e., without turning left or right) and moving forward. This can facilitate the estimation of vehicle motion vector and / or angle detection errors based on the first data. Specifically, the vehicle motion vector and / or angle detection errors for each of the multiple Doppler sensors can be provided with increased accuracy. Simultaneously, the computational workload for estimating vehicle motion vector and / or angle detection errors based on the first data can be reduced.
[0016] In one example, the vehicle motion vector is estimated based on additional third data indicating the vehicle's yaw rate. In this example, the first data does not need to be associated with the vehicle traveling on a straight path. This has the effect that, in any driving situation, the angle detection error of each of the multiple Doppler sensors and / or the vehicle motion vector can be estimated based on the first data. The third data indicating the vehicle's yaw rate can be obtained from a yaw rate sensor that is part of the vehicle. The yaw rate sensor may include the vehicle's compass or, preferably, a gyroscope. Based on the vehicle's yaw rate, the vehicle's heading at each location of the multiple Doppler sensors can be estimated. In other words, the direction of motion of each of the multiple Doppler sensors can be estimated based on the vehicle's yaw rate. For this purpose, rigid body kinematics can be applied, i.e., assuming that the vehicle equipped with the multiple Doppler sensors is a rigid object. Estimating the vehicle motion vector based on the additional third data indicating the vehicle's yaw rate allows for the estimation of the angle detection error of each of the multiple Doppler sensors in a simple and reliable manner. More specifically, the azimuth angle error of each of the multiple Doppler sensors can be obtained with a high level of accuracy.
[0017] In one example, estimating the vehicle motion vector involves solving an optimization problem or determining the median of the Doppler velocity of the first data, corrected by the corresponding angle information of the first data. Additionally or alternatively, determining the angle error involves solving an optimization problem. Solving the optimization problem for estimating the vehicle motion vector may include finding a vehicle motion vector that minimizes the difference between the measured Doppler velocity of an object in the field of view of the Doppler sensor and the modeled Doppler velocity of the object in the field of view of the Doppler sensor. In this case, the modeled Doppler velocity of the object in the field of view of the Doppler sensor includes the measured Doppler velocity corrected by the angle detection error. Solving the optimization problem to determine the angle error may include defining a cost function that minimizes the difference between the measured Doppler velocity of an object in the field of view of the Doppler sensor and the modeled Doppler velocity of the object in the field of view of the Doppler sensor. The cost function may be a function of the angle detection errors of the azimuth and elevation angles of the Doppler sensor. The cost function can be solved for the angle detection errors of the azimuth and elevation angles of the Doppler sensor using orthogonal distance regression or least squares methods. Determining the median of the Doppler velocity of the first data, corrected by the corresponding angle information of the first data, can be understood as the Doppler velocity of objects in the vehicle environment, corrected by angle detection errors, being averaged over the field of view or at least a subset thereof to estimate the vehicle motion vector. This returns to determining the most probable Doppler velocity. Alternatively, averaging may include determining an arithmetic or geometric mean. Considering the angle detection errors of multiple Doppler sensors has the effect that the vehicle motion vector of a vehicle equipped with multiple Doppler sensors can be estimated more accurately.
[0018] In one example, the determined angle detection error includes local angle detection error and / or global angle detection error. Local angle detection error refers to the angle detection error at each point in the field of view of the Doppler sensor, specific to the Doppler sensor system. Therefore, the local angle error of the Doppler sensor can vary across its field of view. This could be, for example, due to the vehicle's outer housing, behind which the Doppler sensor is mounted. The vehicle's outer housing may not include uniform Doppler signal transmission properties across the entire field of view of the Doppler sensor. The vehicle's outer housing may be referred to as a radar fascia, and may include, for example, a vehicle bumper, emblem, or cover. Global angle detection error refers to the angle detection error systematically applied to the entire field of view of the Doppler sensor in the Doppler sensor system. For example, this could be due to misalignment of the Doppler sensor relative to its desired position. This misalignment can occur during vehicle assembly or use, particularly due to deformation of the mounting flange of the vehicle with the Doppler sensor mounted in an accident. The global angular error of a Doppler sensor can be determined by averaging the angular errors of the Doppler sensor across its entire field of view. Averaging can include calculating the geometric mean, arithmetic mean, or median. Alternatively or additionally, averaging can include determining a weighted average. The remaining portion of the angular error that cannot be explained by the global angular error can be defined as the local angular error. The distinction between the global and local angular errors of a Doppler sensor has the effect that even subtle and varying distortions of the Doppler sensor signal across its field of view can be taken into account. Therefore, the Doppler sensor system can be calibrated with high accuracy, and the vehicle motion vector can be obtained with a low level of uncertainty.
[0019] In one example, the method further includes using a stochastic search algorithm and / or gradient descent optimization to enhance the accuracy of the estimated vehicle motion vector. The stochastic search algorithm and / or gradient descent optimization minimize the difference between the modeled Doppler velocity and the measured Doppler velocity of objects in the vehicle environment. The stochastic search algorithm and / or gradient descent optimization iteratively estimate the global and local angle errors of all Doppler sensors in the Doppler sensor system. The stochastic search algorithm and / or gradient descent optimization can be initialized with the azimuth and elevation angles of the first data (i.e., Doppler sensor detection of the vehicle environment). The estimates of the global and local angle errors can be subtracted from the results of the azimuth and elevation angles from previous iterations to achieve convergence of the vehicle motion vector estimation. Therefore, the effect of this example is improved accuracy of the vehicle motion vector, i.e., the difference between the vehicle motion vector provided by this method and the vehicle's true vehicle motion vector is minimized.
[0020] In one example, the trained data model includes an artificial neural network. The artificial neural network may include an input layer capable of receiving first data from multiple Doppler sensors of a Doppler sensor system. This may mean that the number of input nodes included in the input layer is equal to or a multiple of the number of Doppler sensors. In a particular example, the input layer includes a number of input nodes equal to the number of points in the field of view of the multiple Doppler sensors of the Doppler sensor system. In other words, the number of input nodes may be the same as the resolution of the Doppler sensor system. In another example, a separate but identical artificial neural network may be used to calibrate each Doppler sensor of the Doppler sensor system. It should be understood that each separate but identical artificial neural network can be trained using first data derived from the Doppler sensors associated with the artificial neural network. Therefore, even using the same artificial neural network, each artificial neural network can be trained and customized to calibrate the Doppler sensors of the Doppler sensor system associated with that artificial neural network. The artificial neural network can be trained on first data from a training driving sequence. Additionally, vehicle odometer data (i.e., from wheel speed sensors, yaw rate sensors, and / or from a global navigation satellite system) can be obtained as ground-based real-time data during the training driving sequence. Artificial neural networks can be trained to minimize the discrepancy between their predictions (i.e., predicted angle detection errors and predicted vehicle motion vectors) and ground-based data. Therefore, supervised learning methods can be used to train artificial neural networks. It should be understood that the artificial neural network can include output layers with output nodes capable of providing vehicle motion vectors and / or angle detection errors. In a specific example, there can be three output nodes to provide the vehicle motion vector, i.e., one node for each coordinate direction. Additionally or alternatively, each Doppler sensor can have two output nodes to provide the global angle errors of the Doppler sensor's azimuth and elevation angles. Further, additionally or alternatively, the number of output nodes for each Doppler sensor can be twice the number of detection points in that Doppler sensor, i.e., the number of output nodes is twice the resolution of the Doppler sensor. Therefore, local angle errors for azimuth and elevation can be provided for each detection point of each Doppler sensor. Using artificial neural networks to calibrate a Doppler sensor system has the following effects: it enables rapid convergence to an accurate solution for the vehicle motion vector and / or the angle detection error of the Doppler sensor system. Furthermore, it reduces the manual programming work in the development phase of the method. Furthermore, artificial neural networks can adapt to changes in angle detection errors during vehicle use. For example, if the vehicle is already in use, i.e., with the customer, variations in local angle errors for both azimuth and elevation can be taken into account. These local angle errors might occur, for example, due to an accident that causes deformation of the vehicle's outer shell, with the Doppler sensors of the Doppler sensor system mounted behind the outer shell.
[0021] In one example, estimating the vehicle motion vector involves determining a first model of stationary and moving objects based on first data, and estimating an initial vehicle motion vector based on the first model. The first model can be understood as a first coarse estimate of stationary and moving objects in the vehicle environment. Based on the first coarse estimate of stationary objects in the vehicle environment, a first coarse estimate of the vehicle motion vector, i.e., the initial vehicle motion vector, can be obtained. The initial vehicle motion vector can be used in further steps of the method to enhance the convergence speed to a more accurate vehicle motion vector. Convergence from the first coarse estimate of the vehicle motion vector to a more accurate vehicle motion vector can be faster than convergence from a vehicle motion vector obtained based on wheel speed, yaw rate, and / or global navigation satellite system data, or even from a random estimate of the vehicle motion vector.
[0022] In one example, estimating the vehicle motion vector involves determining a second model of stationary and moving objects based on first data and the initial vehicle motion vector. This second model can be understood as a second, more accurate estimate of stationary and moving objects in the vehicle environment. Specifically, the cost function to be minimized and solved for the angle detection errors (i.e., azimuth and elevation angle detection errors) can be defined based on the second model of stationary and moving objects. Providing models of stationary and moving objects in a two-stage process, and thus providing a highly accurate second model, has the effect that even when calibrating the Doppler sensor system using only first data from a single point in time, the angle detection errors can be determined to a satisfactory level of accuracy.
[0023] According to a second aspect, a method is provided for estimating the vehicle motion vector when a vehicle is in motion. The method includes: - Obtain first data indicating the environment of the moving vehicle detected by multiple Doppler sensors installed on the vehicle. - The Doppler sensor system installed in the vehicle is calibrated using the method according to the first aspect, or wherein the Doppler sensor system is set to a state that has already been calibrated using the method according to the first aspect. - Based on the first data, determine the models of stationary and moving objects, and - Estimate the vehicle motion vector based on the model.
[0024] The method according to the second aspect of this disclosure aims to provide an estimate of vehicle motion based on data calibrated to account for the angle detection error of the Doppler sensor system. To this end, first data of objects in the exterior (i.e., the surrounding environment, i.e., the vehicle's environment) of a plurality of Doppler sensors mounted on a Doppler sensor system is obtained. Obtaining the first data can be understood as receiving or determining the first data. Note that the first data can be obtained regardless of any particular yaw rate of the vehicle. In other words, the method according to the second aspect can be performed in any driving scenario of the vehicle. The first data can be corrected by second data, particularly by the angle detection error of the Doppler sensor system, as provided by the method of the first aspect. Alternatively, the Doppler sensor system can be provided in a calibrated state, i.e., the first data obtained by the Doppler sensor system during the execution of the method of the second aspect is already corrected first data, i.e., taking into account the angle detection error of the Doppler sensor system. Based on the corrected first data, models of stationary and moving objects in the vehicle environment are obtained. In particular, the vehicle motion vector can be estimated based on the Doppler velocity of the stationary objects. The method according to the second aspect has the following effect: even in high-dynamic driving scenarios where the vehicle's wheels may slip, and / or when position signals from a global navigation satellite system are unavailable or at least not available in a timely manner, the vehicle's motion vector—that is, the vector indicating the vehicle's position, orientation, and velocity—can be estimated accurately. This may be particularly advantageous for autonomous or at least partially autonomous vehicles that rely primarily on knowing the vehicle's position relative to objects in its environment, since it is not expected that the human driver will provide control input to the vehicle entirely, or at least to a reduced extent. Driving maneuvers of autonomous or at least partially autonomous vehicles are more accurate and reliable if the vehicle knows its precise motion vector within a model of stationary and moving objects in its environment, and if the vehicle's motion vector has been corrected for angle detection errors by a Doppler sensor system.
[0025] The methods of the first and second aspects can be implemented at least in part by a computer, and can be implemented in software or hardware, or both. Furthermore, the methods of the first and second aspects can be executed by computer program instructions running on a device providing data processing capabilities. The data processing device can be a suitable computing device, such as an electronic control module, or it can be a distributed computer system. The data processing device or computer can each include one or more of a processor, memory, data interface, etc.
[0026] According to a third aspect, a data processing apparatus is provided, comprising means for performing the method according to the first aspect and / or for performing the method according to the second aspect. Providing such a data processing apparatus has the effect that the position, orientation, and speed of a vehicle relative to its environment can be determined based on first data without requiring data corresponding to data items from a Global Navigation Satellite System (GNSS). In other words, the provided second data does not require correction from data from a GNSS. This increases the reliability of methods for calibrating Doppler sensor systems that can be mounted on vehicles, particularly in areas where signals from GNSS are not easily obtained, such as in underground parking lots or under bridges. Furthermore, when providing the second data, particularly for high-dynamic driving maneuvers, the accuracy of the provided second data can be improved compared to methods using wheel speed sensors to obtain the rotational speed of the vehicle's wheels, which may slip during high-dynamic driving maneuvers. Moreover, providing such a data processing apparatus has the effect that the vehicle's motion vector can be accurately estimated even in high-dynamic driving scenarios where the vehicle's wheels may slip, and / or when position signals from a GNSS are unavailable or at least not timely available. This can be particularly advantageous for autonomous or at least partially autonomous vehicles that rely primarily on knowing the vehicle's position relative to objects in its environment, as it is not expected that human drivers will provide control input to the vehicle entirely, or at least to a reduced extent. Driving maneuvers of autonomous or at least partially autonomous vehicles are more accurate and reliable if the vehicle knows its precise vehicle motion vectors within a model of stationary and moving objects in its environment, and if these vehicle motion vectors have been corrected for angle detection errors by a Doppler sensor system.
[0027] According to a fourth aspect, a computer program including instructions is provided, which, when executed by a computer, cause the computer to perform the method according to the first aspect and / or the method according to the second aspect. Providing such a computer program has the effect that the position, orientation, and speed of a vehicle relative to its environment can be determined from first data without requiring data corresponding to data items from a Global Navigation Satellite System (GNSS). In other words, the provided second data does not require correction from data from a GNSS. This increases the reliability of methods for calibrating Doppler sensor systems that can be mounted on vehicles, particularly in areas where signals from a GNSS are not easily obtained, such as in underground parking lots, under bridges, or in tunnels. Furthermore, when the second data is provided, particularly for high-dynamic driving maneuvers, the accuracy of the provided second data can be improved compared to methods using wheel speed sensors to obtain the rotational speed of the vehicle's wheels, which may slip during high-dynamic driving maneuvers. Moreover, providing such a data processing device has the effect that the vehicle's motion vector can be accurately estimated even in high-dynamic driving scenarios where the vehicle's wheels may slip, and / or in situations where position signals from a GNSS are unavailable or at least not timely available. This can be particularly advantageous for autonomous or at least partially autonomous vehicles that rely primarily on knowing the vehicle's position relative to objects in its environment, as it is not expected that human drivers will provide control input to the vehicle entirely, or at least to a reduced extent. Driving maneuvers of autonomous or at least partially autonomous vehicles are more accurate and reliable if the vehicle knows its precise vehicle motion vectors within a model of stationary and moving objects in its environment, and if these vehicle motion vectors have been corrected for angle detection errors by a Doppler sensor system.
[0028] According to a fifth aspect, a computer-readable storage medium is provided comprising instructions that, when executed by a computer, cause the computer to perform the method according to the first aspect and / or the method according to the second aspect. Providing such a computer-readable storage medium has the effect that the position, orientation, and speed of a vehicle relative to its environment can be determined based on first data without requiring data corresponding to data items from a Global Navigation Satellite System (GNSS). In other words, the provided second data does not require correction from data from a GNSS. This increases the reliability of methods for calibrating Doppler sensor systems that can be mounted on vehicles, particularly in areas where signals from a GNSS are not easily obtained, such as in underground parking lots or under bridges. Furthermore, when the second data is provided, particularly for high-dynamic driving maneuvers, the accuracy of the provided second data can be improved compared to methods using wheel speed sensors to obtain the rotational speed of the vehicle's wheels, which may slip during high-dynamic driving maneuvers. Moreover, providing such a data processing device has the effect that the vehicle's motion vector can be accurately estimated even in high-dynamic driving scenarios where the vehicle's wheels may slip, and / or when position signals from a GNSS are unavailable or at least not timely available. This can be particularly advantageous for autonomous or at least partially autonomous vehicles that rely primarily on knowing the vehicle's position relative to objects in its environment, as it is not expected that human drivers will provide control input to the vehicle entirely, or at least to a reduced extent. Driving maneuvers of autonomous or at least partially autonomous vehicles are more accurate and reliable if the vehicle knows its precise vehicle motion vectors within a model of stationary and moving objects in its environment, and if these vehicle motion vectors have been corrected for angle detection errors by a Doppler sensor system.
[0029] According to a sixth aspect, a Doppler sensor system that can be installed in a vehicle is provided. The system includes: - Multiple Doppler sensors that can be installed on the vehicle, and - According to the data processing equipment of the third party, The plurality of Doppler sensors are communicatively connected to the data processing device.
[0030] The communication connection between multiple Doppler sensors and the data processing device enables the data processing device to obtain first data detected by the Doppler sensors indicating the vehicle's environment. The first data originates from multiple Doppler sensors that can be installed on the vehicle. Therefore, the methods according to the first aspect and the second aspect can be performed using the Doppler sensor system according to the sixth aspect. In other words, the Doppler sensor system according to the sixth aspect can be calibrated using the method according to the first aspect. Furthermore, the vehicle motion vector of a vehicle equipped with the Doppler sensor system can be estimated reliably and accurately.
[0031] According to a seventh aspect, a calibration system for a Doppler sensor system is provided. The calibration system includes a data processing apparatus according to a third aspect. This has the effect of using the calibration system according to the seventh aspect to perform the method according to the first aspect. In other words, the calibration system according to the seventh aspect can be used to calibrate a Doppler sensor system according to the sixth aspect.
[0032] It should be noted that the above examples can be combined with each other, regardless of the aspects involved. Thus, the methods of the first and second aspects can be combined with the structural features of the Doppler sensor system and calibration system, and similarly, the Doppler sensor system and calibration system can be combined with the features described above regarding the methods of the first and second aspects.
[0033] These and other aspects of this disclosure will become apparent from the examples described below and will be illustrated with reference to the examples described below. Attached Figure Description
[0034] Examples of this disclosure will now be described with reference to the following figures.
[0035] Figure 1 A self-driving vehicle including a Doppler sensor system according to the present disclosure is shown, wherein the position of a following vehicle is estimated by the Doppler sensor system. Figure 2 Shown in a more detailed view Figure 1 The vehicle, as shown, also includes a data processing device according to the present disclosure and a calibration system for a Doppler sensor system according to the present disclosure. Figure 3 It shows the area behind the vehicle's outer shell. Figure 2 The Doppler sensor system of the vehicle's Doppler sensor, Figure 4 It shows Figure 1 and Figure 2The graph shows the Doppler sensor detection of the vehicle's Doppler sensor system. The x-axis of the graph represents the azimuth angles, and the y-axis represents the longitudinal velocity of the detected object relative to the vehicle. Figure 5 A graph showing the detection of a single Doppler sensor is presented, where the x-axis represents the azimuth angle and the y-axis represents the radial velocity of the detected object relative to the vehicle. Figure 6 A comparison of the vehicle's speed during a driving sequence is shown, obtained by the vehicle's internal measurement unit, satellite navigation system, and a method for estimating the vehicle's motion vector while the vehicle is in motion, according to this disclosure. Figure 7 A first embodiment of a method for estimating a vehicle's motion vector while the vehicle is in motion, according to the present disclosure, is shown, wherein the first method includes a method for calibrating a Doppler sensor system that can be mounted to a vehicle, according to the present disclosure. Figure 8 A second embodiment of a method for estimating a vehicle's motion vector while the vehicle is in motion, according to the present disclosure, is shown, wherein the method includes a method for calibrating a Doppler sensor system that can be mounted to a vehicle, according to the present disclosure. Figure 9 A third embodiment of a method for estimating a vehicle's motion vector while the vehicle is in motion, according to the present disclosure, is shown, wherein the method includes a method for calibrating a Doppler sensor system that can be mounted to a vehicle, according to the present disclosure. Detailed Implementation
[0036] The accompanying drawings are merely schematic representations and are intended to illustrate this disclosure only. In principle, identical or equivalent elements have the same reference numerals.
[0037] Figure 1 The vehicle 10 is shown.
[0038] The vehicle 10 includes four wheels 12 (see...) Figure 2 Wheel speed sensors 14 are mounted to each wheel 12 of the vehicle to capture the rotational speed of the associated wheel 12.
[0039] The vehicle 10 also includes a yaw rate sensor. In this example, the yaw rate sensor is a gyroscope 16. The gyroscope 16 is capable of determining the rotational speed of the vehicle about its height axis, i.e., the yaw rate of the vehicle 10. The height axis 18 of the vehicle 10 is defined as an axis perpendicular to the surface of the road 20 on which the vehicle 10 is traveling.
[0040] In addition to the height axis 18, the vehicle 10 also includes a longitudinal axis 22 and a lateral axis 24. The longitudinal axis 22 is oriented parallel to the standard forward direction of travel of the vehicle 10. The lateral axis 24 is perpendicular to both the height axis 18 and the longitudinal axis 22. The height axis 18, the longitudinal axis 22, and the lateral axis 24 form a vehicle coordinate system 26 fixed to the vehicle 10. More precisely, in this example, the origin of the vehicle coordinate system is located on the rear axle 28 of the vehicle 10.
[0041] The vehicle 10 also includes a Doppler sensor system 30 installed on the vehicle 10.
[0042] The Doppler sensor system 30 includes multiple Doppler sensors 32. In this example, four radar sensors 34 are used as Doppler sensors 32. The four radar sensors 34 are arranged at each corner of the vehicle 10. Each radar sensor 34 includes a field of view 36, within which it can detect objects 38 outside the vehicle 10, i.e., objects 38 in the environment of the vehicle 10.
[0043] exist Figure 2 Only one object 38 is shown. This object 38 represents multiple objects 38 that may exist in the environment of the vehicle 10.
[0044] The Doppler sensor system 30 also includes a data processing device 40.
[0045] Four Doppler sensors 32 are communicatively connected to a data processing device 40, enabling the data processing device 40 to receive first data D1, indicating the environment of the vehicle 10, detected by the Doppler sensors.
[0046] The data processing device 40 is also communicatively connected to each wheel speed sensor 14 and the gyroscope 16 of the vehicle 10. Therefore, the data processing device 40 is configured to receive third data D3 from the gyroscope 16 indicating the yaw rate of the vehicle 10.
[0047] The vehicle 10 is able to detect objects 38 in its environment via the Doppler sensor system 30. Figure 1 The image shows a vehicle 10 detecting a following vehicle 42 using a Doppler sensor system 30. The vehicle 10 is traveling in the left lane of road 20. The following vehicle 42 is traveling in the right lane of road 20. Therefore, the Doppler sensor 32, mounted at the right rear corner of the vehicle 10, detects the following vehicle 42 within its field of view 36.
[0048] Figure 1Further illustration of the following vehicle 42 is shown in dashed lines. This additional illustration of the following vehicle 42, also shown in dashed lines, is located in the left lane of road 20, just like the vehicle itself. If the first data D1 from the Doppler sensor 32 (here, the radar sensor 34 at the right rear corner of the vehicle 10) that detects the following vehicle 42 includes an angle detection error, then this additional illustration of the following vehicle 42, shown in dashed lines, indicates that the Doppler sensor system 30 of the vehicle 10 can locate the position of the following vehicle 42.
[0049] like Figure 3 As shown, the angle detection error can be at least partially due to the misalignment of the Doppler sensor 32 relative to its desired mounting position and orientation, such as angular misalignment. If the Doppler sensor 32 is in such a state... Figure 3 The solid line in the figure shows the actual orientation on which the vehicle 10 is mounted, which is different from the actual orientation shown in the figure. Figure 3 The desired orientation, indicated by the dashed line, is tilted relative to the actual orientation. An object 38 (see solid circle) located in a lane adjacent to the vehicle's lane can be perceived by the Doppler sensor system 30 as being in a different position; in this example, it is located in the same lane as the vehicle 10 (see dashed circle). The angle detection error can be referred to as the global detection error of the Doppler sensor 32 of the Doppler sensor system 30. This is because the error caused by geometric misalignment of the Doppler sensor 32 applies to the entire field of view 36 of the Doppler sensor 32 of the Doppler sensor system 30.
[0050] In addition, such as Figure 3 As shown, the Doppler sensor 32 of the Doppler sensor system 30 is mounted to the vehicle 10 behind the outer housing 46. The outer housing 46 of the vehicle 10 may include different materials, different shapes, or different wall thicknesses, thereby incorporating different Doppler signal transmission characteristics. This is in Figure 3The corner portion 48 is thickened compared to the longitudinal portion 50 and the transverse portion 52 of the outer housing 46 of the vehicle 10. Furthermore, multiple portions of the outer housing 46 of the vehicle 10 are arranged at different angles relative to the Doppler sensor 32, and signals emitted from the Doppler sensor 32 and signals reflected back from the object 38 to the emitting Doppler sensor 32 need to propagate through these portions. The different properties or orientations of the outer housing 46 of the vehicle 10 through which the Doppler sensor signal needs to propagate may affect the detection result of the Doppler sensor 32. This is due to the different refraction angles of the Doppler sensor signal at different oriented portions of the outer housing 46 of the vehicle 10 and / or at portions of the outer housing 46 of the vehicle 10 that include different signal propagation characteristics. This error can be referred to as the local detection error of the Doppler sensor 32 of the Doppler sensor system 30. This is because the error caused by the different signal propagation characteristics of the vehicle's outer shell 46 within the field of view 36 of the Doppler sensor 32 may vary in the field of view 36 of the Doppler sensor 32, thus causing local distortion in the perception of the environment of the vehicle 10.
[0051] To account for both global and local detection errors, the vehicle 10 also includes a calibration system 54 for the Doppler sensor system 30. The calibration system 54 includes a data processing device 40. In this example, Figure 2 The data processing device 40 shown is part of both the Doppler sensor system 30 installed on the vehicle 10 and the calibration system 54 for the Doppler sensor system 30 installed on the vehicle 10.
[0052] Data processing device 40 includes a data storage unit 56 and a data processing unit 58. Data storage unit 56 includes a computer-readable storage medium 60. A computer program 62 and a trainable data model 64 are provided on the computer-readable storage medium 60. In this example, the trainable data model 64 is an artificial neural network.
[0053] Computer program 62, and therefore computer-readable storage medium 60, includes instructions that, when executed by data processing unit 58 or more generally a computer, cause the computer or data processing unit 58 to perform a method for estimating a vehicle motion vector while the vehicle is in motion. This method may be referred to as a first method. In this example, the first method estimates the vehicle motion vector of vehicle 10 while vehicle 10 is in motion. The vehicle motion vector describes the motion of vehicle 10 within a stationary global coordinate system 66. In other words, the vehicle motion vector describes the position, orientation, and velocity of vehicle 10 within the global coordinate system 66.
[0054] Furthermore, the computer program 62 and therefore the computer-readable storage medium 60 include instructions that, when executed by the data processing unit 58 (or more generally, the computer), cause the computer or data processing unit 58 to perform a method for calibrating a Doppler sensor system 30 that can be mounted to a vehicle. This method may be referred to as a second method. In this example, the second method calibrates the Doppler sensor system 30 of the vehicle 10. The second method forms part of the first method, as will be explained in conjunction with the following three embodiments.
[0055] Therefore, it can be said that the data storage unit 56 and the data processing unit 58 form an apparatus 63 for executing the first method and an apparatus for executing the second method.
[0056] The first embodiment will be described below. Figure 7 The steps of a first embodiment of a first method for estimating the vehicle motion vector of the vehicle 10 as it moves are shown. The first embodiment can be performed if the yaw rate of the vehicle obtained by the gyroscope 16 is at least temporarily determined to be zero, or if the absolute value of the yaw rate is at least temporarily below a yaw rate threshold.
[0057] In the first step S11 of the first embodiment, the first method includes obtaining first data D1 detected by Doppler sensors in the environment of the vehicle 10 indicating motion. In this example, the first data D1 originates from four radar sensors 34 mounted to the Doppler sensor system 30 of the vehicle 10 at each corner of the vehicle 10. The first data D1 includes multiple detections of the Doppler velocity of an object 38 in the field of view 36 of the four radar sensors 34. The Doppler velocity is understood as the velocity of the object 38 detected by the Doppler sensor 32 along the direction of the direct connection between the object 38 and the Doppler sensor 32 that has detected the object 38. The Doppler velocity can also be referred to as the radial velocity of the object 38 relative to the Doppler sensor 32 that has detected the object 38. The Doppler velocity of the object 38 detected by the Doppler sensor 32 and not further processed by the following method steps can be referred to as the measured Doppler velocity v. m .
[0058] In the second step S12 of the first embodiment, the first method includes calibrating the Doppler sensor system 30 mounted on the vehicle 10. Calibration of the Doppler sensor system 30 mounted on the vehicle is achieved by performing a second method. If the yaw rate of the vehicle 10 obtained from the gyroscope 16 is determined to be zero, or if the absolute value of the yaw rate is below a yaw rate threshold, then only the second method of the first embodiment is performed.
[0059] The first step S121 of the second method in the first embodiment includes distinguishing between stationary objects 38 and moving objects 38 within the first data D1. For this purpose, a vehicle motion vector v is defined.p Vehicle motion vector v p It is a motion vector comprising the vehicle's longitudinal velocity, lateral velocity, and yaw rate relative to the global coordinate system 66. A transformation matrix M is determined, which establishes a transformation based on vector V. m All measurements of Doppler velocity v by the deployed Doppler sensor system 30 m Vehicle motion vector v p The relationship between the standard distribution N and random data items (i.e., random samples). The standard distribution has a mean of zero and σ(v). m ) 2 variance: V m =Mv p +N(0, σ(v m ) 2 (1) The transformation matrix M is based on the actual (i.e., true) azimuth angle φ and the actual (i.e., true) elevation angle ϴ detected by each Doppler sensor. The azimuth angle φ is the angle of object 38 relative to the intermediate axis Doppler sensor 32 in the plane including the longitudinal axis 22 and the lateral axis 24 of the vehicle 10. The elevation angle is the angle of object 38 relative to the height axis 18 in the plane perpendicular to the plane including the longitudinal axis 22 and the lateral axis 24 of the vehicle 10.
[0060] Since the actual azimuth angle φ and the actual elevation angle ϴ are unknown, the transformation matrix M for the measurement is determined. m This is based on the measured azimuth angle φ and the measured elevation angle ϴ detected by each Doppler sensor. The actual azimuth angle φ or elevation angle ϴ is the measured azimuth angle or elevation angle ϴ minus the angle detection error of the azimuth angle φ and the elevation angle ϴ, respectively. The angle detection error of the azimuth angle φ and the elevation angle ϴ will be determined within the range of this second method.
[0061] Measurement transformation matrix M m Establish a matrix V e The so-called estimated Doppler velocity v of the deployed Doppler sensor system e Vehicle motion vector v p The relationship between the normal distribution N and the distribution N: V e =M m v p +N(0, σ(v m ) 2 (2) For each point in the field of view 36 of the Doppler sensor 32 of the Doppler sensor system 30, determine the measured Doppler velocity v. m Compared with the estimated Doppler velocity v eThe absolute difference between them. This absolute difference can be referred to as the measured Doppler velocity v. m Compared with the estimated Doppler velocity v e The residual between the two objects is considered as follows: If the residual is lower than a predetermined residual threshold, the corresponding detection in the field of view 36 of the Doppler sensor 32 is considered as a stationary object 38. If the residual is higher than or equal to the predetermined residual threshold, the corresponding detection in the field of view 36 of the Doppler sensor 32 is considered as a non-stationary object, i.e., a moving object 38.
[0062] The estimation of stationary object 38 is even more reliable if the iterative random sample consensus algorithm is used on a randomly selected subset of the first data D1. The iterative random sample consensus algorithm determines the measured Doppler velocity v on a randomly selected subset of the first data D1. m Compared with the estimated Doppler velocity v e The absolute difference between them is repeatedly iterated to iteratively eliminate non-static objects 38.
[0063] To explain this more graphically, Figure 4 A diagram showing example Doppler sensor detections of objects 38 in the environment surrounding the vehicle 10, performed by the Doppler sensor system 30. It can be observed that multiple Doppler sensor detections are located on the dominant horizontal line, which is plotted as a dashed line and created by dominant stationary structures in the environment surrounding the vehicle 10. The intersection of the dashed line with the y-axis of the diagram represents the longitudinal velocity of the vehicle 10.
[0064] Figure 5 The diagram shows the distribution of the radial velocity of the object 38 detected at the azimuth angle φ of a single Doppler sensor 32, i.e., the measured Doppler velocity v. m Estimated Doppler velocity v e The values are plotted as dashed lines. The estimated Doppler velocity reaches its maximum at an azimuth angle of 0°. The estimated Doppler velocity is zero at an azimuth angle of 90°. The frequency bands whose boundaries lie above and below the estimated Doppler velocity and a predetermined residual threshold are located on the flanks of the estimated Doppler velocity ve. All detections within the frequency band are designated as stationary objects 38 (inliers). All detections outside the frequency band are designated as non-stationary objects (outliers).
[0065] Since multiple Doppler sensors 32 (four radar sensors 34 in this example) are used in the Doppler sensor system 30, the vehicle motion vector v of the vehicle 10 can be estimated based on the stationary object 38 detected by the Doppler sensors 32.p Therefore, the second step S122 of the second method according to the first embodiment includes determining the vehicle motion vector v based on the stationary object 38 identified in the first data D1. p Therefore, the optimization problem is formulated to optimize the measurement of Doppler velocities v at all stationary points in the field of view 36 of the Doppler sensor 32 of the Doppler sensor system 30. m With the estimated Doppler velocity v e Minimize the absolute difference between them. In other words, the second step S122 attempts to determine the vehicle motion vector v that is best aligned with the determined first data D1. p The optimization problem is formulated as follows: (3) Where W is a diagonal weighted matrix, and each diagonal element w i,j Let represent the signal-to-noise ratio of the i-th stationary detection (i.e., point) in the field of view 36 of the j-th Doppler sensor 32. In short, the goal of the formulated optimization problem is to find the vehicle motion vector v. p It minimizes the sum of squared residuals weighted by the signal-to-noise ratio detected by each Doppler sensor in the Euclidean norm. This can be achieved through orthogonal distance regression, least squares method, or similar optimization methods targeting v. p Solve the formulaic optimization problem.
[0066] The subsequent third step S123 of the second method according to the first embodiment includes using the vehicle motion vector v determined in step S122. p This is to refine the identification of stationary and non-stationary objects 38 from the first data D1 in step S121. This is done by using the vehicle motion vector v obtained in the previous step S122 in formula (2). p In order to obtain the matrix v e,new Estimated Doppler velocity v in the middle arrangement e,new This is achieved by adding new values.
[0067] For each detection in the field of view 36 of the Doppler sensor 32 of the Doppler sensor system 30, the measured Doppler velocity v is determined. m With the new estimated Doppler velocity v e,new The absolute difference between them. This absolute difference can be referred to as the measured Doppler velocity v. m Compared with the estimated Doppler velocity v e,newThe new residual is determined by the distance between the two objects. If the new residual is lower than a predetermined new residual threshold, the corresponding detection in the field of view 36 of the corresponding Doppler sensor 32 of the Doppler sensor system 30 is considered a stationary object 38. If the residual is higher than or equal to the predetermined new residual threshold, the corresponding detection in the field of view 36 of the corresponding Doppler sensor 32 of the Doppler sensor system 30 is considered a moving object 38. Note that the predetermined new residual threshold is less than (i.e., lower than) the predetermined residual threshold from step S121. Therefore, the accuracy of determining the stationary object 38 from the Doppler sensor detection is increased.
[0068] In the fourth step S124 of the second method according to the first embodiment, the angular error of the azimuth angle detected by each Doppler sensor is determined. Angle error of elevation angle Therefore, the cost function J i Defined as the measured Doppler velocity of each stationary detection i from step S123 With the new estimated Doppler velocity The squared difference between them. In this step S124, non-stationary detections identified in step S123 are not considered. For each Doppler sensor detection i, the cost function J... i Formulated to depend on azimuth angle Error and elevation angle The error in the cost function J. i To be minimized and solved and : (4) Since in the first embodiment, the vehicle 10 travels in a straight line, i.e., the yaw rate obtained from the gyroscope 16 is zero or at least below the yaw rate threshold, the yaw rate and lateral velocity of the vehicle 10 are zero to simplify the minimization problem. Orthogonal distance regression is used to solve the problem. and Minimize the cost function J. and R Each includes global angle error and local angle error.
[0069] Note that the first data D1 detected by the Doppler sensor from one or more time points can be used in the cost function J. i Minimize. If the first data D1 detected by the Doppler sensor at only one point in time is used, then by applying the cost function J... i Add a regularization term to stabilize the cost function J i The minimum solution: (5) and It is separate punishment and The regularization coefficient has a large value. Adding a regularization term helps mitigate the effects of overfitting and ensures that the angle error is correct. and A robust estimate.
[0070] The angle error has already been solved. and Cost function J i After minimization, the weighted average value is obtained through the field of view 36 of a specific Doppler sensor 32. This is used to obtain the global angular error in the azimuth angle φ of a specific Doppler sensor 32. Similarly, this is achieved by weighted averaging over the field of view 36 of the specific Doppler sensor 32. The global angular error of the elevation angle of a specific Doppler sensor is obtained. The local angular errors of the azimuth angle φ and elevation angle ϴ are defined as the residual errors at each point in the field of view 36 of the Doppler sensor 32, which cannot be explained by the global angular errors of the azimuth angle φ and elevation angle ϴ, respectively. The weight of the weighted average is defined based on the signal-to-noise ratio of each detection in the field of view 36 of the specific Doppler sensor 32.
[0071] In the fifth step S125 of the second method according to the first embodiment, an iterative random search algorithm is applied to achieve the highest precision v. p,best The vehicle motion vector. A random search algorithm estimates the Doppler velocity V. e With the measured Doppler velocity V m The difference between them is minimized. This is achieved by estimating the global angle error and the local angle error in each step of the iteration. The algorithm is initialized with the azimuth angle φ and elevation angle ϴ of the first data D1 (i.e., detected by the Doppler sensor). The estimates of the global angle error and the local angle error are subtracted from the results of the azimuth angle φ and the elevation angle ϴ from the previous iteration step. Due to the vehicle motion vector v p Depending on the true azimuth angle φ and the true elevation angle ϴ detected by the Doppler sensor, the vehicle motion vector v increases as the random search algorithm iterates through its iterations. p The estimates gradually became more accurate.
[0072] Note that stochastic search algorithms can include iterative gradient descent optimization, such as stochastic gradient descent with an adaptive learning rate, to optimize the estimated Doppler velocity V. e Iteratively approximated as the measured Doppler velocity V m .
[0073] The second method is paused once the vehicle 10 stops traveling in a straight line, i.e., once the yaw rate of the vehicle 10 provided by the gyroscope 16 is not zero, or if the absolute value of the yaw rate is no longer below the yaw rate threshold. In short, the second method is paused if the vehicle 10 is traveling on a curve. The latest result of the second method for global and local angular errors in the azimuth angle φ and elevation angle is considered as a calibration factor for the first data D1, indicating the Doppler sensor detection when the vehicle 10 is driving with a yaw rate that is not equal to zero or with an absolute value of yaw rate above the yaw rate threshold. Therefore, when the second method is paused, second data D2, indicating the angular detection errors of the azimuth angle φ and elevation angle ϴ, is provided as a result of the second method. Additionally or alternatively, when the second method is paused, information indicating the vehicle motion vector v is provided. p,best The second data D2 is used as the result of the second method.
[0074] In the third step S13 of the first method according to the first embodiment, the corrected first data D1 is used to determine the model of the stationary and moving objects 38 within the environment of the vehicle 10. For this purpose, formulas (1) and (2) from the first step of the second method are applied. However, the measured Doppler velocity V... m The first data D1 obtained when the vehicle is traveling with a yaw rate that is not equal to zero or with an absolute value of yaw rate that is higher than the yaw rate threshold is corrected by calibration factors of global angle error and local angle error in azimuth angle φ and elevation angle.
[0075] In the fourth step S14 of the first method according to the first embodiment, the vehicle motion vector v of the vehicle 10 when driving with a yaw rate that is not equal to zero or with an absolute value of a yaw rate that is higher than a yaw rate threshold is obtained from the model of stationary and moving objects 38 in the environment of the vehicle 10. p Therefore, formula (3) from the second method is used.
[0076] Therefore, in the first embodiment, in short, the Doppler sensor system 30 is calibrated by the second method as long as the vehicle 10 is traveling in a substantially straight line. The latest result of the second method for global and local angular errors in the azimuth angle φ and elevation angle is regarded as a calibration factor, i.e., a correction factor, used to indicate the first data D1 of the Doppler sensor detection when the vehicle is traveling on a curve. Thus, the vehicle motion vector v of the vehicle 10 is obtained with high accuracy without relying on a global navigation satellite system. p .
[0077] Note that during the use of the vehicle 10, the second method, namely the calibration of the Doppler sensor system 30, can be performed periodically, for example, after a predefined time window. After each execution of the second method, the most recently obtained correction factor is used in the first method.
[0078] Figure 6 A comparison is shown between the speed of vehicle 10 determined by this first embodiment I and the speed of vehicle 10 determined by Global Navigation Satellite System II and by the speed of vehicle 10's internal measurement unit III during the same driving sequence. The internal measurement unit includes a gyroscope 16 and a wheel speed sensor 14. It can be observed that the speed of vehicle 10 obtained by the first embodiment closely matches the vehicle speed obtained by the Global Navigation Satellite System, which is considered to be the true speed of vehicle 10. In contrast, the vehicle speed obtained from the internal measurement unit of vehicle 10 does not match the vehicle speed obtained by the Global Navigation Satellite System. Therefore, the accuracy of the first embodiment exceeds the accuracy of the internal measurement unit of vehicle 10.
[0079] The second embodiment will be described below. Figure 8 The steps of a second embodiment of a first method for estimating the vehicle motion vector of the vehicle 10 during vehicle motion are shown. Only the differences relative to the first embodiment will be explained. The second embodiment can be performed at any yaw rate of the vehicle 10 obtained from the gyroscope 16.
[0080] The first step S21 of the first method according to the second embodiment is the same as the first step S11 of the first method according to the first embodiment.
[0081] The second step S22 of the first method according to the second embodiment includes performing the second method according to the second embodiment.
[0082] Except for allowing any yaw rate of the vehicle, the first step S221 of the second method according to the second embodiment is the same as the first step S121 of the second method according to the first embodiment. The yaw rate is received from the gyroscope 16.
[0083] The second step S222 of the second method according to the second embodiment is the same as the second step S122 of the second method according to the first embodiment, except that in the optimization problem in formula (3), the vehicle motion vector v based on the yaw rate obtained by the gyroscope 16 is used. p Used as an initial guess. Regressed by orthogonal distance or by least squares for v. p Solve the formulaic optimization problem (3).
[0084] Similar to the third step S123 of the second method according to the first embodiment, the third step S223 of the second method according to the second embodiment also includes using the vehicle motion vector v determined in the previous step (here in step S222). pThis is to refine the identification of stationary and non-stationary objects 38 from the first data D1 in step S221. Step S223 of the second method according to the second embodiment is similar to step S123 of the second method according to the first embodiment. However, additionally, based on the measured Doppler velocity V m Compared with the estimated Doppler velocity V e The deviation includes the measured Doppler velocity V. m Each Doppler sensor detects and determines the confidence level.
[0085] In the fourth step S224 of the second method according to the second embodiment, the yaw rate of the vehicle 10 obtained from the gyroscope 16 and the stationary object 38 determined in step S223 are used to estimate the motion direction of each Doppler sensor 32 mounted on the vehicle 10. Non-stationary objects 38 from step S223 are excluded in step S224. For this purpose, the vehicle 10 with the Doppler sensors 32 mounted is understood as a rigid body. Therefore, rigid body kinematics is applied to represent the motion of each Doppler sensor 32 mounted on the vehicle in the global coordinate system 66. For this purpose, it is assumed that the vehicle 10 has no lateral movement at its rear axle. In other words, it is assumed that the vehicle coordinate system 26 does not move along a direction parallel to the rear axle of the vehicle 10. Therefore, based on the yaw rate of the vehicle 10 obtained from the gyroscope 16, a velocity vector in the global coordinate system is determined for each Doppler sensor 32 mounted on the vehicle 10.
[0086] By interpolating the velocity vector calculated for each Doppler sensor 32 mounted on the vehicle 10 with matrix V e、new The estimated Doppler velocity v is arranged and obtained in step S223 in a manner similar to step S123 of the first embodiment. e、new The new values are compared to estimate the global angular error of the azimuth angle φ for each Doppler sensor 32.
[0087] In the fifth step S225 of the second method according to the second embodiment, the first data D1 detected by the Doppler sensors indicating the environment of the vehicle from step S21 is corrected by estimating the global angle error of the azimuth angle φ of each Doppler sensor 32. The corrected first data D1 is then used to provide the vehicle motion vector v using orthogonal distance regression or a similar method. p A more accurate estimate.
[0088] The local angle errors of azimuth φ and elevation ϴ are determined based on first data D1, which is a global angle error correction determined by the azimuth φ of each Doppler sensor 32. The determination of the local angle errors is achieved using a random search algorithm similar to step S125 of the second method according to the first embodiment.
[0089] Finally, the global angle error of the azimuth angle φ is re-estimated, and the global angle error of the elevation angle ϴ is estimated for the first time. This is achieved by weighting the average values over the field of view 36 of the specific Doppler sensor 32. This is used to obtain the global angular error of the azimuth angle φ of a specific Doppler sensor 32. Similarly, this is achieved by weighted averaging over the field of view 36 of the specific Doppler sensor 32. The global angular error of the elevation angle of a specific Doppler sensor 32 is obtained. The local angular errors of the azimuth angle φ and the elevation angle ϴ are determined as the residual errors at each point in the field of view 36 of the Doppler sensor 32, which cannot be explained by the global angular errors of the azimuth angle φ and the elevation angle ϴ, respectively.
[0090] The results of the second method for global and local angle errors in azimuth φ and elevation ϴ are considered as calibration or correction factors for the first data D1 indicating the Doppler sensor detection. Therefore, second data D2 indicating the angle detection errors of azimuth φ and elevation ϴ is provided as the result of the second method according to the second embodiment. Additionally or alternatively, an indication of the vehicle motion vector v is provided. p,best The second data D2 is the result of the second method according to the second embodiment.
[0091] In the third step S23 of the first method according to the second embodiment, the corrected first data D1 is used to determine the models of stationary and moving objects 38 within the environment of the vehicle 10. This step is the same as step S13 of the first method according to the first embodiment, that is, using formulas (1) and (2). The Doppler velocity V to be used in these formulas is... m The first data D1 is corrected by calibration factors for global and local angle errors in the azimuth angle φ and elevation angle ϴ.
[0092] In the fourth step S24 of the first method according to the second embodiment, the vehicle motion vector v of the vehicle 10 is obtained from the model of the stationary and moving objects 38 in the environment of the vehicle 10. p Therefore, formula (3) from the first method is used. Thus, the fourth step S24 of the first method according to the second embodiment is the same as the fourth step S14 of the first method according to the first embodiment.
[0093] The third embodiment will be described below. Figure 9 The steps of a third embodiment of a first method for estimating the vehicle motion vector of the vehicle 10 as it moves are shown. Only the differences relative to the first and second embodiments will be explained. Similar to the second embodiment, the third embodiment can be performed at any yaw rate of the vehicle 10 obtained by the gyroscope 16.
[0094] In the first step S31 of the first method according to the third embodiment, a trainable data model 64 is trained using first data D1 detected by a Doppler sensor indicating the environment of the vehicle 10. When the vehicle 10 is driving a training driving sequence, the Doppler sensor detection originates from four radar sensors 34 mounted on the vehicle 10. During the training driving sequence, odometer data of the vehicle 10 is obtained from wheel speed sensors 14, gyroscopes 16, and / or from a global navigation satellite system. Therefore, the vehicle motion vector of the vehicle 10 during the training driving sequence is determined based on the rotational speed of the wheels 12 of the vehicle 10, the yaw rate of the vehicle, and / or the global position signal from the global navigation satellite system.
[0095] The vehicle motion vectors obtained in this manner are used as ground conditions for training the trainable data model 64. The data model 64 is configured to receive first data D1 as input. The data model 64 is also configured to provide the vehicle motion vector, along with global and local angular errors for azimuth φ and elevation ϴ, as outputs for each Doppler sensor 32. The data model 64 is trained by minimizing the difference between the trainable data model 64's prediction of the vehicle motion vector and the vehicle motion vector determined from wheel speed sensor 14, gyroscope 16, and / or from a global navigation satellite system. The first data D1, obtained during training driving sequences, is used as input to predict the trainable data model 64.
[0096] Once the trainable data model 64 has been trained, in the second step S32 of the first method according to the third embodiment, first data D1 indicating the environment of the vehicle 10 detected by the Doppler sensor is obtained during the test driving sequence of the vehicle. This step is the same as the first step S11 of the first method according to the first embodiment and the first step S21 of the first method according to the second embodiment.
[0097] The third step S33 of the first method according to the third embodiment includes performing the second method according to the third embodiment.
[0098] The second method according to the third embodiment includes providing first data D1 obtained during a test driving sequence as input data to a trained data model 64. The trained data model 64 provides the vehicle motion vector and angle detection errors in azimuth φ and elevation ϴ as outputs during the test driving sequence, which may be referred to as second data D2.
[0099] In the fourth step S34 of the first method according to the third embodiment, a model of the stationary and moving object 38 based on the first data D1 obtained during the test driving cycle is obtained based on the vehicle motion vector and the angle detection error as a correction factor. This step is the same as the third step S23 of the first method according to the second embodiment.
[0100] In the fifth step S35 of the first method according to the third embodiment, the vehicle motion vector v of the vehicle 10 is obtained from the model of the stationary and moving objects 38 in the environment of the vehicle 10. p Therefore, the fifth step S35 of the first method according to the third embodiment is the same as the fourth step S14 of the first method according to the first embodiment or the fourth step S24 of the first method according to the second embodiment.
[0101] As used herein, the phrase “at least one” in relation to a list of one or more entities should be understood to mean at least one entity selected from any one or more entities in the entity list, but not necessarily at least one of each entity specifically listed in the entity list, and does not exclude any combination of entities in the entity list. This definition also allows for the optional presence of entities other than those specifically identified in the entity list referred to by the phrase “at least one,” whether related to or unrelated to those specifically identified entities. Thus, as a non-limiting example, “at least one of A and B” (or equivalently, “at least one of A or B”, or equivalently, “at least one of A and / or B”) could in one example mean at least one (optionally including more than one) A, without B (and optionally including entities other than B); in another example, at least one (optionally including more than one) B, without A (and optionally including entities other than A); and in yet another example, at least one (optionally including more than one) A and at least one (optionally including more than one) B (and optionally including other entities). In other words, the phrases “at least one,” “one or more,” and “and / or” are open-ended expressions that are both connected and separate in operation. For example, each of the expressions “at least one of A, B, and C,” “at least one of A, B, or C,” “one or more of A, B, and C,” “one or more of A, B, or C,” and “A, B, and / or C” can mean a single A, a single B, a single C, A and B together, A and C together, B and C together, A, B, and C together, and optionally, any of the above combined with at least one other entity.
[0102] By studying the accompanying drawings, the disclosure, and the appended claims, those skilled in the art can understand and implement other variations of the disclosed examples in practice with respect to the claimed disclosure. In the claims, the word "comprising" does not exclude other elements or steps, and the indefinite articles "a" or "an" do not exclude multiple. A single processor or other unit can perform the function of several items or steps recited in the claims. The fact that certain measures are recited in mutually different dependent claims does not mean that a combination of these measures cannot be used advantageously. Computer programs can be stored / distributed on suitable media, such as optical storage media or solid-state media provided with or as part of other hardware, but can also be distributed in other forms, such as via the Internet or other wired or wireless telecommunications systems. Any reference numerals in the claims should not be construed as limiting the scope of the claims.
[0103] List of reference numerals 10 bicycles 12 wheels 14 Wheel speed sensors 16 Gyroscope 18 Height axis 20 Roads 22 Longitudinal axis 24. Horizontal axis 26. Vehicle coordinate system 28 Rear Axle 30 Doppler sensor system 32 Doppler sensors 34 Radar Sensors 36 Field of view 38 objects 40 Data processing equipment 42 Following vehicles 46. Outer casing 48. Corner section 50 Vertical section 52. Horizontal section 54 Calibration System 56 data storage units 58 Data Processing Units 60 Computer-readable storage media 62 Computer Programs 63. Apparatus for performing the first method and the second method 64 Data Model 66 Global Coordinate System D1 First Data D2 Second Data D3 Third Data
Claims
1. A method for calibrating a Doppler sensor system (30) that can be mounted to a vehicle (10), the method comprising: - Obtain first data (D1) indicating the environment of the vehicle (10) from multiple Doppler sensors (32) that can be mounted on the vehicle (10), wherein the Doppler sensor detection includes angle information. - Estimate the vehicle motion vector based on the first data (D1), and determine the angle detection error based on the first data (D1) and the estimated vehicle motion vector, or Using the trained data model (64), the angle detection error and / or vehicle motion vector are determined based on the first data (D1), and - Provide second data (D2) indicating the angle detection error and / or the vehicle motion vector.
2. The method according to claim 1, wherein, The first data (D1) is associated with the vehicle (10) traveling on a straight path and / or with the yaw rate of the vehicle below a predetermined yaw rate threshold.
3. The method according to claim 1, wherein, The vehicle motion vector is estimated based on additional third data (D3) indicating the yaw rate of the vehicle (10).
4. The method according to any one of the preceding claims, wherein, Estimating the vehicle motion vector includes solving an optimization problem or determining the median of the Doppler velocity of the first data (D1) corrected by the corresponding angle information of the first data (D1), and / or Determining the angle error involves solving an optimization problem.
5. The method according to any one of the preceding claims, wherein, The determined angle detection error includes local angle detection error and / or global angle detection error. Wherein, the local angle detection error refers to the angle detection error at each point in the field of view (36) of the Doppler sensor (32) specific to the Doppler sensor system (30), and The global angle detection error refers to the angle detection error of the entire field of view (36) of the Doppler sensor (32) of the Doppler sensor system (30) systematically applied.
6. The method according to any one of the preceding claims further includes using a random search algorithm and / or gradient descent optimization to enhance the accuracy of the estimated vehicle motion vector.
7. The method according to any one of the preceding claims, wherein, The trained data model (64) includes an artificial neural network.
8. The method according to any one of claims 1 to 6, wherein, Estimating the vehicle motion vector includes determining a first model of stationary and moving objects (38) based on the first data (D1), and estimating an initial vehicle motion vector based on the first model.
9. The method according to claim 8, wherein, Estimating the vehicle motion vector also includes determining a second model of stationary and moving objects based on the first data (D1) and the initial vehicle motion vector.
10. A method for estimating the vehicle motion vector of a vehicle (10) while the vehicle (10) is in motion, the method comprising: - Obtain first data (D1) indicating the environment of the moving vehicle (10) from multiple Doppler sensors (32) installed on the vehicle (10). - The Doppler sensor system (30) mounted on the vehicle (10) is calibrated using the method according to any one of the preceding claims, or wherein the Doppler sensor system (30) is provided in a state that has been calibrated using the method according to any one of the preceding claims. - Based on the first data (D1), determine the models of the stationary object and the moving object (38), and - Estimate the vehicle motion vector based on the model.
11. A data processing apparatus (40) comprising means (63) for performing the method according to any one of claims 1 to 9 and / or for performing the method according to claim 10.
12. A computer program (62) comprising instructions which, when executed by a computer, cause the computer to perform the method according to claims 1 to 9 and / or the method according to claim 10.
13. A computer-readable storage medium (60) including instructions that, when executed by a computer, cause the computer to perform the method according to claims 1 to 9 and / or the method according to claim 10.
14. A Doppler sensor system (30) that can be installed in a vehicle (10), the system (30) comprising: - Multiple Doppler sensors (32) that can be installed on the vehicle (10), and - The data processing apparatus (40) according to claim 11. The plurality of Doppler sensors (32) are communicatively connected to the data processing device (40).
15. A calibration system (54) for a Doppler sensor system (30), the calibration system (54) comprising the data processing apparatus (40) according to claim 11.
