Real-time reliability evaluation method for enhancing vehicle speed estimation robustness based on data fusion

Abstract

A method of estimating a lateral velocity of a vehicle includes receiving sensor data from a vehicle sensor; determining a physics-based longitudinal velocity estimate of the vehicle using the physics-based model and the sensor data; determining a data-driven longitudinal velocity estimate of the vehicle using the first neural network and the sensor data; determining which of the physical-based longitudinal velocity estimate and the data-driven longitudinal velocity estimate is more reliable using a second neural network to determine a selected longitudinal velocity estimate; determining a lateral velocity of the vehicle using the selected longitudinal velocity estimate; and controlling the vehicle based on the lateral speed.

Description

Technical Field

[0001] The present disclosure relates to a real-time reliability assessment system and method to enhance the robustness of vehicle speed estimation based on data fusion. Background Art

[0002] This introduction generally presents the background of the present disclosure. The work of the inventors described in this introduction, as well as aspects described that may not meet the conditions of the prior art at the time of filing, are not regarded as the prior art of the present disclosure either explicitly or implicitly.

[0003] In a vehicle, the lateral speed can be estimated using the longitudinal speed. However, since the longitudinal speed estimation is unreliable, the lateral speed estimation of the vehicle is sometimes also unreliable. For this reason, a method and system are needed to reliably determine the lateral speed of the vehicle. Summary of the Invention

[0004] The present disclosure describes a method for estimating the lateral speed of a vehicle. In one aspect of the present disclosure, the method includes receiving sensor data from vehicle sensors; determining a physics-based longitudinal speed estimate of the vehicle using a physics-based model and the sensor data; determining a data-driven longitudinal speed estimate of the vehicle using a first neural network and the sensor data; using a second neural network to determine which of the physics-based longitudinal speed estimate and the data-driven longitudinal speed estimate is more reliable to determine a selected longitudinal speed estimate; determining the lateral speed of the vehicle in real time using the selected longitudinal speed estimate; and controlling the vehicle based on the lateral speed. The method described in this paragraph improves vehicle technology by providing a more accurate lateral vehicle speed estimate, thereby improving the control of the vehicle.

[0005] In some aspects of the present disclosure, the sensor data includes the wheel speed of the vehicle, the longitudinal acceleration of the vehicle, the lateral acceleration of the vehicle, the yaw rate of the vehicle, the road wheel angle of the vehicle, and the wheel torque of the vehicle. The first neural network can be a recurrent neural network or another suitable neural network. To determine which of the physics-based longitudinal speed estimate and the data-driven longitudinal speed estimate is more reliable, the method can include determining the reliability of the physics-based longitudinal speed estimate using a second neural network, and determining the reliability of the data-driven longitudinal speed estimate using the second neural network. The method can also include comparing the reliability of the physics-based longitudinal speed estimate with the reliability of the data-driven longitudinal speed estimate to determine which of the physics-based longitudinal speed estimate and the data-driven longitudinal speed estimate is more reliable. The method can include determining a final longitudinal speed using a first extended Kalman filter based on the selected longitudinal speed estimate. The method can include: comparing the reliability of the physics-based longitudinal speed estimate with a predetermined reliability threshold to determine whether the reliability of the physics-based longitudinal speed estimate is less than the predetermined reliability threshold; comparing the reliability of the data-driven longitudinal speed estimate with the predetermined reliability threshold to determine whether the reliability of the data-driven longitudinal speed estimate is less than the predetermined reliability threshold; and in response to determining that both the reliability of the data-driven longitudinal speed estimate and the reliability of the physics-based longitudinal speed estimate are less than the predetermined reliability threshold, increasing the covariance of the first extended Kalman filter. The method can include determining the lateral speed of the vehicle using a second extended Kalman filter based on the final longitudinal speed previously determined using the first extended Kalman filter.

[0006] The present disclosure also describes a real-time reliability assessment system to enhance the robustness of vehicle speed estimation based on data fusion. The system includes a plurality of sensors and a controller in communication with the sensors. The controller is programmed to execute the above method.

[0007] The present disclosure also describes a tangible, non-transitory machine-readable medium that includes machine-readable instructions that, when executed by a processor, cause the processor to execute the above method.

[0008] Other applicable fields of the present invention will become apparent from the detailed description provided below. It should be understood that the specification and specific examples are for illustrative purposes only and are not intended to limit the scope of the present disclosure.

[0009] The above features and advantages of the currently disclosed systems and methods, as well as other features and advantages, are apparent when considered in conjunction with the accompanying drawings, which include the claims and exemplary embodiments. Description of the Drawings

[0010] The present disclosure will be more fully understood from the detailed description and the accompanying drawings, in which:

[0011] Figure 1 It is a schematic diagram of a vehicle including a real - time reliability evaluation system that enhances the robustness of vehicle speed estimation based on data fusion.

[0012] Figure 2 It is a real - time reliability evaluation method that enhances the robustness of vehicle speed estimation based on data fusion.

[0013] Figure 3 It is a longitudinal speed arbitration and covariance online adaptive method. Detailed implementation manners

[0014] Now, several embodiments of the present disclosure illustrated in the accompanying drawings will be referred to in detail. Whenever possible, the same or similar reference numerals are used in the drawings and the description to denote the same or similar parts or steps.

[0015] Refer to Figure 1 , the vehicle 10 generally includes a body 12 and a plurality of wheels 14 connected to the body 12. The vehicle 10 can be an autonomous vehicle. In the described embodiment, the vehicle 10 is described as a sedan in the illustrated embodiment, but it should be understood that other vehicles can also be used, including trucks, scooters, sport utility vehicles (SUVs), recreational vehicles (RVs), etc.

[0016] The vehicle 10 also includes one or more sensors 24 connected to the body 12. The sensors 24 sense observable conditions of the external environment and / or the internal environment of the vehicle 10. As a non - limiting example, the sensors 24 can include one or more cameras, one or more light detection and ranging (LIDAR) sensors, one or more radars, one or more global positioning system (GPS) transceivers, one or more inertial measurement units (IMU), one or more accelerometers, one or more vehicle speed sensors, one or more wheel speed sensors, one or more yaw rate sensors, one or more gyroscopes, one or more proximity sensors, one or more ultrasonic sensors, one or more thermal imaging sensors, and / or other sensors 24. Each sensor 24 is configured to generate a signal indicating the sensed observable conditions (i.e., sensor data) of the external environment and / or the internal environment of the vehicle 10.

[0017] Vehicle 10 includes a vehicle controller 34 that communicates with a sensor 24. The vehicle controller 34 includes at least one vehicle processor 44 and a vehicle non - transitory computer - readable storage device or medium 46. The vehicle processor 44 can be a custom or commercially available processor, a central processing unit (CPU), a graphics processing unit (GPU), an auxiliary processor among several processors associated with the vehicle controller 34, a semiconductor - based microprocessor (in the form of a microchip or chipset), a macroprocessor, a combination thereof, or generally a device for executing instructions. The vehicle non - transitory computer - readable storage device or medium 46 can include volatile and non - volatile memory such as, for example, read - only memory (ROM), random access memory (RAM), and keep - alive memory (KAM). KAM is a persistent or non - volatile memory that can be used to store various operating variables when the vehicle processor 44 is powered down. The vehicle non - transitory computer - readable storage device or medium 46 can be implemented using a variety of storage devices such as PROM (programmable read - only memory), EPROM (electrically programmable read - only memory), EEPROM (electrically erasable programmable read - only memory), flash memory, or any other electrical, magnetic, optical, or combination storage device capable of storing data, some of which represents executable instructions used by the vehicle controller 34 in controlling the vehicle 10. The vehicle controller 34 of the vehicle 10 can be programmed to execute method 100 ( Figure 2 ) and method 200 ( Figure 3 ).

[0018] The instructions can include one or more separate programs, each program including an ordered list of executable instructions for implementing a logical function. When executed by the vehicle processor 44, the instructions receive and process signals from sensors, execute logic, calculations, methods, and / or algorithms for automatically controlling components of the vehicle 10, and generate control signals based on the logic, calculations, methods, and / or algorithms to automatically control components of the vehicle 10. Although Figure 1 shows a single vehicle controller 34, embodiments of the vehicle 10 can include multiple vehicle controllers 34 that communicate via a suitable communication medium or a combination thereof and cooperate to process sensor signals, execute logic, calculations, methods, and / or algorithms, and generate control signals to automatically control features of the vehicle 10. The vehicle controller 34 is part of a system 21 for estimating the lateral speed of the vehicle 10.

[0019] Vehicle 10 includes one or more actuators 26 that control one or more vehicle features such as, but not limited to, the propulsion system, the driveline system, the steering system, and the braking system of the vehicle 10. In various embodiments, the vehicle features can also include internal and / or external vehicle features such as, but not limited to, doors, trunks, and cab features such as air, music, lighting, etc.

[0020] Figure 2 is a flowchart of method 100. Method 100 begins at block 102. At block 102, vehicle controller 34 receives sensor data from sensors 24 of vehicle 10 and / or an estimate based on the sensor data. The sensor data can include, but is not limited to, the wheel speed of vehicle 10, the yaw acceleration of vehicle 10, the longitudinal acceleration of vehicle 10, the lateral acceleration of vehicle 10, the yaw rate of vehicle 10, the road wheel angle of vehicle 10, and / or the wheel torque of vehicle 10. Then, method 100 proceeds to block 104.

[0021] At block 104, vehicle controller 34 determines a physics-based longitudinal speed estimate of vehicle 10 using a physics-based model and sensor data from sensors 24. To determine the physics-based longitudinal speed estimate, vehicle controller 34 can use a physics-based model that is described using the following formula:

[0022]

[0023]

[0024] where:

[0025] a x is the longitudinal acceleration of vehicle 10;

[0026] a y is the lateral acceleration of vehicle 10;

[0027] is the yaw acceleration of vehicle 10;

[0028] ω z is the yaw rate of vehicle 10;

[0029] v x is the longitudinal speed of vehicle 10;

[0030] v y is the lateral speed of vehicle 10;

[0031] ψ is the yaw angle of vehicle 10;

[0032] x A is the position of vehicle 10 relative to the horizontal axis;

[0033] y A is the position of vehicle 10 relative to the vertical axis;

[0034] is the GPS speed of vehicle 10 in the eastward direction;

[0035] is the GPS speed of vehicle 10 in the northward direction;

[0036] γ GPS is the GPS yaw angle of vehicle 10;

[0037] is the yaw rate of vehicle 10 determined by GPS or IMU;

[0038] is the physics-based longitudinal speed estimate of vehicle 10.

[0039] In the physics-based model, longitudinal acceleration, lateral acceleration, and yaw acceleration can be used as inputs. Further, it is assumed that the longitudinal acceleration and lateral acceleration of vehicle 10 are gravity-compensated. Additionally, when the vehicle sideslip is low (i.e., close to zero), the heading angle and yaw angle are similar. After determining the physics-based longitudinal speed estimate of vehicle 10, method 100 continues to block 106.

[0040] At block 106, vehicle controller 34 determines a data-driven longitudinal speed estimate of vehicle 10 using a first neural network and sensor data. Since the data-driven longitudinal speed estimate may only be used when the physics-based longitudinal speed estimate is inaccurate, the first neural network may be smaller and computationally efficient. At block 106, vehicle controller 34 receives certain inputs. As a non-limiting example, the inputs can include the wheel speed, longitudinal acceleration, lateral acceleration, yaw rate, yaw acceleration, road wheel angle, and wheel torque of vehicle 10. These inputs are in the input space. Then, vehicle controller 34 uses a dimensionality reduction technique to transform the input data set into a data set with significantly reduced dimensions. As a non-limiting example, vehicle controller 34 can use principal component analysis (PCA) to perform dimensionality reduction on the input data set. The output of principal component analysis is the principal component. The principal component is used as the input to the first neural network. Then, the first neural network outputs a data-driven longitudinal speed estimate of vehicle 10. As a non-limiting example, the first neural network can be a recurrent neural network (RNN) or a non-linear autoregressive network with exogenous input (NARX). The first neural network can be trained by first testing multiple input settings based on network accuracy. Then, the accuracy and interference time of different networks are compared. Next, based on the selected network, the hyperparameters are adjusted to have the most accurate and least interference time. Then, method 100 continues to block 108.

[0041] At block 108, the vehicle controller 34 performs a reliability assessment on the data-driven longitudinal speed estimate and the physics-based longitudinal speed estimate to determine which of the physics-based longitudinal speed estimate and the data-driven longitudinal speed estimate is more reliable. A second neural network can be used to predict the reliability of the data-driven longitudinal speed estimate and the physics-based longitudinal speed estimate. To train the second neural network, the ground truth speed of the vehicle 10 is measured using sensor 24 (e.g., a real-time kinematic sensor). Then, the error of the estimated speed (i.e., the data-driven longitudinal speed estimate and the physics-based longitudinal speed estimate) is calculated using the ground truth speed. The error can be determined using a weighted error function as follows:

[0042] E W =(w)E A +(1 - w)E r

[0043] E w is the weighted error;

[0044] w is the sigmoid weighting function;

[0045] E A is the absolute error between the ground truth speed and the estimated speed;

[0046] E r is the relative error between the ground truth speed and the estimated speed.

[0047] The weighted error for all data points is calculated and used as the target variable for speed reliability assessment. After selecting the possible input features, dimensionality reduction techniques such as PCA are used to reduce the complexity of the input space and the neural network model. Multiple neural networks are trained using the new input feature set and the weighted error as the target variable. Then, the regressor with the best performance (i.e., the network) is used to predict the error of the speed estimate. A high error indicates low reliability, while a low error indicates high reliability. Then, method 100 proceeds to block 110.

[0048] At block 110, the vehicle controller 34 arbitrates between the data-driven longitudinal speed estimate and the physics-based longitudinal speed estimate. Specifically, the vehicle controller 34 selects the most reliable speed estimate between the data-driven longitudinal speed estimate and the physics-based longitudinal speed estimate. The most reliable speed estimate can be referred to as the selected longitudinal speed estimate. Then, method 100 proceeds to block 112.

[0049] At block 112, the vehicle controller 34 determines whether to update (e.g., increase) the covariance of the first extended Kalman filter online based on the reliability of the data-driven longitudinal speed estimate and the physics-based longitudinal speed estimate, as described in detail below. Regardless of whether the covariance is fixed or updated, the selected longitudinal speed estimate is fed into the first extended Kalman filter to determine the final longitudinal speed. Then, method 100 proceeds to block 114.

[0050] At block 114, the vehicle controller 34 uses a second extended Kalman filter to determine the lateral speed of the vehicle 10 based on the final longitudinal speed previously determined using the first extended Kalman filter. Then, method 100 proceeds to block 116.

[0051] At block 116, the vehicle controller 34 controls the vehicle 10 using the previously determined lateral speed. As a non-limiting example, the vehicle controller 34 can adjust the torque of one or more actuators 26 (e.g., electric motor, internal combustion engine) and / or actuate other actuators 26 (e.g., electronic limited slip differential (eLSD) and aerodynamic elements) based on the lateral speed previously determined at block 114.

[0052] Figure 3 It is a flowchart of the longitudinal speed arbitration and covariance online adaptation method 200. Method 200 starts at block 202 and block 204. At block 202, the vehicle controller 34 determines the physics-based longitudinal speed estimate as described above. At block 204, the vehicle controller 34 determines the data-driven longitudinal speed estimate as described above. Then, method 100 continues to block 206. At block 206, the vehicle controller 34 determines the reliability assessment of the physics-based longitudinal speed estimate and the data-driven longitudinal speed estimate as described above. Next, method 100 proceeds to block 208 and block 210.

[0053] At block 208, the vehicle controller 34 compares the reliability of the physics-based longitudinal speed estimate with the reliability of the data-driven longitudinal speed estimate to determine which of the physics-based longitudinal speed estimate and the data-driven longitudinal speed estimate is more reliable. If the reliability of the physics-based longitudinal speed estimate is greater than the reliability of the data-driven longitudinal speed estimate, then method 200 proceeds to block 212. At block 212, the vehicle controller 34 selects the physics-based longitudinal speed estimate as the longitudinal speed for determining the lateral speed of the vehicle 10. If the reliability of the data-driven longitudinal speed estimate is greater than the reliability of the physics-based longitudinal speed estimate, then method 200 continues to block 214. At block 214, the vehicle controller 34 selects the data-driven longitudinal speed estimate as the longitudinal speed for determining the lateral speed of the vehicle 10.

[0054] At block 210, the vehicle controller 34 performs a critical reliability assessment. The critical reliability assessment requires comparing each of the physics-based longitudinal speed estimate and the data-driven longitudinal speed estimate with a predetermined reliability threshold. Next, method 200 proceeds to block 216. At block 216, the vehicle controller 34 determines whether both the physics-based longitudinal speed estimate and the data-driven longitudinal speed estimate are unreliable. Specifically, if the reliability of the data-driven longitudinal speed estimate and the reliability of the physics-based longitudinal speed estimate are both less than the predetermined reliability threshold, method 200 proceeds to block 218. At block 218, the vehicle controller 34 increases the covariance of the first extended Kalman filter in real time. If the reliability of the data-driven longitudinal speed estimate and / or the reliability of the physics-based longitudinal speed estimate is equal to or greater than the predetermined reliability threshold, method 200 proceeds to block 220. At block 220, the vehicle controller 34 fixes the covariance of the first extended Kalman filter. In other words, the covariance of the first extended Kalman filter remains unchanged.

[0055] Although the above describes exemplary embodiments, it does not mean that these embodiments describe all possible forms encompassed by the claims. The words used in the specification are descriptive rather than restrictive, and it should be understood that various changes can be made without departing from the spirit and scope of the present disclosure. As previously mentioned, the features of the various embodiments can be combined to form further embodiments of the presently disclosed systems and methods that may not be explicitly described or illustrated. Although the various embodiments may be described as having advantages over one or more desired features or being superior to other embodiments or prior art embodiments, those of ordinary skill in the art recognize that trade-offs can be made to one or more features or characteristics to achieve the desired overall system attributes, depending on the specific application and implementation. These attributes can include, but are not limited to, cost, strength, durability, life cycle cost, marketability, appearance, packaging, size, suitability, weight, manufacturability, ease of assembly, etc. Thus, embodiments that are described as less desirable in one or more characteristics compared to other embodiments or prior art embodiments are not outside the scope of the present disclosure and may be desirable for a particular application.

[0056] The drawings are in simplified form and are not drawn to precise scale. Directional terms, such as top, bottom, left, right, up, above, over, below, beneath, rear, and front, may be used with respect to the drawings for convenience and clarity of description only. These directional terms and similar directional terms should not be construed as in any way limiting the scope of the present disclosure.

[0057] This document describes embodiments of the present disclosure. However, it should be understood that the disclosed embodiments are merely examples, and other embodiments may take various alternative forms. The accompanying drawings are not necessarily drawn to scale; some features may be enlarged or minimized to show details of particular components. Thus, the specific structural and functional details disclosed herein should not be construed as limiting, but merely as a representative basis for teaching one skilled in the art to practice the systems and methods of the present disclosure in different ways. As will be understood by one of ordinary skill in the art, the various features illustrated and described with reference to any one figure may be combined with features illustrated in one or more other figures to produce embodiments that are not explicitly illustrated or described. Combinations of the illustrated features provide representative embodiments for typical applications. However, for a particular application or implementation, various combinations and modifications of features consistent with the teachings of the present disclosure may be required.

[0058] Embodiments of the present disclosure may be described herein in terms of functional and / or logical block components and various processing steps. It should be understood that these block components may be implemented by a plurality of hardware, software, and / or firmware components configured to perform the specified functions. For example, embodiments of the present disclosure may employ various integrated circuit components, such as memory elements, digital signal processing elements, logic elements, look-up tables, etc., which may perform various functions under the control of one or more microprocessors or other control devices. Additionally, those skilled in the art will understand that embodiments of the present disclosure may be used in conjunction with a variety of systems, and the systems described herein are merely exemplary embodiments of the present disclosure.

[0059] For the sake of brevity, techniques related to signal processing, data fusion, signaling, control, and other functional aspects of the system (as well as the various operating components of the system) may not be described in detail herein. Additionally, the connecting lines shown in the various figures included herein are intended to represent example functional relationships and / or physical couplings between the various elements. It should be noted that alternative or additional functional relationships or physical connections may exist in embodiments of the present disclosure.

[0060] The description is merely illustrative in nature and is in no way intended to limit the present disclosure, its applications, or uses. The broad teachings of the present disclosure may be implemented in a variety of forms. Thus, while the present disclosure includes specific examples, the true scope of the present disclosure should not be so limited since other modifications will become apparent after studying the drawings, the specification, and the appended claims.

Claims

1. A method for estimating a lateral velocity of a vehicle, comprising: receiving sensor data from a sensor of the vehicle; determining a physics-based longitudinal velocity estimate of the vehicle using a physics-based model and the sensor data; determining a data-driven longitudinal velocity estimate of the vehicle using a first neural network and the sensor data; determining, using a second neural network, which of the physics-based longitudinal velocity estimate and the data-driven longitudinal velocity estimate is more reliable to determine a selected longitudinal velocity estimate; determining in real time a lateral velocity of the vehicle using the selected longitudinal velocity estimate; as well as The vehicle is controlled based on the lateral velocity.

2. The method according to claim 1, wherein: The sensor data includes a wheel speed of the vehicle, a longitudinal acceleration of the vehicle, a lateral acceleration of the vehicle, a yaw rate of the vehicle, a road wheel angle of the vehicle, and a wheel torque of the vehicle, and the first neural network is a recurrent neural network.

3. The method according to claim 2, wherein: Determining, using the second neural network, which of the physics-based longitudinal velocity estimate and the data-driven longitudinal velocity estimate is more reliable includes: determining a reliability of the physics-based longitudinal velocity estimate using the second neural network; and The reliability of the data-driven longitudinal velocity estimate is determined using the second neural network.

4. The method of claim 3 further comprising comparing the reliability of the physics-based longitudinal velocity estimate with the reliability of the data-driven longitudinal velocity estimate to determine which of the physics-based longitudinal velocity estimate and the data-driven longitudinal velocity estimate is more reliable. 5 . The method of claim 4 , further comprising determining a final longitudinal velocity based on the selected longitudinal velocity estimate using a first extended Kalman filter.

6. The method according to claim 5, further comprising: comparing the reliability of the physics-based longitudinal velocity estimate to a predetermined reliability threshold to determine whether the reliability of the physics-based longitudinal velocity estimate is less than the predetermined reliability threshold; comparing the reliability of the data-driven longitudinal velocity estimate to the predetermined reliability threshold to determine whether the reliability of the data-driven longitudinal velocity estimate is less than the predetermined reliability threshold; as well as In response to determining that the reliability of the data-driven longitudinal velocity estimate and the reliability of the physics-based longitudinal velocity estimate are both less than the predetermined reliability threshold, increasing a covariance of the first extended Kalman filter. 7 . The method of claim 6 , further comprising determining a lateral velocity of the vehicle using a second extended Kalman filter based on the final longitudinal velocity previously determined using the first extended Kalman filter.

8. A system for estimating lateral velocity of a vehicle, comprising: a plurality of sensors, wherein each of the plurality of sensors is configured to generate sensor data; a controller in communication with the plurality of sensors, wherein the controller is programmed to: receiving sensor data from a plurality of sensors of the vehicle; determining a physics-based longitudinal velocity estimate of the vehicle using a physics-based model and the sensor data; determining a data-driven longitudinal velocity estimate of the vehicle using a first neural network and the sensor data; determining, using a second neural network, which of the physics-based longitudinal velocity estimate and the data-driven longitudinal velocity estimate is more reliable to determine a selected longitudinal velocity estimate; determining a lateral velocity of the vehicle using the selected longitudinal velocity estimate; and The vehicle is controlled based on the lateral velocity.

9. The system according to claim 8, wherein: The sensor data includes a wheel speed of the vehicle, a longitudinal acceleration of the vehicle, a lateral acceleration of the vehicle, a yaw rate of the vehicle, a road wheel angle of the vehicle, and a wheel torque of the vehicle, and the first neural network is a recurrent neural network.

10. The system according to claim 9, wherein: The controller is programmed to: determining a reliability of the physics-based longitudinal velocity estimate using the second neural network; and The reliability of the data-driven longitudinal velocity estimate is determined using the second neural network.

Images