Device for estimating a wheel slip rate of a vehicle and device for estimating a driving speed using the same
A deep learning-based apparatus accurately estimates wheel slip and vehicle speed for four-wheel drive vehicles off-road by processing engine and wheel data, addressing sensor noise and road condition challenges, thereby improving estimation accuracy.
Patent Information
- Application Number
- DE102023131684
- Authority / Receiving Office
- DE · DE
- Patent Type
- Applications
- Current Assignee / Owner
- Priority Date
- 2023-05-09
- Filing Date
- 2023-11-14
- Publication Date
- 2024-11-14
AI Technical Summary
Existing methods struggle to accurately estimate wheel slip rate and vehicle speed for four-wheel drive vehicles operating off-road due to sensor noise and changing road conditions, leading to significant errors in speed calculation.
An apparatus utilizing a deep learning-based wheel slip estimation model, incorporating a storage unit, wheel slip estimation unit, and vehicle speed estimation unit, which processes engine torque, speed, longitudinal and lateral acceleration, yaw rate, and wheel speed to accurately calculate wheel slip and vehicle speed, employing a long-short-term memory network and loss functions for precise estimation.
The apparatus provides accurate estimation of wheel slip rate and vehicle speed even in off-road conditions, reducing errors and ensuring reliable vehicle speed calculation for four-wheel drive vehicles.
Smart Images

Figure 00000000_0000_ABST
Abstract
Description
BACKGROUND 1. TECHNICAL FIELD
[0001] The present disclosure / invention relates to an apparatus for estimating (e.g., calculating) a wheel slip rate of a vehicle and an apparatus for estimating (e.g., calculating) a vehicle speed using the same (e.g., using the apparatus for estimating a wheel slip rate). 2. DESCRIPTION OF RELATED TECHNOLOGY
[0002] Recently, driving on off-road surfaces such as sand, mud, and gravel has become increasingly popular. Interest in off-road vehicles is also growing, as orders for vehicles are increasing in many overseas countries where off-road surfaces make up the majority of the land.
[0003] Conversely, wheel slippage often occurs off-road. The wheels may be in a state where the vehicle cannot move due to the (e.g., lack of) resistance (e.g., friction, e.g., grip) of the road surface (e.g., the road surface); for example, a vehicle may become stuck. For this reason, it is difficult to accurately estimate (e.g., calculate) the off-road wheel slip rate. For general vehicles (e.g., vehicles not driving on off-road surfaces), the slip rate and driving speed are estimated (e.g., calculated).
[0004] An example of a related art technique for calculating a wheel slip rate is described in a Korean patent publication, "Vehicle Slip Rate Calculator" (Application No.: 10-2003-0101251, hereinafter referred to as "related art"). Referring to the related art vehicle slip rate calculator, a method for calculating a vehicle speed using an acceleration sensor and an angular velocity sensor installed in the vehicle is shown, thereby calculating a wheel slip rate.
[0005] Conversely, the method described in the prior art can determine a relatively accurate wheel slip rate because the sensor generates less noise on a flat road surface. However, on an off-road surface, not only does the slope of the road surface (e.g., the road surface, e.g., the area being traveled on) constantly change, but significant noise is also generated in the sensor reading due to random irregularities in the road surface. Therefore, the method described in the prior art can make it difficult to determine an accurate wheel slip rate.
[0006] It can be difficult to determine an accurate wheel slip rate when a four-wheel drive (e.g., four-wheel drive) vehicle is traveling off-road. In a front- or rear-wheel drive vehicle, wheel slip occurs at a wheel that is moving and transmitting power to the road surface. Accordingly, for a front- or rear-wheel drive vehicle, the wheel slip rate can be calculated by estimating (e.g., calculating) the average speed of the remaining non-driven wheels that are not experiencing wheel slip as the vehicle speed.
[0007] However, since most off-road vehicles are four-wheel drive and all four wheels are used for driving, wheel slip can occur on all four wheels. Therefore, using the average of the vehicle speeds calculated for each wheel can introduce a large error in the vehicle speed due to the speed calculated for a wheel with wheel slip.
[0008] Therefore, there is a need for a technology that can accurately estimate (e.g. calculate) the driving speed of a vehicle, even in an off-road vehicle with four-wheel drive. BRIEF EXPLANATION OF THE INVENTION
[0009] One aspect (e.g., an object) of the present disclosure / invention is to provide an apparatus in which the wheel slip rate (e.g., the rate of wheel spin due to lack of grip / friction between the wheel and the traveling surface) of a vehicle traveling off-road (e.g., off-road, e.g., off-road) can be accurately estimated (e.g., calculated).
[0010] Another object of the present disclosure / invention is to provide an apparatus with which the traveling speed of a vehicle traveling off-road can be accurately estimated (e.g., calculated).
[0011] According to one aspect of the present disclosure / invention, an apparatus for estimating (e.g., calculating) a wheel slip rate includes a storage unit that stores a wheel slip estimation model (e.g., a wheel slip calculation model, e.g., a data set for estimating (e.g., calculating) the wheel slip, e.g., a neural network for estimating (e.g., calculating) the wheel slip), and a wheel slip estimation unit (e.g., a wheel slip calculation unit) that estimates (e.g., calculates) wheel slip information using the wheel slip estimation model based on driving information.
[0012] The driving information may include at least one of engine torque (e.g., internal combustion engine torque), speed of an engine (e.g., an internal combustion engine), longitudinal acceleration (e.g., longitudinal acceleration), lateral acceleration (e.g., transverse acceleration), yaw rate (e.g., rotation rate about a vertical vehicle axis), and / or wheel speed (e.g., wheel rotational speed, e.g., wheel speed) of each wheel.
[0013] The wheel slip estimation model can be learned using a deep learning network (e.g., a multi-layer learning network, such as a network with at least one multi-layer neural network).
[0014] The wheel slip estimation model can be learned using a long-short-term memory network (e.g., an LSTM network, a network with a long short-term memory, such as a neural network in whose learning process the technique called “long short-term memory” was / is used).
[0015] In the wheel slip estimation model, at least one of a smooth L1 loss function and a negative Gaussian log-likelihood (e.g., a Gaussian negative log-likelihood, e.g., an NLL) loss function may be used as a loss function (e.g., a function for comparing a target output value and an expected output value, e.g., of a neural network to estimate how well a neural network models training / learning data).
[0016] The wheel slip estimation model can be learned (e.g., by a neural network) by using a wheel slip rate determined with a global positioning system (e.g., a GPS) as a correct value.
[0017] According to another aspect of the present disclosure / invention, an apparatus for estimating (e.g., calculating) a traveling speed includes a receiving unit that acquires (e.g., acquires, e.g., receives) traveling information of a vehicle and a storage unit that stores a wheel slip estimation model (e.g., a wheel slip calculation model). The apparatus also includes a wheel slip estimation unit that estimates (e.g., calculates) wheel slip information using the wheel slip estimation model based on the traveling information. The apparatus also includes a traveling speed estimation unit (e.g., a traveling speed calculation unit) that estimates (e.g., calculates) a traveling speed of the vehicle based on wheel slip information.
[0018] The receiving unit may receive the driving information using a network provided (e.g., provided) in the vehicle.
[0019] The driving information may include at least one of engine torque, engine speed (e.g., combustion engine speed), longitudinal acceleration, lateral acceleration, yaw rate, and / or wheel speed of each wheel.
[0020] The wheel slip estimation model can be learned using a deep learning network.
[0021] The device for estimating a driving speed may further include (for example, have) a preprocessing unit that receives and preprocesses the driving information from the receiving unit. The preprocessing unit may forward the driving information to the wheel slip estimation unit.
[0022] The preprocessing unit may preprocess the received driving information using standardization (e.g., standardized procedures).
[0023] The wheel slip information may include an average and a variance (e.g. deviation) of the wheel slip rates.
[0024] The vehicle speed estimation unit can estimate (e.g., calculate) a vehicle speed for each wheel of the vehicle. The vehicle speed estimation unit can also estimate (e.g., calculate) the vehicle's vehicle speed by applying a weight to the vehicle speed estimated (or calculated) for each wheel.
[0025] The vehicle speed estimation unit may, in estimating the vehicle speed for each wheel, use a different method for estimating the vehicle speed based on a magnitude of the wheel slip rate estimated (e.g., calculated) by the wheel slip estimation unit.
[0026] The vehicle speed estimation unit may determine a weighting of the vehicle speed estimated for each wheel based on a deviation value (e.g., variance value) estimated by the wheel slip estimation unit.
[0027] The vehicle speed estimation unit may further include a post-processing unit that post-processes wheel slip rate information estimated by the wheel slip estimation unit and the vehicle speed estimated by the vehicle speed estimation unit.
[0028] The post-processing unit can set the vehicle speed and wheel slip to “0” (zero) when a product of a wheel rotational angular velocity and a dynamic radius of each wheel is a preset value or less.
[0029] The post-processing unit can post-process the driving speed using an exponential moving average. BRIEF DESCRIPTION OF THE DRAWINGS
[0030] The above and other aspects, features and advantages of the present disclosure / invention should be more clearly understood from the following detailed description taken in conjunction with the accompanying drawings in which: Fig. 1 is a block diagram of an apparatus for estimating (e.g., calculating) a wheel slip rate and for estimating (e.g., calculating) a vehicle speed using the same according to an embodiment, Fig. 2 is a block diagram of an apparatus for estimating (e.g., calculating) the wheel slip rate according to an embodiment, Fig. 3 is a block diagram illustrating a learning process of a wheel slip estimation model (e.g., wheel slip calculation model) according to an embodiment, Fig.4 is a block diagram illustrating a method for estimating (e.g., calculating) a traveling speed of a vehicle according to an embodiment; and Fig. 5 is a diagram illustrating an estimated (e.g., calculated) vehicle speed, an average vehicle speed calculated for each wheel, and a vehicle speed calculated by a global positioning system (GPS), according to one embodiment. DETAILED DESCRIPTION
[0031] Since the present disclosure / invention is susceptible to various modifications and embodiments, some embodiments are illustrated in the drawings and described in detail. However, this is not intended to limit the present disclosure / invention to the described embodiments, and should be understood to include all modifications, changes, and substitutions included within the present invention as defined by the appended claims.
[0032] Terms such as "first" and "second" may be used to describe various components, but the components should not be limited by these terms. These terms are used merely to distinguish between the individual components. For example, a first element may be referred to as a second element, and similarly, a second element may be referred to as a first element, without departing from the scope of the present disclosure / invention. The use of "and / or" includes any combination of multiple related elements or any one of multiple related elements.
[0033] The terms used in this application are for the purpose of describing embodiments only and are not intended to limit the present disclosure / invention. Singular terms include plural terms unless the context clearly indicates otherwise. Throughout this application, the terms "comprise," "have," and the like, as well as variations thereof, are intended to indicate that a feature, number, step, operation, component, part, or combination thereof is described in the specification. It is to be understood, however, that these terms do not preclude the possibility that one or more other features, numbers, steps, operations, components, parts, or combinations thereof are present or added.
[0034] Unless otherwise defined, all terms used herein, including technical or scientific terms, have the same meaning as commonly understood by one of ordinary skill in the art to which this disclosure / invention relates. Terms such as those defined in commonly used dictionaries should be interpreted to have a meaning consistent with their meaning in the context of the relevant art and, unless specifically defined in this application, should not be interpreted in an ideal or overly formal sense.
[0035] Embodiments of the present disclosure / invention will be described in more detail below with reference to the accompanying drawings. When a component, device, element, unit, or the like of the present disclosure / invention is described as having a purpose or performing an operation, function, or the like, the component, device, element, or unit should be considered to be "configured to" fulfill that purpose or perform that operation or function. Furthermore, each component, device, element, unit, and the like may separately embody or be included as part of a processor and memory, such as a non-transitory computer-readable medium.
[0036] Fig.1 is a block diagram of an apparatus for estimating a wheel slip rate and an apparatus for estimating a vehicle speed using the same according to an embodiment.
[0037] Referring to Fig. 1, a device for estimating a wheel slip rate may include a storage unit 400 and a wheel slip estimation unit (e.g., wheel slip calculation unit) 300. Furthermore, a device for estimating a traveling speed of a vehicle may include a receiving unit 100, a device for estimating a wheel slip rate (e.g., wheel slip rate calculation device), and a traveling speed estimation unit (e.g., a traveling speed calculation unit) 500.
[0038] The receiving unit 100 can receive information about the driving state of the vehicle. The receiving unit 100 can receive, for example, engine speeds (e.g., internal combustion engine speed) per minute (RPM), engine torque, longitudinal acceleration, lateral acceleration, yaw rate, wheel speed of each wheel, and the like in the vehicle's Controller Area Network (CAN) signals.
[0039] The receiving unit 100 can receive information about the driving state of the vehicle via a network provided in the vehicle. For example, the receiving unit 100 can receive information about the driving state, such as engine speed, engine torque, longitudinal acceleration, lateral acceleration, yaw rate, and wheel speed of each wheel, via the vehicle's CAN network.
[0040] The receiving unit 100 may collect (for example, be) time-series data of vehicle driving information for a predetermined period of time. Furthermore, the receiving unit 100 may generate (for example, collect) time-series data a predetermined number of times (for example, 30 times) at predetermined time intervals (for example, 20 milliseconds). A preset time interval may be referred to as a sampling time, and a preset number of time points may be referred to as a window size (for example, a sampling, for example, a sampling accuracy).
[0041] Table 1 illustrates the results of a test of the performance as a function of the window size and the sampling time in the wheel slip rate estimating apparatus and the vehicle speed estimating apparatus using the wheel slip rate estimating apparatus according to one embodiment. [Table 1] Window size [sampling] 10 20 30 40 50 Sampling time [ms] 10 MAE [%] 2.05 1.91 1.78 1.67 1.91 RMSE [%] 3.75 3.52 3.58 3.17 3.52 20 MAE [%] 1.95 1.77 1.69 1. 66 1. 64 RMSE [%] 3.58 3.41 3.21 3.21 3.12 50 MAE [%] 1. 94 1.80 1.84 1.76 1.81 RMSE [%] 3.62 3.41 3.49 3.49 3.41 100 MAE [%] 2.07 1.98 1. 82 1. 94 1.86 RMSE [%] 3.61 3.73 3.47 3.69 3.59
[0042] The mean absolute error (MAE) can be an average (e.g., mean) of the absolute error values between the actual model and the estimation model. The mean absolute error can be considered to have a smaller error if the average error is smaller (e.g., the smaller the average error, the smaller the average absolute error). It also allows for an intuitive understanding of the average performance of the estimation model.
[0043] Furthermore, the root mean square error (RMSE) can be the square root of the average of the squares of the errors between the actual model and the estimated model. As with the mean absolute error, the smaller the square error, the smaller the error (e.g., the absolute error, e.g., the actual error) can be assessed.
[0044] Furthermore, according to preliminary tests, the wheel slip estimation time (e.g., the time required to estimate the wheel slip) was too long for all window sizes when the sampling time was 10 milliseconds (ms). Even with a sampling time of 50 ms and window sizes of 40 and 50 samples, the estimation time was similarly long. In one embodiment of the present disclosure / invention, a sampling time of 20 ms and a window size of 30 samples may be adopted in consideration of the accuracy and wheel slip estimation time exemplified in Table 1. In an apparatus for estimating a wheel slip rate and estimating a vehicle speed using the same according to one embodiment of the present disclosure / invention, the sampling time is set to 50 ms and the window size is set to 30 samples. Therefore, the accuracy of the estimated value can be guaranteed, and the estimation time can be reduced.However, the window size and sampling time may be applied in various ways depending on the specifications of the embedded personal computer (PC) or processor employed in the apparatus for estimating a wheel slip rate and the apparatus for estimating a vehicle speed using the same, and the deep learning model applied.
[0045] Back to Fig. 1: A preprocessing unit 200 may perform a predetermined conversion to input the input data received via the receiving unit 100 to the wheel slip estimation unit 300 to be described later. For example, the preprocessing unit 200 may preprocess the input data received via the receiving unit 100 using a standardization method.
[0046] Methods for preprocessing the input data by the preprocessing unit 200 may include normalization and standardization. Normalization may be preprocessing that transforms the input data between (e.g., to a range of values with absolute values between) 0 and 1 by changing the difference between each value and a minimum value compared to the difference between the maximum value and the minimum value of the input data. Furthermore, standardization may be preprocessing that transforms the input data into a normal distribution with a mean of 0 and a standard deviation of 1.
[0047] The results of testing the performance of the wheel slip rate estimator and vehicle speed estimator using the wheel slip rate estimator according to an embodiment with various preprocessing methods are shown in Table 2 below. [Table 2] Performance indicator No preprocessing normalization standardization MAE Unable to train 3.06 [%] 2.77[%l RMSE 4.78[%] 4.39[%] R2 0.760 [-] 0.798 [-]
[0048] The coefficient of determination (R2 value, e.g., coefficient of determination, e.g., accuracy) can indicate how strongly the estimated model correlates with the actual model. The R2 value can also be a ratio between the variance (e.g., deviation) of the predicted value and the variance (e.g., deviation) of the actual value. The closer the coefficient of determination is to 1, the more accurate the model is likely to be. The closer the coefficient of determination is to 0, the less accurate the model.
[0049] As shown in Table 1, it is impossible to learn the wheel slip estimation unit 300 without preprocessing. It can be confirmed that the case where standardization is used as the preprocessing method has higher reliability than the case where normalization is used.
[0050] A wheel slip estimation model 310 formed from (e.g., consisting of, e.g., implemented as) an artificial neural network for which learning (e.g., the learning process) has been completed may be stored in the storage unit 400. The artificial neural network may be implemented based on a long-short-term memory (LSTM), as described below.
[0051] The storage unit 400 is a recording medium suitable for storing the wheel slip estimation model 310. The storage unit 400 may, for example, comprise magnetic media such as hard disks, floppy disks, and magnetic tapes, optical media such as CD-ROM (e.g., Compact Disk Read Only Memory) and DVD (e.g., Digital Video Disk), magneto-optical media such as floppy disks, and / or semiconductor memories such as flash memory or EPROM (e.g., Erasable Programmable ROM) or SSDs (e.g., solid-state drives) based thereon.
[0052] The wheel slip estimation unit 300 may calculate the wheel slip rate and a variance value (e.g., deviation value) thereof for each moving wheel of the vehicle using the wheel slip estimation model (310 in Fig. 2) estimate (e.g. calculate).
[0053] The vehicle speed estimation unit 500 may estimate (e.g., calculate) the vehicle speed using the wheel slip rate estimated by the wheel slip estimation unit 300 and the variance value of the wheel slip rate.
[0054] The wheel slip estimation unit 300 and the vehicle speed estimation unit 500 may be implemented by a non-volatile memory (not shown) and a processor (not shown). The non-volatile memory may be configured to store data related to algorithms configured to control the operation of various components of the vehicle or software instructions for reproducing the algorithms. The processor may be configured to perform a process described below using data stored in a corresponding memory. The memory and processor may be implemented as individual chips. Alternatively, the memory and processor may be implemented as a single chip integrated with each other. A processor may take the form of one or more processors.
[0055] A post-processing unit 600 may post-process the data estimated by the wheel slip estimation unit 300 and the vehicle speed estimation unit 500.
[0056] The post-processing unit 600 may limit the amount of change (e.g., the amount of change) of the estimated data to prevent a phenomenon in which the estimated data changes rapidly (e.g., too quickly). For example, the post-processing unit 600 may limit the amount of change so that the wheel slip rate estimated by the wheel slip estimation unit 300 does not exceed 100% / s (e.g., a change in the wheel slip rate of more than 100% per second). Furthermore, the post-processing unit 600 may limit the change in the vehicle speed, such as the vehicle acceleration, estimated by the vehicle speed estimation unit 500 to a certain value (e.g., 4.9 m / s^2). Therefore, the required data can be stably estimated by preventing a phenomenon in which the data rapidly jumps (e.g., jumps rapidly) due to an operational error, external noise, or the like.
[0057] Furthermore, since the wheel slip estimation can only be estimated (e.g., calculated) when the vehicle is moving, accurate estimation cannot be performed for a state where the vehicle is stationary (e.g., in a state) because learning cannot be performed. Consequently, the wheel slip rate and the traveling speed can be set to zero when the vehicle is stationary (e.g., stopped). The stationary state (e.g., stopped state) of the vehicle can be determined based on the product of the rotational angular velocity of the wheel and the dynamic radius. For example, if the largest value of the product of the rotational angular velocity of the wheel and the dynamic radius calculated from the wheel mounted on the vehicle is equal to the set value (e.g., 0.36 km / h (kilometers per hour)) or less, it can be determined that the vehicle is at a standstill.
[0058] The post-processing unit 600 may set a lower limit for the wheel slip rate and an upper limit for the vehicle speed. Referring to Equation 5 described below, if the wheel slip rate estimated by the wheel slip estimation unit 300 is close to -100%, the estimated vehicle speed may deviate. If the estimated vehicle speed deviates, the vehicle's brake control unit may rapidly brake, and the possibility of getting stuck may greatly increase.
[0059] Furthermore, according to the results of the analysis of actual test data, the wheel slip rate has a negative value only at low speeds, that is, in sections with a driving speed of 2 km / h or less. The negative value of the wheel slip rate may not have a significant impact on the estimation of the vehicle's actual driving speed.
[0060] Consequently, the post-processing unit 600 may apply the lower limit of the wheel slip rate estimated by the wheel slip estimation unit 300 in the process of calculating the traveling speed of the vehicle. For example, if the wheel slip rate estimated by the wheel slip estimation unit 300 is less than or equal to a set value (e.g., predefined value) (e.g., -25%), a wheel slip rate as a set value (e.g., -25%) may be applied to the estimation of the traveling speed. Similarly, for the traveling speed estimated by the traveling speed estimation unit 500, the post-processing unit 600 may apply the upper limit of the set value (e.g., 1.33 times) of the wheel speed calculated by the product of the wheel's rotational angular velocity and the dynamic radius.
[0061] Furthermore, the post-processing unit 600 may estimate the vehicle speed and the wheel slip rate differently under a preset low-speed condition to prevent a divergence between the wheel slip rate and the vehicle speed in low-speed situations. This is intended to prevent a large error from occurring due to a time delay between a vehicle speed and a wheel speed estimated by a global positioning system (GPS) under a low-speed condition. The preset low-speed condition may be a condition where the wheel speed, calculated as the product of the wheel's rotational angular velocity and the dynamic radius, is equal to or less than a first set value (e.g., 1.5 km / h).For the wheels provided in the vehicle, the difference between a wheel speed calculated for the wheel with maximum wheel speed and a wheel speed calculated for the wheel with minimum wheel speed, which is calculated as the product of the rotational angular velocity and the dynamic radius of the wheel, is (for example preferably) within a second target value (e.g. 0.36 km / h).
[0062] The post-processing unit 600 may also use an exponential moving average (EMA, e.g., exponential moving average). The exponential moving average (EMA) may be a moving average that gives more weight and significance to recent data points.
[0063] Components of the vehicle speed estimation device can exchange information by being connected via a wired or wireless network provided in the vehicle. For example, the components of the vehicle speed estimation device can exchange data via the network communication means available in the vehicle, such as Ethernet, Media Oriented Systems Transport (MOST), Flexray, Controller Area Network (CAN), Local Interconnect Network (LIN), Internet, Long Term Evolution (LTE), 5G, Wi-Fi, Bluetooth, Near Field Communication (NFC), Zigbee, radio frequencies (RF), or the like.
[0064] Fig. 2 is a block diagram of an apparatus for estimating wheel slip rate according to an embodiment. Fig. 3 is a block diagram illustrating a learning process of a wheel slip estimation model 310 according to one embodiment.
[0065] Referring to Fig.2, according to one embodiment, the wheel slip estimation model 310 encodes input data using LSTM and decodes the data through a linear layer to estimate the wheel slip rate of each wheel and variance values. Wheel speed information can be prevented from being lost by the wheel slip estimation model 310 using a skip connection (e.g., a (e.g., direct) connection from one layer of a neural network not only to a next layer, but also, or, for example, only to a layer after but one or an even more distant layer). More specifically, the wheel slip estimation model 310 can prevent wheel speed information from being lost by combining an encoded signal and a wheel speed signal before encoding using LSTM. Furthermore, the LSTM used for encoding can have three layers (e.g., consist of three layers).A linear layer used for decoding may have four layers and may also include an activation function between the respective linear layers. The activation function may add nonlinearity to the output data. The activation function according to one embodiment of the present disclosure / invention may be a leaky rectified linear unit (ReLU, e.g., a leaky rectified linear activation function, e.g., a leaky rectified linear unit).
[0066] The LSTM, the skip connection, the linear layer and the activation function itself are well-known technologies, so a detailed description is omitted.
[0067] Referring to Fig.3 illustrates a method for learning the wheel slip estimation model 310 according to an embodiment as an example. The wheel slip estimation model 310 may be learned by comparing a value calculated using GPS with a value calculated by the wheel slip estimation model 310. More specifically, deep learning may be performed on the value calculated using GPS and the value estimated by the wheel slip estimation model 310 by backpropagating the errors.
[0068] The wheel slip estimation model 310 may calculate the wheel slip rate based on the product of the wheel's rotational angular velocity and the wheel's dynamic radius, which are input (e.g., fed) via the receiving unit 100, and the speed value received from the GPS. A wheel slip rate calculated using GPS may be referred to as a reference wheel slip rate. An equation for calculating the reference wheel slip rate may be the same as Equation 1. The reference wheel slip rate may be used as a correct value in the deep learning process of the wheel slip estimation model 310 to train the wheel slip estimation model 310. λi={0if Vx≤1.5 m / s and |Vx−Rωi|≤0.36 m / sRωi−Vxmax (Vx, Rωi, ϵ)otherwise
[0069] In equation 1, λ is the wheel slip rate, V xthe longitudinal speed (km / h) of the vehicle measured by GPS, R is the dynamic radius of the wheel including the tire, ω is the rotational angular velocity of the wheel, and ε can be a non-zero constant as a threshold value. Furthermore, i can refer to any wheel of the vehicle. For example, in a four-wheel drive vehicle, the left front wheel (FL), the right front wheel (FR), the left rear wheel (RL), and the right rear wheel (RR) can be distinguished by setting different i values. Furthermore, the product Rω i from the dynamic radius of the wheel including the tire and the rotational angular velocity of the wheel can be called the wheel speed. By defining ε as a positive non-zero constant, ε can additionally be a value that prevents a divergence of the wheel slip rate when V x and Rω iwhich are in the denominator of the equation for calculating the wheel slip rate have a value of 0.
[0070] Since in the low-speed section (for example in the low-speed section) there is a large time delay between V x and Rω i There is also a problem where a large error occurs. Therefore, in the low-speed section V x be set to 0. The low speed section may refer to a section in which the wheel speed is less than or equal to the first set value (e.g. 1.5 km / h) and the difference between a maximum wheel speed and a minimum wheel speed calculated based on the wheels installed in the vehicle is less than or equal to the second set value (e.g. 0.36 km / h).
[0071] According to one embodiment of the present disclosure, positive (+) slip may occur when the wheel speed is higher than the vehicle's ground speed. Conversely, negative (-) slip may occur when the vehicle's ground speed is higher than the wheel speed.
[0072] Back to Fig. 3: The wheel slip estimation model 310 may perform error backpropagation learning using a loss function. As in an example shown in Equation 2, the wheel slip estimation model 310 may be formed from a combination of a smooth L1 loss function and a Gaussian negative log-likelihood loss function as the loss function. L=lL1+αlNLL
[0073] In equation 2, L can be a loss function, l L1 can be a smooth L1 loss function and I NLL can be a Gaussian NLL loss function.
[0074] The smooth L1 loss function (l L1 ) can be adjusted so that the wheel slip rate estimated by the wheel slip model (e.g., the wheel slip estimation model 310) can estimate a similar value to the GPS-derived wheel slip rate, e.g., the reference wheel slip rate. The smooth L1 loss function can be represented by the following equation, Equation 3. lL1={0.5(μ^i−λi)2β, if |μ^i−λi|<β|μ^i−λi|−0.5∗β, otherwise
[0075] As shown in Equation 4 as an example, the Gaussian negative log-likelihood loss function (l NLL ) learn the mean and variance values of the hatch rate by learning the distribution of the data. lNLL=12(log(max(σ2, ϵ)+(μ^i−λi)2max(σ2, ϵ))
[0076] In equation 4, L can be a loss function, (l L1 ) can be a smooth L1 loss function and (l NLL) can be a Gaussian NLL loss function. In equations 3 and 4, (λ i ) is the reference wheel slip rate calculated as in Equation 1 based on the GPS signal, (û) is the wheel slip rate estimated by the wheel slip estimation unit 300, and (σ 2 ) is the variance value estimated by the wheel slip estimation unit 300. Furthermore, (ε) and (β) are non-zero constants.
[0077] Back to Fig. 3: The wheel slip estimation model 310 may perform error backpropagation learning using the wheel slip rate estimated by the wheel slip estimation model 310 and a reference wheel slip rate calculated from the GPS.
[0078] The wheel slip estimation model 310 learned by the above method may be stored in the storage unit 400. The wheel slip estimation unit 300 may estimate the wheel slip rate and the variance value using the wheel slip estimation model 310 stored in the storage unit 400.
[0079] Fig. 4 is a block diagram illustrating a method for estimating the traveling speed of a vehicle according to an embodiment.
[0080] Referring to Fig. 4 together with Fig. 1, the vehicle speed estimation unit 500 can calculate the vehicle speed (V ̂ x,i) in various ways according to (for example, based on) the wheel slip rate. If the wheel slip rate is less than the set value (e.g., 0), the vehicle speed estimation unit 500 estimates the vehicle speed using Equation 5. If the wheel slip rate is greater than or equal to the set (e.g., preset) value (e.g., 0), the vehicle speed estimation unit 500 may estimate the vehicle speed (V ̂ x,i ) of each wheel present in the vehicle according to equation 6. V^x,i=Rωiλi+1 V^x,i=(1−λi)Rωi
[0081] The traveling speed estimation unit 500 can estimate the traveling speed of the vehicle based on the traveling speed estimated for each wheel. Specifically, the traveling speed of the vehicle can be calculated by multiplying the weight (e.g., the weighting, e.g., the mathematical weight) of each wheel by the ratio of the traveling speed estimated for each wheel and then adding them.
[0082] The calculation of the driving speed (V ̂ x1 ) of the vehicle can be represented as in equation 7 below, the equation for calculating the weight (w i) can be represented (expressed, for example) as Equation 8. The vehicle's running speed estimated based on Equation 7 can be referred to as the first running speed, which is to be distinguished from the vehicle's low-speed running speed, which will be described later. V^x1=∑i=14wiV^x,i wi=(1σi2+ϵ) / ∑i=14(1σi2+ϵ)
[0083] In order to prevent the traveling speed estimated by the traveling speed estimation unit 500 from having a large error from the traveling speed determined by the GPS, the traveling speed estimation unit 500 may further use another method for estimating the traveling speed at low speed.
[0084] In detail, the vehicle speed estimation unit 500 can estimate the vehicle speed ( V ̂ x2 ) according to equation 9 and the wheel slip rate (λ i) according to Equation 10 under the low-speed condition. The low-speed condition may be a condition where, in the low-speed section, the wheel speed is less than or equal to the first set (e.g., preset) value (e.g., 1.5 km / h). The difference between a maximum wheel speed and a minimum wheel speed calculated based on the wheels present in the vehicle is the second set (e.g., preset) value (0.36 km / h) or less. Furthermore, the traveling speed of the vehicle estimated under the low-speed condition may be referred to as a second traveling speed, which is different from the above-mentioned first traveling speed. V^x2=14∑i=14Rωi λi=Rωi−V^x2max (V^x2, Rωi, ϵ)
[0085] Fig.5 is a graph illustrating the estimated traveling speed, the average traveling speed calculated by each wheel, and the traveling speed calculated by GPS according to one embodiment.
[0086] Referring to Fig. 5, the vehicle speed that can be verified by GPS is shown as a "true" line. The vehicle speed resulting from the average of the wheel speeds of the respective wheels of the vehicle according to the prior art is shown as a "Wheel Speed (Average)" graph. The estimated vehicle speed according to an embodiment of the present disclosure / invention is expressed as an "estimated speed."
[0087] As in Fig.5 as an example, the estimated speed according to an embodiment of the present disclosure / invention has fewer errors relative to the actual vehicle speed that can be verified by GPS compared to the speed estimated according to the related art. For example, it can be confirmed that the estimated speed according to an embodiment of the present disclosure / invention has improved accuracy compared to the related art.
[0088] The methods according to embodiments of the present disclosure / invention may be implemented in the form of program instructions executable by various computing means and recorded on a computer-readable medium. Computer-readable media may include program instructions, data files, data structures, and the like, alone or in combination. Program instructions recorded on a computer-readable medium may be designed and configured specifically for the present disclosure / invention, or they may be known and usable by those skilled in the art of computer software.
[0089] Examples of computer-readable media include hardware devices specifically configured to store and execute program instructions, such as read-only memory (ROM), random access memory (RAM), flash memory, and the like. Examples of program instructions include high-level language codes that can be executed by a computer using an interpreter or the like, as well as machine language codes generated by a compiler. The hardware device described above can be configured to operate as at least one software module to perform the operations described in the present disclosure / invention, and vice versa.
[0090] As explained above, the slip rate of a vehicle driving off-road can be accurately estimated using a deep-learned model (e.g., a model generated based on deep learning) for wheel slip estimation.
[0091] Furthermore, the traveling speed of a vehicle traveling off-road can be accurately estimated based on the vehicle's wheel slip rate. The wheel slip rate and traveling speed can be accurately estimated in the same device or in separate devices or devices.
[0092] While exemplary embodiments have been illustrated and described above, it should be apparent to those skilled in the art that modifications and variations may be made without departing from the scope of the present invention as defined by the appended claims. QUOTES CONTAINED IN THE DESCRIPTION
[0000] This list of documents submitted by the applicant was generated automatically and is included solely for the convenience of the reader. This list is not part of the German patent or utility model application. The DPMA assumes no liability for any errors or omissions. Cited patent literature
[0000] 10-2003-0101251
[0004]
Claims
[1] An apparatus for estimating a wheel slip rate, the apparatus comprising: a storage unit (400) storing a wheel slip estimation model (310), and a wheel slip estimation unit (300) that estimates wheel slip information using the wheel slip estimation model (310) based on driving information. [2] The apparatus according to claim 1, wherein the driving information includes at least one of an engine torque, a rotational speed (rpm) of an engine, a longitudinal acceleration, a lateral acceleration, a yaw rate, and a wheel rotation speed of each wheel. [3] The apparatus of claim 1 or 2, wherein the wheel slip estimation model (310) is learned using a deep learning network. [4] The apparatus of any one of claims 1 to 3, wherein the wheel slip estimation model (310) is learned using a long-short-term memory (LSTM) network. [5] The apparatus of claim 4, wherein the LSTM network has a sampling time of 20 milliseconds and a window size of 30 samples. [6] The apparatus according to any one of claims 1 to 5, wherein in the wheel slip estimation model (310) at least one of a smooth L1 loss function and a Gaussian negative log likelihood (NLL) loss function is used as the loss function. [7] The apparatus according to any one of claims 1 to 6, wherein the wheel slip estimation model (310) is learned by applying a wheel slip rate determined using a global positioning system (GPS) as a correct value. [8] Device for estimating a driving speed, comprising: a receiving unit (100) which detects driving information of a vehicle, a storage unit (400) storing a wheel slip estimation model (310), a wheel slip estimation unit (300) that estimates wheel slip information using the wheel slip estimation model (310) based on the driving information, and a traveling speed estimation unit (500) that estimates a traveling speed of the vehicle based on the wheel slip information. [9] The apparatus according to claim 8, wherein the receiving unit (100) receives the driving information via a network provided in the vehicle. [10] The apparatus according to claim 8 or 9, wherein the driving information includes at least one of an engine torque, an engine speed, a longitudinal acceleration, a lateral acceleration, a yaw rate, and a wheel rotation speed for each wheel. [11] The apparatus of any one of claims 8 to 10, wherein the wheel slip estimation model (310) is learned using a deep learning network. [12] The apparatus according to any one of claims 8 to 11, further comprising a preprocessing unit (200) configured to receive and preprocess the driving information from the receiving unit (100) and then transmit the driving information to the wheel slip estimation unit (300). [13] The apparatus of claim 12, wherein the preprocessing unit (200) is configured to preprocess the received driving information using standardization. [14] Apparatus according to any one of claims 8 to 13, wherein the wheel slip information comprises an average and a variance of the wheel slip rates. [15] The apparatus according to any one of claims 8 to 14, wherein the vehicle speed estimation unit (500) is configured to to estimate a driving speed for each wheel of the vehicle, and to estimate the vehicle's speed by applying a weight to the estimated speed for each wheel. [16] The apparatus according to claim 15, wherein the vehicle speed estimation unit (500) is configured to differently apply, in estimating the vehicle speed for each wheel, a method for estimating the vehicle speed based on a magnitude of a wheel slip rate estimated by the wheel slip estimation unit (300). [17] The apparatus according to claim 15 or 16, wherein the vehicle speed estimation unit (500) is configured to determine a weight of the vehicle speed estimated for each wheel based on a deviation value estimated by the wheel slip estimation unit (300). [18] The apparatus according to any one of claims 8 to 17, further comprising a post-processing unit (600) configured to post-process wheel slip rate information estimated by the wheel slip estimation unit (300) and the vehicle speed estimated by the vehicle speed estimation unit (500). [19] The apparatus according to claim 18, wherein the post-processing unit (600) is configured to set the vehicle speed and the wheel slip rate to "0" (zero) when the product of a wheel rotational angular velocity and a dynamic radius of each wheel is a preset value or less. [20] The apparatus of claim 18 or 19, wherein the post-processing unit (600) is configured to post-process the vehicle speed using an exponential moving average.
Citation Information
Patent Citations
Slip rate producing device
KR100534831B1
Cited By
Method for detecting wheel slippage of a vehicle wheel
DE102024209655A1
