Suspension system

Abstract

To provide a sensorless suspension system that supports learning by extracting unfavorable states of a neural network when substituting sensor functions with the neural network.SOLUTION: A suspension system for calculating sensor information and state information of a suspension in association with each other, by machine learning includes: a weight parameter storage unit; a vehicle state estimation unit; a first vehicle behavior calculation unit that calculates a first physical value based on estimation results; a second vehicle behavior calculation unit that calculates a second physical value based on the sensor information; an estimation accuracy verification unit that outputs estimation accuracy of the state of the suspension by comparing the first physical value and the second physical value; and a traveling data management unit that instructs learning of weight parameters based on an output result of the estimation accuracy.SELECTED DRAWING: Figure 1

Description

Technical Field

[0001] The present invention relates to a suspension system.

Background Art

[0002] In a suspension control system that controls the posture of a vehicle by changing the damping force of a suspension spring according to road surface conditions and driving conditions, control values are calculated by detecting sensor data obtained from sensors that detect the piston speed and spring up-and-down speed of the suspensions of the four wheels of the vehicle, and the suspensions are controlled. Regarding this sensor function, from the viewpoints of cost reduction and reduction of fitting man-hours, there is a technology that realizes sensorless by mounting a neural network that realizes a logic or an estimation function for estimating sensor values from CAN (Controller Area Network) data.

[0003] However, the driving roads are diverse, and it is not easy for the above-described sensor value estimation technology to replace the sensor function on all roads. Particularly when a neural network is applied, it may be difficult to continuously replace the sensor with the learned neural network mounted at the time of shipment. Therefore, in a suspension control system, a continuous learning system for improving the coverage of roads is necessary.

[0004] Furthermore, in the field of image recognition where machine learning may be applied, it is easy for a human to identify an image with a low recognition rate and additional learning can be performed specifically for that image. On the other hand, in a sensor value estimation technology applying a neural network assuming sensorless, it is difficult to identify an unfavorable road condition because there is no sensor value serving as teacher data.

[0005] For example, Patent Document 1 below discloses a damper control system that can receive feedback data regarding the behavior of a vehicle and perform control of the characteristics of a damper with high response performance and robustness while executing a machine learning algorithm. [Prior art documents] [Patent Documents]

[0006] [Patent Document 1] Japanese Patent Publication No. 2021-17168 [Overview of the project] [Problems that the invention aims to solve]

[0007] Building upon conventional technologies, achieving sensorless suspension systems that estimate the suspension state from other sensor information without directly detecting it with sensors requires continuous learning to improve the accuracy of machine learning models by extracting states where sensor value estimation is inefficient. In light of this, the present invention aims to provide a sensorless suspension system that, when replacing sensor functions with neural networks, extracts states where the neural network is inefficient and supports its learning process. [Means for solving the problem]

[0008] A suspension system that correlates sensor information acquired from multiple sensors installed in a vehicle with suspension state information installed in the vehicle and calculates using machine learning, comprising: a weight parameter storage unit that stores weight parameters calculated by machine learning; a vehicle state estimation unit that outputs an estimation result of the suspension state based on the sensor information and the weight parameters; a first vehicle behavior calculation unit that calculates a first physical value of a physical quantity related to the behavior of the vehicle based on the output estimation result; a second vehicle behavior calculation unit that calculates a second physical value of the physical quantity based on the sensor information; an estimation accuracy verification unit that outputs the estimation accuracy of the suspension state by the vehicle state estimation unit by comparing the first physical value and the second physical value; and a driving data management unit that instructs a learning management server to learn the weight parameters based on the output result of the estimation accuracy by the estimation accuracy verification unit. [Effects of the Invention]

[0009] When the suspension state is estimated from other sensor information without directly detecting it with a sensor, a sensorless suspension system can be provided that extracts states that the neural network is not good at and supports learning. [Brief explanation of the drawing]

[0010] [Figure 1] Functional block diagram of a suspension system according to the first embodiment of the present invention. [Figure 2A] Figure 1 shows the vehicle state estimation unit represented by a neural network. [Figure 2B] Diagram showing the relationship between input and output datasets when training a neural network. [Figure 3A] Flowchart for vehicle status estimation and learning notification on the vehicle side [Figure 3B] Flowchart for updating vehicle-side weight data [Figure 3C] Flowchart for server-side weight data learning [Figure 3D] Flowchart for handling anomaly detection on the server side [Figure 4A] Example terminal screen for visualizing CAN data, vehicle behavior estimation results, and vehicle behavior physical value calculation results. [Figure 4B] Another example of a terminal screen intended for visualizing vehicle data. [Figure 4C] An example of a system that allows the driver to review the results of vehicle behavior estimation. [Figure 5] Functional block diagram of a suspension system according to a second embodiment of the present invention [Figure 6] An example of the internal configuration of the wheel speed reliability determination unit shown in Figure 5. [Figure 7] Functional block diagram of a suspension system according to a third embodiment of the present invention. [Figure 8] Example of a selector control signal in the third embodiment [Figure 9]Flowchart of the Third Embodiment [Figure 10] Functional Block Diagram of Suspension System According to the Fourth Embodiment

[0011] Hereinafter, embodiments of the present invention will be described with reference to the drawings. The following description and drawings are examples for explaining the present invention, and for the sake of clarity of explanation, appropriate omissions and simplifications have been made. The present invention can be implemented in various other forms. Unless otherwise limited, each component may be singular or plural.

[0012] In the drawings, the positions, sizes, shapes, ranges, etc. of each component shown may not represent the actual positions, sizes, shapes, ranges, etc. in order to facilitate understanding of the invention. Therefore, the present invention is not necessarily limited to the positions, sizes, shapes, ranges, etc. disclosed in the drawings.

[0013] (First Embodiment of the Present Invention and Overall Configuration of the Apparatus) (Fig. 1) Assuming a sensorless suspension system without an in-vehicle suspension sensor, a suspension system that controls damping force to improve the ride comfort of an automobile has a configuration of a vehicle 101, a learning management server 102, and an Internet environment 103. The vehicle 101 has a configuration of a suspension ECU (Electronic Control Unit) 104, an ECU 105, sensors 106a... 106b, a sensor ECU 107, a CAN 108, an active suspension 109 (hereinafter referred to as suspension 109), a display unit 110, and an interface (I / F) 111.

[0014] The suspension ECU 104 has a vehicle state estimation unit 112, a weight data management unit 113, a weight parameter storage unit 114, a suspension control value calculation unit 115, a first pitch rate calculation unit 116, a second pitch rate calculation unit 117, an estimation accuracy verification unit 118, a running data management unit 119, and a running data storage unit 120.

[0015] The suspension 109 suppresses vibrations transmitted from the road surface via the tires when the vehicle 101 is in motion, which are felt by the driver and passengers seated in the seats. The damping force characteristics of the suspension 109 are controlled by the suspension ECU 104, and the suspension control value calculation unit 115 of the suspension ECU 104 derives the control values ​​of the suspension 109.

[0016] The learning management server 102 is a server that manages the learning of the vehicle state estimation unit 112 included in the suspension ECU 104, and transmits information to the vehicle 101 via the internet environment 103. The ECU 105 transmits the information from the learning management server 102 transmitted from the internet environment 103 to the CAN 108. As a result, information is shared within the suspension ECU 104 with each functional unit via the CAN 108. The sensor ECU 107 acquires information from sensors 106a...106b and transmits it to the CAN 108, thereby sharing sensor information with each functional unit of the vehicle 101.

[0017] The following describes the various functions of the suspension ECU 104. The suspension ECU 104 acquires information from or outputs information to the CAN 108 via interface 111. The weight data management unit 113 reads the weight parameters stored in the weight parameter storage unit 114 (weight parameter memory unit) to determine the calculation specifications to be used for estimation by the vehicle state estimation unit 112. The weight data management unit 113 also acquires weight parameters learned from the learning management server 102 via CAN 108 and updates the weight parameters.

[0018] The vehicle state estimation unit 112 refers to sensor information acquired by sensors 106a...106b as specific information from the data constantly transmitted within the CAN 108. Sensors 106a...106b detect the rotational speed (wheel speed) of one of the four wheels of the vehicle 101, as well as longitudinal acceleration, lateral acceleration, vertical acceleration, yaw rate, roll rate, pitch rate, etc., in the vehicle behavior system, and transmit this sensor information to the CAN 108. In Figure 1, sensors 106a...106b are shown as if there were only two types, but in reality, they measure the aforementioned physical values, and it goes without saying that there are many of them. Furthermore, if the vehicle 101 on which the vehicle state estimation unit 112 is installed is a learning vehicle, it is equipped with acceleration sensors for the suspension, and that data is also transmitted to the CAN 108.

[0019] The vehicle state estimation unit 112 acquires sensor information from each of the sensors 106a...106b. The vehicle state estimation unit 112 also acquires the values ​​of the weight parameters read from the weight parameter storage unit 114 by the weight data management unit 113 as data related to the vehicle state during driving. Based on the acquired sensor information and the acquired weight parameters, the vehicle state estimation unit 112 outputs estimation results for the physical quantities related to suspension control for each of the four wheels.

[0020] The suspension control value calculation unit 115 calculates control values ​​for controlling the damping force of the suspension 109 based on the physical quantities estimated by the vehicle state estimation unit 112. By calculating control values ​​using estimation in this way, it becomes unnecessary to install acceleration sensors to measure the vertical speed of the sprung mass and the stroke speed of the piston, as in conventional systems, thus contributing to cost reduction.

[0021] Next, we will explain the verification of the estimation accuracy of the vehicle state estimation unit 112. First, the vehicle state estimation unit 112 transfers the estimated sprung mass vertical speeds of the four-wheel suspension (state information of the suspension 109) to the first pitch rate calculation unit 116. The first pitch rate calculation unit 116 calculates the pitch rate (first physical value) based on the transferred sprung mass vertical speeds of the four wheels. Meanwhile, the four-wheel speed data (sensor information) included in the CAN 108, acquired from multiple sensors 106a...106b provided on the vehicle 101, is transferred to the second pitch rate calculation unit 117. The second pitch rate calculation unit 117 calculates the pitch rate (second physical value) based on the transferred four-wheel speed data.

[0022] This invention focuses on the fact that the pitch rate can be calculated using both the vertical velocity of the sprung mass and the wheel speed of the four wheels. The aim is to verify the estimation accuracy of the first pitch rate derived from the estimation result by using the second pitch rate as a model. By comparing the two pitch rates derived from calculations based on the vertical velocity of the sprung mass and the wheel speed of the four wheels, if they are similar in value, it is determined that the estimation accuracy of the vertical velocity of the sprung mass estimated by the vehicle state estimation unit 112 is good.

[0023] The estimation accuracy verification unit 118 compares the two pitch rates (first physical value and second physical value) calculated by the first pitch rate calculation unit 116 and the second pitch rate calculation unit 117. If the difference between the two pitch rates is less than a predetermined threshold, the estimation unit 118 determines that the estimation result output by the vehicle state estimation unit 112 is not problematic. On the other hand, if the difference between the two pitch rates is greater than a predetermined threshold, the estimation unit 118 determines that the estimation result output by the vehicle state estimation unit 112 is problematic and identifies a state of poor output accuracy for the vehicle state estimation unit 112. When the estimation accuracy verification unit 118 identifies a state of poor output accuracy for the vehicle state estimation unit 112, it assigns a problem flag to the time series dataset input to the vehicle state estimation unit 112 via CAN 108.

[0024] In this way, the vehicle state estimation unit 112 outputs information about the difference between the first physical value and the second physical value regarding the state of the suspension 109, which is the estimated result output by the vehicle state estimation unit 112, as the estimation accuracy. Note that the first physical value and the second physical value are not limited to pitch rate, but may be other vehicle behavior data that can be calculated based on multiple pieces of information or data, such as roll rate.

[0025] In this way, by comparing the pitch rate calculated from the sprung mass vertical velocity estimated by the neural network (the pitch rate calculated by the first pitch rate calculation unit 116) with the pitch rate calculated from the wheel speed information from CAN 108 that does not go through the neural network (the pitch rate calculated by the second pitch rate calculation unit 117), it is determined whether the accuracy of the sprung mass vertical velocity estimation by the vehicle state estimation unit 112 is low.

[0026] The weight parameters used in the neural network of the vehicle state estimation unit 112 are parameters obtained by training data created using a verification vehicle equipped with an acceleration sensor for the suspension, separate from the vehicle 101.

[0027] The estimation accuracy verification unit 118 transfers the datasets flagged as problematic to the driving data management unit 119. The driving data management unit 119 stores the transferred datasets in the driving data storage unit 120. Through this process, the driving data storage unit 120 accumulates datasets that the vehicle state estimation unit 112 is not good at. Based on the accumulated datasets that the driving data management unit 119 is not good at, the driving data management unit 119 instructs the learning management server 102 to update the weight parameters for the driving environment of the vehicle 101. The learning management server 102 performs learning to update the weight parameters in accordance with the instructions from the driving data management unit 119, which were made based on the output of the estimation accuracy verification unit 118. In this way, the learning management server 102 learns the weight parameters based on the output results of the estimation accuracy verification unit 118 in the suspension ECU 104 mounted on the vehicle 101.

[0028] (Figure 2A) The neural network of the vehicle state estimation unit 112 is composed of a three-layer hierarchical neural network in which elements of an input layer (number of elements i) 201, a hidden layer (number of elements j) 202, and an output layer (number of elements K) 203 are hierarchically connected. The input layer element group 201 of the neural network includes an input element group 201a for the wheel speed time series and an input element group 201b for the vehicle behavior time series, and the state of the suspension 109 is estimated using these two element groups.

[0029] The number of elements in the hidden layer group 202 is generally determined by the number of elements in the input layer group 201 and the output layer group 203, but it should be the number that maximizes the accuracy of state estimation by the neural network. The output layer group 203 outputs the estimated instantaneous value of the piston speed or the sprung-up and down speed of the suspension 109 mounted on the vehicle. Although Figure 2A shows a multilayer perceptron as an example of a neural network, a recurrent neural network, which is advantageous for learning time-series data, may also be used.

[0030] Each element of the input layer element group 201 and each element of the hidden layer element group 202 are connected by weights W1ij (i=1~I, j=1~J), and each element of the hidden layer element group 202 and the output layer element 203 are connected by weights W2jk (j=1~J, k=1). This weight information (weight parameters) is represented by a matrix of weights W1ij and W2jk. The weight parameters are determined in advance by machine learning and stored in the weight parameter storage unit 114. The vehicle state estimation unit 112 retrieves the weight parameters from the weight parameter storage unit 114 via the weight parameter management unit 113 and performs calculations using a neural network. In this example, the simplest hidden layer element group 202 is shown as a fully connected neural network with one layer, but the system is not limited to this.

[0031] (Figure 2B) The upper diagram shows wheel speed as an example of the information from CAN108. The neural network training is performed by driving a test vehicle equipped with acceleration sensors that measure piston speed and sprung mass vertical speed for controlling the suspension 109, as described above, and is carried out using time-series data 204 of the data contained in the CAN of the test vehicle and sensor data (time-series data of on-board sensors) 205.

[0032] Specifically, discrete values ​​sampled from the time-series data 204 within the window 206 are set in the neural network's input layer elements 201 (Figure 2A). Then, the corresponding instantaneous sensor data values ​​207 are set in the neural network's output layer elements 203 (Figure 2A). Regarding the relationship between the window 206 and the instantaneous sensor data values ​​207, if the width of the window 206 is, for example, 1 second, the instantaneous sensor data values ​​207 are designed to be estimated from the driving history of 1 second prior. Therefore, if the sampling interval is, for example, 20 m / second, there will be 50 sampling points within the window 206 (= 1 second ÷ 20 m / second), and the combination of these 50 data points and the instantaneous sensor data values ​​207 is defined as the training dataset. Then, by sliding the window 206 and the instantaneous sensor data values ​​207 in the time direction every 20 m / second, the dataset is multiplied by n. This allows for training of the neural network using a large number of datasets.

[0033] (Figure 3A) This section explains the estimation of the vehicle state and the notification of the need for learning in vehicle 101. First, in step S101, the vehicle state estimation unit 112 determines whether vehicle 101 is running. If vehicle 101 is running, the process proceeds to steps S102 and S103; otherwise, the process ends.

[0034] In step S102, the second pitch rate calculation unit 117 calculates the pitch rate from the wheel speed. Meanwhile, in step S103, the vehicle state estimation unit 112 estimates the sprung mass vertical velocity. In step S104, the first pitch rate calculation unit 116 calculates the pitch rate based on the sprung mass vertical velocity estimated in step S103, and the process proceeds to step S105.

[0035] In step S105, the estimation accuracy verification unit 118 calculates the estimation accuracy (error) based on the two pitch rates calculated in steps S102 and S104. In step S106, as an example of a predetermined threshold, if this estimation accuracy (error) is greater than 1 degree, the process proceeds to step S106; otherwise, it returns to step S101. Note that although an error of 1 degree was used as the threshold for judging the estimation accuracy using the pitch rate, this value is not the only one that can be used.

[0036] In step S107, the estimation accuracy verification unit 118 assigns a learning requirement flag to data where the estimation accuracy (error) is greater than 1 degree and the estimation accuracy is deemed poor. In step S108, the driving data management unit 119 writes the driving data with the learning requirement flag assigned to it to the driving data storage unit 120. In step S109, the driving data management unit 119 counts the number of learning requirement flags.

[0037] In step S110, the driving data management unit 119 determines whether the number of learning flags stored in the driving data storage unit 120 has exceeded 1000 counts. This 1000 count becomes the condition for transferring the data accumulated in the driving data storage unit 120 to the learning management server 102. If the number of learning flags exceeds 1000 counts, the process proceeds to step S111, and the driving data with the learning flags assigned is transferred to the learning management server 102, which is the driving data server. Note that while the criterion for the number of learning flags to transfer data to the learning management server 102 is set at 1000, it is not limited to this value.

[0038] (Figure 3B) The following describes the data update of weight parameters in vehicle 101. In step S201, the weight data management unit 113 determines whether or not weight parameters have been received from the learning management server 102. If weight parameters have been received, the process proceeds to step S202; otherwise, step S201 is repeated.

[0039] In step S202, the weight data management unit 113 stores the acquired weight parameter data in the weight parameter storage unit 114. In step S203, the weight data management unit 113 determines whether the weight parameters have been stored in the storage unit 114. If the weight parameters have been stored, the process proceeds to step S204; otherwise, step S203 is repeated.

[0040] In step S204, the weight data management unit 113 switches the weight parameters and repeats the flow from step S201, and proceeds to the flow of step S205. In step S205, the vehicle state estimation unit 112 estimates the vehicle state based on the weight parameters and outputs the estimation result.

[0041] In step S206, the estimation accuracy verification unit 118 determines whether there is an abnormality in the estimation output of the vehicle state estimation unit 112. If there is, the process proceeds to step S207; otherwise, the flow from step S201 is repeated. In step S207, the estimation accuracy verification unit 118 assigns a learning requirement flag to the driving data with abnormalities. In step S208, the estimation accuracy verification unit 118 notifies the developers of the vehicle state estimation unit 112, who possess a terminal with a terminal screen 407 as shown in Figure 4B below, or inspectors who perform maintenance and inspections, of the abnormality in the estimation output, and then the flow from step S201 is repeated.

[0042] (Figure 3C) The learning of weight parameters in the learning management server 102 will now be described. In step S301, the learning management server 102 receives driving data from the vehicle 101. Next, in step S302, the learning management server 102 performs training on the neural network. Then, in step S303, the learning management server 102 stores the weight parameters. Next, in step S304, the learning management server 102 transfers the weight parameters to the weight data management unit 113, and the flow ends.

[0043] (Figure 3D) This is the flow for detecting and responding to anomalies in vehicle state estimation on the learning management server 102. In step S401, the learning management server 102 determines whether there is an anomaly notification regarding vehicle state estimation from the estimation accuracy verification unit 118. If there is an anomaly notification, the server proceeds to step S402; otherwise, step S401 is repeated. In step S402, the learning management server 102 verifies the output of the vehicle state estimation unit 112.

[0044] In step S403, the learning management server 102 determines whether an abnormality has occurred in the output of the vehicle state estimation unit 112. If an abnormality is determined to have occurred, the process proceeds to step S404; otherwise, the process proceeds to step S406.

[0045] In step S404, the latest weight parameter data stored in the learning management server 102 is saved. Then, in step S405, learning is performed again in the learning management server 102, and the flow of step S402 is repeated. In step S406, since it was determined that no abnormality occurred in step S403, the weight parameters are transferred from the learning management server 102 to the vehicle 101. Note that although the retraining was explained as saving the latest weight parameters and learning with new initial values, it may also be an additional learning process using the latest weight parameters as initial values.

[0046] (Figure 4A) In the measurement device screen 401 used in a driving test, which visualizes data related to CAN system data, vehicle behavior estimation results, and calculated vehicle behavior physical values, the upper section displays multiple types of data 402, 403 (CAN system data) referenced from CAN 108, the middle section displays vehicle behavior estimation results 404 output by the vehicle state estimation unit 112, and the lower section displays time-series data 405 of vehicle behavior physical values. Furthermore, there is a file output button 406 for outputting and storing a dataset as text data in CSV format for input into a neural network, for example. The button 406 may be a physical button on the device, selected by touch operation, or selected using a pointing device. It may also be selected by touch operation using fingers in conjunction with a touch sensor function.

[0047] (Figure 4B) The terminal screen 407 displays side-by-side the pitch rate 408 calculated by the first pitch rate calculation unit 116 based on the output of the neural network, which is the output of the vehicle state estimation unit 112, and the pitch rate 409 calculated by the second pitch rate calculation unit 117, which is the wheel speed of all four wheels and is not based on the output of the neural network. Furthermore, the accuracy judgment result 410 is displayed alongside the two pitch rates, indicating a high level if the difference between the two pitch rates is greater than or equal to a predetermined value, and a low level if the difference between the two pitch rates is less than a predetermined value.

[0048] (Figure 4C) Figure 4C shows how the results of vehicle behavior estimation can be viewed by the user through the display unit 110, with the user being the driver as an example. The display content of the instrument panel 411 includes the vehicle behavior estimation results 412. For example, lamp 413 lights up when the accuracy judgment result 410 is high level and the error is large. The display meter 414 shows a graph that represents the percentage of distance or time during which the accuracy judgment result 410 (Figure 4B) is high level relative to the distance traveled or the total travel time.

[0049] As a result, the estimation accuracy output by the estimation accuracy verification unit 118 is presented to users such as the developers of the vehicle condition estimation unit 112, inspectors who perform maintenance and inspections and who have terminals, and the driver of the vehicle 101, through the display unit 110. This allows them to check the accuracy of the suspension system to which the neural network is applied and to recognize low accuracy conditions.

[0050] (Second embodiment) (Figure 5) The second embodiment focuses on the fact that the calculation of the pitch rate derived from wheel speed has different reliability depending on the weather and road surface conditions. Compared to the first embodiment, the suspension system has a weather information server 503, a GPS sensor 504, a stereo camera 505, and a wheel speed reliability determination unit 506. The wheel speed reliability determination unit 506 determines the reliability of the output of the estimation accuracy verification unit 118 based on information from outside the vehicle 101, such as the weather information server 503 and information from the on-board stereo camera 505 and GPS sensor 504. In this way, by referring to the road surface conditions and focusing on the fact that the reliability of the pitch rate changes depending on the road surface conditions, the aim is to extract the neural network's weak points by reflecting the reliability determination results.

[0051] (Figure 6) The wheel speed reliability determination unit 506 includes a road surface μ estimation unit 601 and a wheel speed reliability calculation unit 602. The information referenced by the wheel speed reliability determination unit 506 includes information from the stereo camera 505, GPS sensor 504, weather information server 503, and information on the condition of the tires mounted on the vehicle 101. It references this information and inputs it into the road surface μ estimation unit 601.

[0052] The road surface μ estimation unit 601 has the function of detecting conditions in which the road surface is generally prone to slipping, such as when puddles have formed due to rainfall, or when the road surface μ is low due to snow or ice, based on information from the stereo camera 505. However, this does not necessarily require the stereo camera 505; any device capable of similar detection will suffice. For example, the vehicle's position information obtained from the GPS sensor 504 may be used to query the weather information server 503 and estimate the road surface conditions.

[0053] In addition, factors that are determined on a regular basis, other than those that change depending on the surrounding conditions as described above, such as whether the ground is sandy or whether the dirt surface has a coarse grain, may be managed in association with the map, and the system may detect whether the road surface is prone to slipping or wheel spin based on the information from the GPS sensor 504.

[0054] The wheel speed reliability determination unit 506 calculates the reliability of the wheel speed information based on the road surface information output by the road surface μ estimation unit 601, the vehicle 101 information, and the condition of the tires mounted on the vehicle 101 (e.g., grip force), and outputs the calculation result to the estimation accuracy verification unit 118. Generally, the road surface μ is considered to be around 0.8 for paved dry roads, in the range of 0.4 to 0.6 for paved wet roads, in the range of 0.2 to 0.5 for snow-covered roads, and in the range of 0.1 to 0.2 for icy roads. The total friction coefficient μ' is estimated from this road surface μ and the condition of the tires, and the physical quantity at which slippage begins is calculated by integrating it with the weight N of the vehicle 101.

[0055] The estimation accuracy verification unit 118 (Figure 5) verifies the estimation accuracy based on the reliability of the wheel speed information calculated by the wheel speed reliability determination unit 506, the pitch rate calculated by the first pitch rate calculation unit 116, and the pitch rate calculated by the second pitch rate calculation unit 117. If the reliability of the wheel speed information determined by the wheel speed reliability determination unit 506 is high, the process proceeds as in the first embodiment. However, if the reliability of the wheel speed information is low, the system continues estimating the vehicle state using the neural network or switches to another alternative means.

[0056] The aforementioned road surface μ is just one example, and if tire conditions cannot be obtained, the calculation of μ' can be simplified. Furthermore, the accuracy of the vehicle weight N may be improved by including the weight of the occupants and cargo in addition to the vehicle's own weight.

[0057] (Third embodiment) (Figure 7) The suspension ECU 104 according to this embodiment is newly equipped with a selector 703 and an estimated value verification / correction unit 705 compared to the first embodiment. In addition, the vehicle state estimation units 112a to 112d are installed for each of the four wheels and work in coordination with each other. The vehicle state estimation units 112a to 112d are installed for FL (front left), FR (front right), RL (rear left), and RR (rear right), respectively.

[0058] The selector 703 selects three estimation results from four vehicle state estimation units 112a to 112d, creating different combinations. The first pitch rate calculation unit 116 calculates four types of pitch rates (first physical values) based on the three estimation results selected by the selector 703. This is because, assuming the vehicle 101 is a rigid body, the first physical value (pitch rate) can be calculated using the estimation results for three wheels (sprung mass vertical velocity), and therefore, four types of first physical values ​​can be calculated by selecting the three estimation results in different ways.

[0059] The reason for calculating in this way is that, in the comparison by the estimated value verification / correction unit 705, for example, if only the sprung mass vertical velocity of one wheel is of low accuracy, then the pitch rate calculated using the sprung mass vertical velocity of the other three wheels excluding that one wheel is likely to be nearly identical to the pitch rate derived from the wheel speed, while the three pitch rates including that one wheel are likely to be less identical to the pitch rate derived from the wheel speed. In this way, the estimated value verification / correction unit 705 determines that the estimated sprung mass vertical velocity values ​​of the three wheels used to calculate the pitch rate that matches the pitch rate derived from the wheel speed are highly accurate, and transfers the corresponding estimated sprung mass vertical velocity values ​​for the three wheels to the suspension control value calculation unit 115, excluding the remaining wheel.

[0060] Furthermore, as mentioned above, the pitch rate can be calculated if data for three wheels is available, so the estimated value verification / correction unit 705 uses the correct pitch rate derived from the wheel speed and the sprung mass vertical speeds of the three wheels to inversely calculate the sprung mass vertical speed of the one wheel that was determined to be low accuracy. The estimated value verification / correction unit 705 then transfers the inversely calculated low-accuracy sprung mass vertical speed of the one wheel to the suspension control value calculation unit 115. In other words, the estimated value verification / correction unit 705, which performs estimation accuracy verification, verifies the four types of estimation results and derives the estimation results by inverse calculation.

[0061] Thus, if a high-precision pitch rate is obtained for at least one system, the estimation results of the three systems of sprung mass vertical velocity used to derive it can be determined to be highly accurate, while the estimation results of the remaining system of sprung mass vertical velocity can be determined to be low-precision. If the sprung mass vertical velocity is low-precision, it can be adopted as the estimation result by inversely calculating using the pitch rate calculated by the second pitch rate calculation unit 117 and the three systems of sprung mass vertical velocity, and can be treated as training data for the neural network that realizes the vehicle state estimation unit 112 performed by the learning management server 102.

[0062] Furthermore, if all four difference values, which are based on a comparison between each of the four first physical values ​​and the second physical value, are smaller than a predetermined threshold, the estimated value verification / correction unit 705 adopts the four estimated results output from the four vehicle state estimation units 112a to 112d as the state of the suspension 109 and transfers these estimated results to the suspension control value calculation unit 115 (steps S10 and S18 in Figure 9, described later).

[0063] The estimated value verification / correction unit 705 transfers the input dataset from one of the vehicle state estimation units 112a to 112d, which output low-accuracy sprung mass vertical speeds, and the sprung mass vertical speeds calculated by the estimated value verification / correction unit 705 to the driving data management unit 119, which then stores the input dataset and sprung mass vertical speeds in the driving data storage unit 120.

[0064] When a certain number of data sets (for example, 1000) are stored in the driving data storage unit 120, the driving data management unit 119 uses the count as a transfer condition to transfer the data sets to the learning management server 102 via the interface 111.

[0065] As a result, the learning management server 102 can store datasets that the vehicle state estimation unit 112 is not good at, and by performing additional learning using the stored data, the accuracy of the vehicle state estimation units 112a to 112d can be improved.

[0066] (Figure 8) Let's explain the selector 703. In Figure 8, the selector 703 has four phases based on the vehicle state estimation period 801, which is the operating period of the vehicle state estimation unit 112, and shows how to select three from the FL selection signal 802, FR selection signal 803, RL selection signal 804, and RR selection signal 805. For example, if the period 801 is 20 m / s, the combination of selecting three wheels will differ in each of the four divisions of the 20 m / s period.

[0067] As shown in the example in Figure 8, in the first period, the FL selection signal 802, FR selection signal 803, and RL selection signal 804 are at high levels, so the three types FL, FR, and RL are selected. In the second period, the FR selection signal 803, RL selection signal 804, and RR selection signal 805 are at high levels, so the three types FR, RL, and RR are selected. In the third period, the FL selection signal 802, RL selection signal 804, and RR selection signal 805 are at high levels, so the three types FL, RL, and RR are selected. In the fourth period, the FL selection signal 802, FR selection signal 803, and RR selection signal 805 are at high levels, so the three types FL, FR, and RR are selected.

[0068] Note that in Figure 8, there is one estimated value verification / correction unit 705 and the control signal is for serial processing; however, if there are four estimated value verification / correction units 705, it is generally performed using parallel processing.

[0069] (Figure 9) The flowchart for the third embodiment will now be described. In step S1, the value of the counter in the driving data management unit 119 is set to 0 (reset). In step S2, the first pitch rate calculation unit 116 performs pitch rate calculation A using the sensor estimate calculated using the sprung mass vertical speeds of wheels FL, FR, and RL. In step S3, the first pitch rate calculation unit 116 performs pitch rate calculation B using the sensor estimate calculated using the sprung mass vertical speeds of wheels FR, RL, and RR. In step S3, the first pitch rate calculation unit 116 performs pitch rate calculation C using the sensor estimate calculated using the sprung mass vertical speeds of wheels RL, RR, and FL. In step S4, the first pitch rate calculation unit 116 performs pitch rate calculation D using the sensor estimate calculated using the sprung mass vertical speeds of wheels RR, FL, and FR.

[0070] In step S6, the estimated value verification / correction unit 705 calculates the result obtained by subtracting the calculation result derived from the wheel speed calculated by the second pitch rate calculation unit 117 from the calculation result A calculated in step S2. In step S7, the estimated value verification / correction unit 705 calculates the result obtained by subtracting the calculation result derived from the wheel speed calculated by the second pitch rate calculation unit 117 from the calculation result B calculated in step S3. In step S8, the estimated value verification / correction unit 705 calculates the result obtained by subtracting the calculation result derived from the wheel speed calculated by the second pitch rate calculation unit 117 from the calculation result C calculated in step S3. In step S9, the estimated value verification / correction unit 705 calculates the result obtained by subtracting the calculation result derived from the wheel speed calculated by the second pitch rate calculation unit 117 from the calculation result D calculated in step S5.

[0071] In step S10, the estimated value verification / correction unit 705 determines whether the difference between the four results calculated in steps S6 to S9 is smaller than a predetermined threshold. If the difference between the four calculated results is small, the process proceeds to step S18, where the estimated value verification / correction unit 705 adopts the four estimated results. After that, the process returns to step S2 to continue determining whether the four estimated results can be used (whether the difference is smaller than a predetermined threshold).

[0072] If the difference between the four results calculated in step S10 is large, the process proceeds to step S11. In step S11, the estimated value verification / correction unit 705 determines whether the difference between the calculation results calculated in step S6 is smaller than a predetermined threshold. If the difference is not smaller than a predetermined threshold, in step S12, the estimated value verification / correction unit 705 determines whether the difference between the calculation results calculated in step S7 is smaller than a predetermined threshold. If the difference is not smaller than a predetermined threshold, in step S13, the estimated value verification / correction unit 705 determines whether the difference between the calculation results calculated in step S8 is smaller than a predetermined threshold. If the difference is not smaller than a predetermined threshold, in step S14, the estimated value verification / correction unit 705 determines whether the difference between the calculation result D calculated in step S9 is smaller than a predetermined threshold. If the difference in step S14 is not smaller than a predetermined threshold, the process returns to step S2.

[0073] In step S15, if the difference calculated in step S6 in step S11 is smaller than a predetermined threshold, the sensor estimate calculated using the sprung vertical velocities of wheels FL, FR, and RL in pitch rate calculation A of step S2 is adopted. In step S16, the sprung vertical velocities of the remaining wheel RR are calculated based on this, and in step S17, the calculated sprung vertical velocities of wheel RR are stored in the dataset. Note that the calculation flow for wheel FL (steps S18 to S20), wheel FR (steps S21 to S23), and wheel RL (steps S24 to S26) are the same as the flow from steps S15 to S17, so the description is omitted.

[0074] In step S27, the driving data storage unit 120 performs counter processing. In step S28, if the counter processing count from step S27 exceeds 1000, in step S29, the driving data storage unit 120 sends the updated dataset to the learning management server 102, and in step S30, the dataset in the driving data storage unit 120 is deleted, and the process returns to step S1 to repeat the above flow. If the counter processing count from step S28 does not exceed 1000, the process returns to step S2 to repeat the above flow.

[0075] (Fourth embodiment) (Figure 10) The fourth embodiment has a configuration that allows for retraining (parameter updating) in a virtual environment using only flags indicating strengths and weaknesses in driving data. In contrast to the previously described embodiment, which improved the estimation accuracy of sprung mass vertical speed and piston speed output by the vehicle state estimation unit 112 while driving, the fourth embodiment reproduces the driving environment in a virtual environment and improves the estimation accuracy of sprung mass vertical speed and piston speed in the virtual environment.

[0076] Specifically, the accuracy status is transferred to the learning management server 102 along with road information (including road profiles) acquired by existing sensors such as the GPS sensor 504 and the stereo camera 604. The learning management server 102 then reproduces this data through simulation to create training data, thereby performing the learning process.

[0077] This embodiment is characterized in that, in addition to the second embodiment, a vehicle behavior simulator 1103 is added to the learning management server 102, and a map information server 1104 used to reproduce real terrain and road shapes in the virtual environment is also included in the configuration. The map information handled by the map information server 1104 may be, for example, the open-source OSM (Open Street Map), or other paid or free map data. This map data may include elevation / gradient information, and as data collection progresses, information such as road profiles may be applied.

[0078] Furthermore, by adding information from the weather information server 503 to the map information server 1104, the reproducibility of the real environment in the virtual environment will improve. Therefore, the learning management server 102 learns weight parameters based on the second sensor information acquired by the second sensors (weather information server 503, map information server 1104) mounted on the vehicle 101 and the virtual environment information learned outside the vehicle 101. In this way, by utilizing the vehicle behavior simulator 1103 of the learning management server 102 in an advanced virtual environment, it becomes possible to improve the estimation accuracy of the sprung mass vertical speed and piston speed output by the vehicle state estimation unit 112, even without actual vehicle measurements.

[0079] According to the embodiments of the present invention described above, the following effects and advantages are achieved.

[0080] (1) A suspension system that calculates using machine learning by associating sensor information acquired from multiple sensors 106 provided on the vehicle 101 with state information of the suspension 109 provided on the vehicle 101 comprises a weight parameter storage unit 114 that stores weight parameters calculated by the machine learning, and a vehicle state estimation unit 112 that outputs an estimation result of the state of the suspension 109 based on the sensor information and weight parameters. A first vehicle behavior calculation unit 116 that calculates a first physical value of a physical quantity related to the behavior of the vehicle 101 based on the output estimation result, a second vehicle behavior calculation unit 117 that calculates a second physical value of a physical quantity based on the sensor information, and an estimation accuracy verification unit 118 that outputs the estimation accuracy of the state of the suspension 109 by the vehicle state estimation unit 112 by comparing the first physical value and the second physical value. A driving data management unit 119 that instructs the learning management server 102 to learn weight parameters based on the output result of the estimation accuracy by the estimation accuracy verification unit 118. This approach allows us to identify the neural network's weaknesses in vehicles that do not have suspension sensors, thereby supporting its learning process.

[0081] (2) The vehicle 101 has four vehicle state estimation units 112, each of which is linked to one of the four wheels of the vehicle 101, and a selector 703 that selects three of the four estimation results output from the four vehicle state estimation units 112. In this way, the vehicle state estimation unit 112 can estimate the state of the suspension 109 using only the estimation results for three wheels.

[0082] (3) The first vehicle behavior calculation unit 116 calculates four types of first physical values ​​from the three estimation results selected by the selector 703. The estimation accuracy verification unit 118 adopts the four estimation results output from the four vehicle state estimation units 112 as the state of the suspension if the four difference values ​​based on the comparison of each of the four types of first physical values ​​with the second physical value are smaller than a predetermined threshold. The estimation accuracy verification unit 118 adopts the three estimation results selected by the selector 703 to calculate the first physical value as the state of the suspension 109 if the difference value based on the comparison of any one of the four calculated first physical values ​​with the second physical value is larger than a predetermined threshold. The first vehicle behavior calculation unit 116 recalculates the four types of first physical values ​​from the three estimation results selected by the selector 703. In this way, the estimation accuracy can be judged taking into account the low-accuracy vehicle state estimation results.

[0083] (4) The estimation accuracy verification unit 118 calculates the first physical value for the estimation results not selected by the selector 703 using the three estimation results selected by the selector 703 and the second physical value. In this way, the first physical value can be calculated even for estimation results not selected by the selector 703.

[0084] (5) The estimation accuracy output by the estimation accuracy verification unit 118 is presented to the user through the display unit 110. This makes it possible to visualize the estimation accuracy of the state of the suspension 109.

[0085] (6) The first and second physical values ​​are the pitch rate or roll rate of the vehicle 101. In this way, the first and second physical values ​​can be determined based on multiple vehicle behavior data.

[0086] (7) The learning management server 102 learns weight parameters based on the second sensor information acquired by the second sensor mounted on the vehicle 101 and the virtual environment information learned outside the vehicle 101. This improves the estimation accuracy of the sprung mass vertical velocity and piston velocity in the virtual environment.

[0087] (8) The learning management server 102 learns the weight parameters based on the computer installed in the vehicle 101. In this way, the weight parameters can be updated without installing the learning management server 102 on the vehicle 101 side.

[0088] It should be noted that the present invention is not limited to the embodiments described above, and various modifications and combinations of other configurations can be made without departing from the spirit of the invention. Furthermore, the present invention is not limited to having all the configurations described in the embodiments described above, and may also include configurations in which some of those configurations are omitted. [Explanation of symbols]

[0089] 101 vehicles 102 Learning Management Server 103 Network 104 Suspension ECU 105 Vehicle ECU 106 Sensors 107 Sensor ECU 108 CAN 109 Active Suspension 110 Display section 111 Interface (I / F) 112 Vehicle condition estimation unit 113 Weight Data Management Unit 114 Weight parameter storage unit (memory unit) 115 Suspension control value calculation unit 116 First pitch rate calculation unit 117 Second Pitch Rate Calculation Unit 118 Estimation Accuracy Verification Unit 119 Driving Data Management Department 120 Driving data storage unit 201 Input layer elements of a neural network 202 Hidden Layer Elements of a Neural Network 203 Output layer elements of a neural network 204 Time series data 205 Sensor Data 206 windows 207 Instantaneous Sensor Data 401 Measuring device screen 402 Wheel speed data (CAN system data) 403 Front and rear acceleration data (CAN system data) 404 Piston speed data (estimated vehicle condition) 405 Pitch Rate Data (Vehicle Behavior Physical Values) 406 File Output Button 407 Terminal screen 408 Pitch rate derived from neural networks 409 Non-neural network-derived pitch rate 410 Accuracy judgment result 411 Instrument Panel 412 Vehicle Behavior Estimation Results 413 Result Display Lamp 414 Odometer 503 Weather Information Server 504 GPS Sensor 505 Stereo Camera 506 Wheel speed reliability determination unit 601 Road surface μ estimation section 703 Selector 705 Estimate Verification / Correction Unit 801 Vehicle condition estimation cycle 802~805 Selection Signal 1103 Vehicle Behavior Simulator 1104 Map Information Server

Claims

1. A suspension system that correlates sensor information obtained from multiple sensors installed in a vehicle with the state information of the suspension installed in the vehicle and calculates using machine learning, A weight parameter storage unit that stores the weight parameters calculated by the aforementioned machine learning, A vehicle state estimation unit that outputs an estimation result of the suspension state based on the sensor information and the weight parameters, A first vehicle behavior calculation unit calculates a first physical value of a physical quantity relating to the behavior of the vehicle based on the outputted estimation result, A second vehicle behavior calculation unit calculates a second physical value of the physical quantity based on the sensor information, The estimation accuracy verification unit outputs the estimation accuracy of the suspension state by the vehicle state estimation unit by comparing the first physical value and the second physical value, The system includes a driving data management unit that instructs the learning management server to learn the weight parameters based on the output result of the estimation accuracy verification unit. Suspension system.

2. A suspension system according to claim 1, The vehicle has four vehicle state estimation units, each of which is in conjunction with one of the four wheels of the vehicle. The system includes a selector that selects three of the four estimation results output from the four vehicle state estimation units. Suspension system.

3. A suspension system according to claim 2, The first vehicle behavior calculation unit calculates four types of first physical values ​​from the three estimation results selected by the selector, The estimation accuracy verification unit, if the four difference values ​​obtained by comparing each of the four first physical values ​​with the second physical value are smaller than a predetermined threshold, adopts the four estimation results output from the four vehicle state estimation units as the state of the suspension. The estimation accuracy verification unit, if the difference value based on the comparison with the second physical value for any one of the four first physical values ​​calculated is smaller than a predetermined threshold, adopts the three estimation results selected by the selector to calculate the first physical value as the suspension state. If the difference value based on the comparison between the four calculated first physical values ​​and the second physical values ​​is greater than a predetermined threshold, the first vehicle behavior calculation unit recalculates the four first physical values ​​from the three estimation results selected by the selector. Suspension system.

4. A suspension system according to claim 3, The estimation accuracy verification unit calculates the estimation result that was not selected by the selector using the three estimation results selected by the selector and the second physical value. Suspension system.

5. A suspension system according to claim 1, The estimation accuracy output by the estimation accuracy verification unit is presented to the user through the display unit. Suspension system.

6. A suspension system according to claim 1, The first physical value and the second physical value are the pitch rate or roll rate of the vehicle. Suspension system.

7. A suspension system according to claim 1, The learning management server learns the weight parameters based on the second sensor information acquired by the second sensor mounted on the vehicle and the virtual environment information learned outside the vehicle. Suspension system.

8. The suspension system according to claim 4, The learning management server learns the weight parameters based on the computer installed in the vehicle. Suspension system.

Images