Suspension system

Abstract

A suspension system in which sensor information and state information of a suspension are associated and calculation is performed by machine learning, the suspension system being provided with: a weighted parameter storage unit; a vehicle state estimation unit; a first vehicle behavior calculation unit that calculates a first physical value on the basis of the estimation result; a second vehicle behavior calculation unit that calculates a second physical value on the basis of the sensor information; an estimated accuracy verification unit that compares the first physical value with the second physical value and outputs an estimated accuracy of the state of the suspension; and a travel data management unit that instructs learning of a weighting parameter on the basis of the result of the output of the estimated accuracy.

Description

Technical Field

[0001] The 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 the suspension on the spring according to the road conditions and the operating conditions, the suspension is controlled by calculating the control value by detecting sensor data obtained from sensors that detect the piston speed and the up and down speed of the spring of the four-wheel suspension of the vehicle. Regarding the sensor function, from the perspective of cost reduction and appropriate labor reduction, there is a technology that realizes sensorless by installing a neural network, which realizes the logic and estimation function of estimating the sensor value based on CAN (Controllerarea Network) data.

[0003] However, the driving roads are diverse, and it is not easy to replace the sensor function on all roads with the above sensor value estimation technology. In particular, when a neural network is applied, it may be difficult to continuously replace the sensor with the learned neural network installed at the factory. Therefore, a continuous learning system is required in the suspension control system to improve the comprehensiveness of the road.

[0004] Furthermore, in the field of image recognition where machine learning is sometimes applied, it is easy for humans to identify images with low recognition rates by simply performing additional learning specifically on those images. In contrast, in sensor value estimation technology that uses a sensorless neural network, it is difficult to identify road conditions that are not an area of ​​expertise because there are no sensor values ​​to serve as training data.

[0005] For example, Patent Document 1 listed below discloses a shock absorber control system that receives feedback data on the behavior of a vehicle, can execute a machine learning algorithm, and can control the characteristics of the shock absorber with high responsiveness and robustness. Prior art literature Patent Literature

[0006] Patent Document 1: Japanese Patent Application Publication No. 2021-17168 Summary of the invention Problem that the invention aims to solve

[0007] According to the prior art, in order to achieve sensorless operation in which the suspension state is estimated based on other sensor information instead of directly detecting the suspension state with a sensor, it is necessary to further extract the weak states of sensor value estimation and continuously learn to improve the accuracy of the machine learning model. In view of this, the purpose of the present invention is to provide a sensorless suspension system that extracts the weak states of a neural network to assist learning when a neural network replaces the sensor function. Technical means of solving problems

[0008] A suspension system that associates sensor information obtained from multiple sensors of a vehicle with state information of a suspension of the vehicle and performs calculations through machine learning, the suspension system comprising: a weighted parameter storage unit that stores weighted parameters calculated through the machine learning; a vehicle state estimation unit that outputs an estimation result of the state of the suspension based on the sensor information and the weighted 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 state of the suspension 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 weighted parameters based on the output result of the estimation accuracy of the estimation accuracy verification unit. Effects of the Invention

[0009] The present invention provides a sensorless suspension system that extracts the weak states of a neural network to assist learning when the suspension state is estimated based on other sensor information without directly detecting the suspension state using a sensor. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] Figure 1 This is a functional block diagram of a suspension system according to a first embodiment of the present invention. Figure 2A It is represented by the composition of the neural network Figure 1 FIG. 2 is a diagram of a vehicle state estimation unit. Figure 2B It is a relationship diagram of the input and output data sets when learning a neural network. Figure 3A This is a flowchart related to vehicle state estimation and learning requirement notification on the vehicle side. Figure 3B This is a flowchart related to updating weighted data on the vehicle side. Figure 3C This is a flowchart related to weighted data learning on the server side. Figure 3D This is a flowchart related to abnormality detection on the server side. Figure 4A This is an example of a terminal screen for visualizing data related to CAN data, vehicle behavior estimation results, and vehicle behavior physical value calculation result data. Figure 4B This is another example of a terminal screen that is envisioned to visualize vehicle data. Figure 4CThis is an example of when the driver can confirm the result of the vehicle behavior estimation. Figure 5 This is a functional block diagram of a suspension system according to a second embodiment of the present invention. Figure 6 yes Figure 5 An example of the internal structure of the wheel speed reliability determination unit. Figure 7 This is a functional block diagram of a suspension system according to a third embodiment of the present invention. Figure 8 This is an example of a control signal of the selector in the third embodiment. Fig. 9 This is a flowchart of the third embodiment. Fig.10 This is a functional block diagram of a suspension system according to a fourth embodiment. DETAILED DESCRIPTION

[0011] Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings. The following description and drawings are examples for illustrating the present invention, and appropriate omissions and simplifications have been made for clarity of the description. The present invention may also be implemented in various other ways. Unless otherwise specified, each constituent element may be singular or plural.

[0012] In order to facilitate understanding of the invention, the position, size, shape, range, etc. of each component shown in the drawings may not represent the actual position, size, shape, range, etc. Therefore, the present invention is not necessarily limited to the position, size, shape, range, etc. disclosed in the drawings.

[0013] (First Embodiment of the Present Invention and Overall Configuration of the Device) ( Figure 1 ) A sensorless suspension system that controls damping force to improve the ride comfort of an automobile without a vehicle-mounted suspension sensor is provided, and includes a vehicle 101, a learning management server 102, and a network environment 103. The vehicle 101 includes 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 the suspension 109), a display unit 110, and an interface (I / F) 111.

[0014] The suspension ECU 104 includes a vehicle state estimation unit 112 , a weighted data management unit 113 , a weighted 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 driving data management unit 119 , and a driving data storage unit 120 .

[0015] When the vehicle 101 is running, the suspension 109 suppresses the vibrations felt by the driver and passengers sitting on the seats from the vibrations transmitted from the road surface via the tires. The suspension ECU 104 controls the damping force characteristics of the suspension 109, and the suspension control value calculation unit 115 of the suspension ECU 104 derives the control value 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 network environment 103. The ECU 105 transmits the information of the learning management server 102 transmitted from the network environment 103 to the CAN 108. Thus, in the suspension ECU 104, the information is shared with each functional unit via the CAN 108. In addition, the sensor ECU 107 acquires information from the sensors 106a ... 106b, transmits it to the CAN 108, and shares the information with each functional unit of the vehicle 101.

[0017] The functional units of the suspension ECU 104 are described. The suspension ECU 104 obtains information from the CAN 108 or outputs information to the CAN 108 via the interface 111. The weighted data management unit 113 determines the calculation specifications for the estimation of the vehicle state estimation unit 112 by reading the weighted parameters stored in the weighted parameter storage unit 114 (weighted parameter storage unit). In addition, the weighted data management unit 113 obtains the weighted parameters learned from the learning management server 102 via the CAN 108 and updates the weighted parameters.

[0018] The vehicle state estimation unit 112 refers to the sensor information that can be obtained as specific information by the sensors 106a...106b in the data that is normally transmitted to the CAN 108. In addition to detecting the rotation speed (wheel speed) of one of the four wheels of the vehicle 101, the sensors 106a...106b also detect the front and rear acceleration, left and right acceleration, up and down acceleration, yaw rate, roll rate, pitch rate, etc. in the vehicle behavior system, and transmit them to the CAN 108 as sensor information. Figure 1 Although the diagram shows two sensors 106a…106b, there are actually multiple sensors measuring the above physical values. In addition, when the vehicle 101 equipped with the vehicle state estimation unit 112 is a learning vehicle, an acceleration sensor for the suspension is mounted on the vehicle, and its data is also transmitted to the CAN 108.

[0019] The vehicle state estimation unit 112 acquires sensor information from each sensor 106a ... 106b. In addition, the vehicle state estimation unit 112 acquires the value of the weighted parameter read from the weighted parameter storage unit 114 by the weighted data management unit 113 as data related to the vehicle state during driving. The vehicle state estimation unit 112 outputs the estimation result of the physical quantity related to the suspension control of each of the four wheels based on the acquired sensor information and the acquired weighted parameter.

[0020] The suspension control value calculation unit 115 calculates a control value for controlling the damping force of the suspension 109 based on the physical quantity estimated by the vehicle state estimation unit 112. In this way, by performing calculations using the estimated control value, it is not necessary to provide an acceleration sensor for measuring the sprung vertical velocity or the stroke velocity of the piston as in the past, which can contribute to cost reduction.

[0021] Next, the verification of the estimation accuracy of the vehicle state estimation unit 112 is described. First, the vehicle state estimation unit 112 transmits the estimated sprung up-and-down speeds of the suspension of the four wheels (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 transmitted sprung up-and-down speeds of the four wheels. On the other hand, the wheel speed data (sensor information) of the four wheels contained in the CAN 108 obtained from the multiple sensors 106a...106b possessed by the vehicle 101 is transmitted 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 transmitted wheel speed data of the four wheels.

[0022] The present invention focuses on the fact that the pitch rate can be calculated by either the sprung up-and-down velocity or the wheel speeds of the four wheels, and aims to verify the estimation accuracy of the first pitch rate from the estimation result by using the second pitch rate as a sample. The two pitch rates derived by the calculation based on the sprung up-and-down velocity and the wheel speeds of the four wheels are compared, and if they are the same value, it is determined that the estimation accuracy of the sprung up-and-down velocity estimated by the vehicle state estimation unit 112 is good.

[0023] The estimation accuracy verification unit 118 compares the two pitch rates (the first physical value and the 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 value, the estimation accuracy verification unit 118 determines that there is no problem with the estimation result output by the vehicle state estimation unit 112. On the other hand, if the difference between the two pitch rates is greater than a predetermined threshold value, the estimation accuracy verification unit 118 determines that there is a problem with the estimation result output by the vehicle state estimation unit 112, and determines that the output accuracy of the vehicle state estimation unit 112 is in a poor state. If the estimation accuracy verification unit 118 determines that the output accuracy of the vehicle state estimation unit 112 is in a poor state, a problematic flag is assigned to the time series data set input to the vehicle state estimation unit 112 via the CAN 108.

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

[0025] In this way, by comparing the pitch rate calculated based on the sprung up and down velocity estimated by the neural network (the pitch rate calculated by the first pitch rate calculation unit 116) and the pitch rate calculated based on the wheel speed information from the CAN 108 not passing through the neural network (the pitch rate calculated by the second pitch rate calculation unit 117), it is determined whether the estimation accuracy of the sprung up and down velocity by the vehicle state estimation unit 112 is low.

[0026] The weighting parameters used in the neural network of the vehicle state estimation unit 112 are parameters obtained by generating training data and learning in a verification vehicle or the like equipped with an acceleration sensor for suspension, which is different from the vehicle 101 .

[0027] The estimation accuracy verification unit 118 transmits the data set to which the problematic flag is assigned to the driving data management unit 119. The driving data management unit 119 stores the transmitted data set in the driving data storage unit 120. By performing such a process, the data set that the vehicle state estimation unit 112 is not good at is stored in the driving data storage unit 120. The driving data management unit 119 instructs the learning management server 102 to update the weighted parameters regarding the driving environment of the vehicle 101 based on the stored data set that is not good at. The learning management server 102 performs learning for updating the weighted parameters based on the instruction of the driving data management unit 119 based on the output of the estimation accuracy verification unit 118. In this way, the learning management server 102 learns the weighted parameters based on the output result 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 hierarchical neural network composed of three layers, in which the elements of the input layer (number of elements i) 201, the hidden layer (number of elements j) 202, and the 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 of the wheel speed time series and an input element group 201b of the vehicle behavior time series, and the state of the suspension 109 is estimated by these two element groups.

[0029] The number of elements in the hidden layer element group 202 is generally determined by the number of elements in the input layer element group 201 and the output layer element group 203, but is set to a number that maximizes the accuracy of the state estimation of the neural network. The estimated instantaneous value of the piston speed or the sprung vertical speed of the vehicle-mounted suspension 109 is output from the output layer element group 203. Figure 2A In FIG. 1 , a multilayer perceptron is shown as an example of a neural network, but a recurrent neural network that 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 combined by weight W1ij (i=1~I, j=1~J), and each element of the hidden layer element group 202 and the output layer element 203 are combined by weight W2jk (j=1~J, k=1). These weighted information (weighted parameters) are represented by the determinants of weights W1ij and weights W2jk. The weighted parameters are obtained in advance through machine learning and stored in the weighted parameter storage unit 114. The vehicle state estimation unit 112 obtains the weighted parameters from the weighted parameter storage unit 114 via the weighted parameter management unit 113, thereby implementing calculations using a neural network. In addition, in this example, the simplest hidden layer element group 202 is a 1-layer fully connected neural network, but it is not limited to this.

[0031] ( Figure 2B ) The figure in the upper section is a figure that takes the wheel speed in the information of CAN 108 as an example. The learning of the neural network is performed by driving a verification vehicle equipped with an acceleration sensor for measuring the piston speed and the sprung vertical speed for controlling the suspension 109 as described above, and the learning is implemented using the time series data 204 and sensor data (time series data of the vehicle-mounted sensor) 205 of the data contained in the CAN of the verification vehicle.

[0032] Specifically, the discrete values ​​obtained by sampling the data contained in the window 206 in the time series data 204 are set in the input layer element group 201 ( Figure 2A ). Then, in the output layer element 203 of the neural network ( Figure 2A) and the corresponding sensor data instantaneous value 207 is set in the window 206. Regarding the relationship between the window 206 and the sensor data instantaneous value 207, if the width of the window 206 is set to, for example, 1 second, the sensor data instantaneous value 207 is set to the specification estimated based on the driving history 1 second ago. Therefore, if the sampling interval is set to, for example, 20 m / second, the sampling points in the window 206 are 50 (=1 second ÷ 20 m / second), and the combination of the 50 data groups and the sensor data instantaneous value 207 is defined as a learning data set. Then, by sliding the window 206 and the sensor data instantaneous value 207 in the time direction every 20 m / second, the data set is enlarged n times. In this way, learning of a neural network using many data sets can be performed.

[0033] ( Figure 3A ) The estimation of the vehicle state and the notification of the need for learning in the vehicle 101 will be described. First, in step S101, the vehicle state estimation unit 112 determines whether the vehicle 101 is traveling. If the vehicle 101 is traveling, the process proceeds to step S102 and step S103. If not, the process ends.

[0034] In step S102, the second pitch rate calculation unit 117 calculates the pitch rate based on the wheel speed. On the other hand, in step S103, the vehicle state estimation unit 112 estimates the sprung vertical velocity. In step S104, the first pitch rate calculation unit 116 calculates the pitch rate based on the sprung 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 step S102 and step S104. In step S106, as an example of a predetermined threshold, if the estimation accuracy (error) is greater than 1 degree, the process proceeds to step S106, and if it is less than 1 degree, the process returns to step S101. In addition, although the error of 1 degree is used as the threshold value for determining the estimation accuracy using the pitch rate, it is not limited to this value.

[0036] In step S107, a learning-needed flag is assigned to the data that is determined by the estimation accuracy verification unit 118 to have an estimation accuracy (error) greater than 1 degree and poor estimation accuracy. In step S108, the driving data management unit 119 writes the driving data assigned with the learning-needed flag to the driving data storage unit 120. In step S109, the driving data management unit 119 counts the number of learning-needed flags.

[0037] In step S110, the driving data management unit 119 determines whether the number of learning-required flags stored in the driving data storage unit 120 is greater than 1000 counts. The 1000 counts become a condition for transmitting the data stored in the driving data storage unit 120 to the learning management server 102. In the case where the number of learning-required flags is greater than 1000 counts, the process proceeds to step S111, and the driving data to which the learning-required flags are assigned is transmitted to the driving data server, i.e., the learning management server 102. In addition, the reference number of learning-required flags is set to 1000 as a condition for transmitting data to the learning management server 102, but it may not be limited to this value.

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

[0039] In step S202, the weighted data management unit 113 stores the acquired weighted parameter data in the weighted parameter storage unit 114. In step S203, the weighted data management unit 113 determines whether the storage of the weighted parameters in the storage unit 114 is completed. If the storage of the weighted parameters is completed, the process proceeds to step S204, and if not, step S203 is repeated.

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

[0041] In step S206, the estimation accuracy verification unit 118 determines whether the estimation output of the vehicle state estimation unit 112 is abnormal. If there is an abnormality, the process proceeds to step S207. If there is no abnormality, the process from step S201 is repeated. In step S207, the estimation accuracy verification unit 118 assigns a learning-needed flag to the driving data with abnormality. In step S208, the estimation accuracy verification unit 118 notifies the vehicle state estimation unit 112 of the abnormality of the estimation output. Figure 4B After the developer of the vehicle state estimating unit 112 of the terminal or the inspector who performs maintenance inspection or the like views the terminal screen 407 shown, the flow from step S201 is repeated.

[0042] ( Figure 3C ) The learning of the weighted parameters in the learning management server 102 is described. In step S301, the learning management server 102 receives the driving data from the vehicle 101. Next, in step S302, the learning management server 102 performs the learning of the neural network. Next, in step S303, the learning management server 102 stores the weighted parameters. Next, in step S304, the learning management server 102 transmits the weighted parameters to the weighted data management unit 113, and the process ends.

[0043] ( Figure 3D ) In the learning management server 102, this is a process corresponding to the detection of an abnormality in the estimation of the vehicle state. In step S401, the learning management server 102 determines whether there is an abnormality notification about the estimation of the vehicle state from the estimation accuracy verification unit 118. If there is an abnormality notification, it proceeds to step S402, and if not, it repeats step S401. 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 it is determined that an abnormality has occurred, the process proceeds to step S404, and if it is determined that no abnormality has occurred, the process proceeds to step S406.

[0045] In step S404, the latest weighted parameter data stored in the learning management server 102 is saved. Next, in step S405, learning is performed again in the learning management server 102, and the process of step S402 is repeated. In step S406, since it is determined in step S403 that no abnormality has occurred, the weighted parameters are transmitted from the learning management server 102 to the vehicle 101. In addition, the re-learning is described in terms of saving the latest weighted parameters and learning with new initial values, but additional learning with the latest weighted parameters as initial values ​​may also be performed.

[0046] ( Figure 4A ) In the measuring device screen 401 in the driving test, which visualizes the data related to the CAN system data, the vehicle behavior estimation result, and the calculation result data of the vehicle behavior physical value, the upper stage shows the various data 402 and 403 (CAN system data) referenced from the CAN 108, the middle stage shows the vehicle behavior estimation result 404 output by the vehicle state estimation unit 112, and the lower stage shows the time series data 405 of the vehicle behavior physical value. In addition, for example, it is assumed that there is a file output button 406, which is used to output and store the data set as text data in the form of a file in CSV format so as to be input to the neural network. In addition, the button 406 can be a physical button provided by the device, or it can be selected by touch operation, or it can be selected by a pointing device, etc. In addition, it can also be selected by touch operation of a finger in combination with the touch sensor function.

[0047] ( Figure 4B ) The pitch rate 408 calculated by the first pitch rate calculation unit 116 based on the output of the neural network as the output of the vehicle state estimation unit 112 and the pitch rate 409 calculated by the second pitch rate calculation unit 117 not based on the output of the neural network as the wheel speeds of the four wheels are presented in parallel on the terminal screen 407. Furthermore, the accuracy determination result 410 is presented in parallel with the two pitch rates, and the accuracy determination result 410 is presented in a high level display when the difference between the two pitch rates is greater than a predetermined value, and in a low level display when the difference between the two pitch rates is less than the predetermined value.

[0048] ( Figure 4C ) Figure 4C 4 is a diagram showing a situation where a user can check the result of vehicle behavior estimation through the display unit 110. Here, the user is taken as an example of a driver. The display content of the instrument panel 411 includes the vehicle behavior estimation result 412. For example, the light 413 is a light that lights up when the accuracy determination result 410 is a high level and the error is large. In addition, the display instrument 414 shows the accuracy determination result 410 ( Figure 4B ) is a graph of the distance of a high-level driving situation, or the ratio of the time of a high-level to the driving distance and driving time.

[0049] Therefore, the estimated accuracy output by the estimated accuracy verification unit 118 is prompted to users such as the developer of the vehicle state estimation unit 112, the inspector who holds the terminal for maintenance inspection, the driver of the vehicle 101, etc. through the display unit 110, thereby confirming whether the accuracy of the suspension system to which the neural network is applied is good and being able to identify low-precision states.

[0050] (Second embodiment) ( Figure 5 ) The second embodiment is an embodiment that focuses on the different reliability of the calculation of the pitch rate from the wheel speed depending on the weather or road surface conditions. As a functional unit that has been changed and added to the first embodiment, the suspension system has a weather information server 503, a GPS sensor 504, a stereo camera 505 (stereo camera), 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 the outside of the vehicle 101, such as the weather information server 503, the vehicle-mounted stereo camera 505, and the GPS sensor 504. Therefore, the purpose is to focus on the reference road surface state, or the reliability of the pitch rate changes according to the road surface state, and reflect the reliability determination result, so as to extract the weak state of the neural network.

[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 wheel speed reliability determination unit 506 refers to information from the stereo camera 505, information from the GPS sensor 504, information from the weather information server 503, information on the condition of the tires mounted on the vehicle 101, etc. The information is referred to and input to the road surface μ estimation unit 601.

[0052] The road surface μ estimation unit 601 has a function of detecting that the road surface on which the vehicle is generally traveling is in a slippery state, such as a state of water accumulation due to rainfall, or a state of low road surface μ due to snow accumulation or ice, based on information from the stereo camera 505. In this regard, even if the stereo camera 505 is not used, as long as the same detection can be performed, the position information of the vehicle that can be obtained from the GPS sensor 504 can be used to inquire the weather information server 503 to estimate the road surface condition.

[0053] In addition, stably determined elements other than the elements that change according to the surrounding conditions as described previously, such as information on sand or coarse-grained dirt roads, can also be managed in association with the map, and information from the GPS sensor 504 can be used to detect whether the road surface is slippery or whether the wheels are prone to spinning.

[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 information of the vehicle 101, and the information of the condition (e.g., grip) of the tires mounted on the vehicle 101, and outputs the calculation result to the estimation accuracy verification unit 118. Generally, the road surface μ is set to be about 0.8 for a paved dry road surface, 0.4 to 0.6 for a paved wet road surface, 0.2 to 0.5 for a snowy road, and 0.1 to 0.2 for an icy road. The total friction coefficient μ' is estimated based on the road surface μ and the condition of the tires, and the physical quantity of the slip is calculated by integrating it with the weight N of the vehicle 101.

[0055] Estimation accuracy verification unit 118 ( Figure 5 ) The estimation accuracy is verified 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. When the reliability of the wheel speed information determined by the wheel speed reliability determination unit 506 is high, the same processing as the first embodiment is performed. However, when the reliability of the wheel speed information is low, the estimation of the vehicle state by the neural network is continued, or the process is switched to another alternative means.

[0056] The road surface μ is an example, and when the tire condition cannot be obtained, the calculation of μ' can be simplified. The weight N of the vehicle 101 can also be calculated by adding the weight of the passengers and the load to the weight of the vehicle itself, thereby improving accuracy.

[0057] (Third embodiment) ( Figure 7 ) Compared with the first embodiment, the suspension ECU 104 of this embodiment further includes a selector 703 and an estimated value verification / correction unit 705. In addition, the vehicle state estimation units 112a to 112d are provided for each of the four wheels and cooperate with each other. The vehicle state estimation units 112a to 112d are provided for FL (front left), 112b for FR (front right), 112c for RL (rear left), and 112d for RR (rear right).

[0058] The selector 703 selects three estimation results from the four vehicle state estimation units 112a to 112d to make different combinations. The first pitch rate calculation unit 116 calculates four pitch rates (first physical values) based on the three estimation results selected by the selector 703. This is because if the vehicle 101 is a rigid body, the first physical value (pitch rate) can be calculated using the estimation results (sprung vertical speed) of the three wheels, and thus the four first physical values ​​can be calculated using the selection method of the three estimation results.

[0059] The reason for such calculation is that, for example, when the sprung vertical velocity of only one wheel is low-precision in the comparison of the estimated value verification / correction unit 705, the pitch rate calculated using the sprung vertical velocity of the other three wheels excluding the one wheel is likely to be substantially consistent with the pitch rate from the wheel speed, and the three pitch rates including the one wheel are likely to be inconsistent with the pitch rate from the wheel speed. Thus, the estimated value verification / correction unit 705 determines that the sprung vertical velocity estimated values ​​of the three wheels used in the calculation of the pitch rate consistent with the pitch rate from the wheel speed are high-precision, transmits the corresponding sprung vertical velocity estimated values ​​of the three wheels to the suspension control value calculation unit 115, and excludes the remaining one wheel.

[0060] In addition, as described above, if there is data of three wheels, the pitch rate can be calculated, so the estimated value verification / correction unit 705 uses the pitch rate of the correct answer from the wheel speed and the sprung up and down speeds of the three wheels to perform inverse calculation to determine the sprung up and down speed of one wheel with low accuracy. The estimated value verification / correction unit 705 transmits the inversely calculated low-precision sprung up and down speed of one wheel to the suspension control value calculation unit 115. That is, the estimated value verification / correction unit 705 that performs the estimation accuracy verification verifies four types of estimation results and derives the estimation result by inverse calculation.

[0061] In this way, if a high-precision pitch rate of one system is obtained, the estimation result of the sprung vertical speed of the three systems used to derive the pitch rate can be determined to be high-precision, and the estimation result of the sprung vertical speed of the remaining one system can be determined to be low-precision. In the case of the low-precision sprung vertical speed, the pitch rate calculated by the second pitch rate calculation unit 117 and the sprung vertical speed of the three systems are used for inverse calculation, and can be adopted as the estimation result, and can be processed as training data for the neural network that realizes the vehicle state estimation unit 112 implemented by the learning management server 102.

[0062] In addition, when all four difference values ​​based on the comparison between the four first physical values ​​and the second physical value are smaller than the predetermined threshold value, 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 transmits the estimated results to the suspension control value calculation unit 115 (described later). Fig. 9 Step S10, step S18).

[0063] The estimated value verification / correction unit 705 transmits the input data group of one of the vehicle state estimation units 112a~112d that outputs the sprung up and down speed set to a low precision and the sprung up and down speed calculated by the estimated value verification / correction unit 705 to the driving data management unit 119, and the driving data management unit 119 stores the input data set and the sprung up and down speed in the driving data storage unit 120.

[0064] After storing a certain number (eg, 1000) of data sets in the traveling data storage unit 120 , the traveling data management unit 119 transmits the data sets to the learning management server 102 via the interface 111 , using the count as a transmission condition.

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

[0066] ( Figure 8 ) The selector 703 is described below. Figure 8 , it is shown how the selector 703 sets four stages based on the vehicle state estimation cycle 801 which is the operation cycle of the vehicle state estimation unit 112, and selects three from the FL selection signal 802, the FR selection signal 803, the RL selection signal 804, and the RR selection signal 805. For example, if the cycle 801 is 20 m / sec, the combination of selecting the three rounds is different when 20 m / sec is divided into four parts.

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

[0068] In addition, in this Figure 8 In the example, when the estimated value verification / correction unit 705 is one system, the control signal is implemented by serial processing. If the estimated value verification / correction unit 705 is four systems, it is generally implemented by parallel processing.

[0069] ( Fig. 9 ) The flowchart of the third embodiment is described. In step S1, the value of the counter of 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 estimated value calculated using the sprung up-down speed of the wheels FL, FR, and RL. In step S3, the first pitch rate calculation unit 116 performs pitch rate calculation B using the sensor estimated value calculated using the sprung up-down speed of the wheels FR, RL, and RR. In step S3, the first pitch rate calculation unit 116 performs pitch rate calculation C using the sensor estimated value calculated using the sprung up-down speed of the wheels RL, RR, and FL. In step S4, the first pitch rate calculation unit 116 performs pitch rate calculation D using the sensor estimated value calculated using the sprung up-down speed of the wheels RR, FL, and FR.

[0070] In step S6, the estimated value verification / correction unit 705 calculates a value obtained by subtracting the calculation result 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 a value obtained by subtracting the calculation result 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 a value obtained by subtracting the calculation result 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 a value obtained by subtracting the calculation result 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 of the four results calculated in steps S6 to S9 is less than a predetermined threshold value. If the difference of the four calculated results is small, the process proceeds to step S18, and the estimated value verification / correction unit 705 uses four estimated results. Thereafter, in order to continue to determine whether the four estimated results can be used (whether the difference is less than a predetermined threshold value), the process returns to step S2.

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

[0073] In step S11, when the difference calculated in step S6 is less than a predetermined threshold value, the sensor estimated value calculated using the sprung vertical speeds of the wheels FL, FR, and RL in the pitch rate calculation A in step S2 is adopted in step S15. In step S16, the sprung vertical speed of the remaining wheel RR is calculated based on this, and the sprung vertical speed of the wheel RR calculated in step S17 is stored in the data set. In addition, since the calculation process of the wheel FL, i.e., steps S18 to S20, the calculation process of the wheel FR, i.e., steps S21 to S23, and the calculation process of the wheel RL, i.e., steps S24 to S26, are the same as the process of steps S15 to S17, their description is omitted.

[0074] In step S27, the travel data storage unit 120 performs counting processing. In step S28, if the count processing number in step S27 exceeds 1000, then in step S29, the travel data storage unit 120 sends an updated data set to the learning management server 102, and in step S30, the data set in the travel data storage unit 120 is deleted, and the process returns to step S1 to repeat the above process. If the count processing number in step S28 does not exceed 1000, the process returns to step S2 to repeat the above process.

[0075] (Fourth embodiment) ( Fig.10 ) The fourth embodiment has a configuration that allows relearning (parameter updating) to be performed using only a virtual environment as long as there is a flag indicating proficiency / disadvantage of driving data. In the above-mentioned embodiment, the estimation accuracy of the sprung vertical velocity and the piston velocity output by the vehicle state estimation unit 112 is improved while driving. In contrast, in the fourth embodiment, the driving environment is reproduced in a virtual environment, and the estimation accuracy of the sprung vertical velocity and the piston velocity is improved in the virtual environment.

[0076] Specifically, the accuracy status is transmitted to the learning management server 102 together with the road information (including the road profile) acquired by the GPS sensor 504 and the stereo camera 604 as the existing sensors, and is reproduced by simulation in the learning management server 102, thereby implementing learning by generating training data.

[0077] The feature of this embodiment is that, in 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 the real terrain and road shape in a virtual environment is further included in the configuration. The map information possessed by the map information server 1104 is, for example, open source OSM (OpenStreet Map), or other paid / free map data. These map data may include elevation / slope information, and as data collection progresses, information such as road contours may also be applied.

[0078] In addition, if the information of the weather information server 503 is given to the map information server 1104, the reproducibility of the real environment in the virtual environment is improved. Therefore, the learning management server 102 learns the weighting parameters based on the second sensor information (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, if the vehicle behavior simulator 1103 of the learning management server 102 is used in an advanced virtual environment, the estimation accuracy of the sprung vertical velocity and the piston velocity output by the vehicle state estimation unit 112 can be improved even if it is not necessarily the measurement of the actual vehicle.

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

[0080] (1) The suspension system associates sensor information acquired from a plurality of sensors 106 provided in the vehicle 101 with state information of a suspension 109 provided in the vehicle 101, and performs calculations by machine learning. The suspension system includes: 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 the weight parameters. The suspension system includes: 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 the 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. The suspension system includes a driving data management unit 119 that instructs the learning management server 102 to learn weighted parameters based on the output result of the estimation accuracy of the estimation accuracy verification unit 118. In this way, in a vehicle without a suspension sensor, the weak state of the neural network can be extracted to assist learning.

[0081] (2) The vehicle 101 includes four vehicle state estimating units 112 that cooperate with the four wheels of the vehicle 101, respectively, and includes a selector 703 that selects three estimation results from four estimation results output by the four vehicle state estimating units 112. Thus, the vehicle state estimating unit 112 can estimate the state of the suspension 109 based on only the estimation results of the three wheels.

[0082] (3) The first vehicle behavior calculation unit 116 calculates four first physical values ​​based on the three estimation results selected by the selector 703. When four difference values ​​based on the comparison between each of the four first physical values ​​and the second physical value are smaller than a predetermined threshold value, 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. When the difference value based on the comparison between any one of the four first physical values ​​calculated and the second physical value is smaller than a predetermined threshold value, the estimation accuracy verification unit 118 adopts the three estimation results selected by the selector 703 for calculating the first physical value as the state of the suspension 109. When the difference value based on the comparison between the four first physical values ​​calculated and the second physical value is larger than a predetermined threshold value, the first vehicle behavior calculation unit 116 recalculates the four first physical values ​​based on the three estimation results selected by the selector 703. In this way, it is possible to determine the estimation accuracy of the estimation result taking into account the low-precision vehicle state.

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

[0084] (5) The estimated accuracy output by the estimated accuracy verification unit 118 is presented to the user via the display unit 110. This allows the estimated accuracy of the state of the suspension 109 to be visualized.

[0085] (6) The first physical value and the second physical value are the pitch rate or the roll rate of the vehicle 101. Thus, the first physical value and the second physical value can be obtained from a plurality of vehicle behavior data.

[0086] (7) The learning management server 102 learns weighting 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 vertical velocity and the piston velocity in the virtual environment.

[0087] (8) The learning management server 102 learns the weighting parameters using a computer mounted on the vehicle 101. In this way, the weighting parameters can be updated without providing the learning management server 102 on the vehicle 101 side.

[0088] In addition, the present invention is not limited to the above-mentioned embodiments, and various modifications or other configurations can be combined within the scope of the gist thereof. In addition, the present invention is not limited to the configuration having all the configurations described in the above-mentioned embodiments, and also includes a configuration in which a part of the configuration is deleted. 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 Department 111 interface (I / F) 112 Vehicle state estimation unit 113 Weighted Data Management Department 114 weighting parameter storage unit (storage unit) 115 Suspension control value calculation unit 116 1st pitch rate calculation unit 117 2nd pitch rate calculation unit 118 Estimation Accuracy Verification Department 119 Driving Data Management Department 120 driving data storage unit 201 Input layer components of neural network 202 Hidden layer component group of neural network 203 Output layer elements of neural network 204 Time Series Data 205 sensor data 206 windows 207 sensor data instantaneous value 401 Measurement device screen 402 Wheel speed data (CAN system data) 403 front and rear acceleration data (CAN system data) 404 Piston speed data (vehicle state estimated value) 405 pitch rate data (vehicle movement physical value) 406 File Output Button 407 terminal screen 408 Pitch rate from neural network 409 Pitch rate from non-neural network 410 accuracy determination results 411 Dashboard 412 Vehicle Behavior Estimation Result 413 judgment result indicator light 414 Odometer 503 Weather Information Server 504GPS sensor 505 Stereo Camera 506 Wheel speed reliability determination unit 601 Road surface μ estimation unit 703 Selector 705 Estimated value verification / correction unit 801 Vehicle status estimation cycle 802~805 selection signal 1103 Vehicle Behavior Simulator 1104 map information server.

Claims

1. A suspension system that associates sensor information obtained from a plurality of sensors provided on a vehicle with state information of a suspension provided on the vehicle, performs calculations through machine learning, The suspension system is characterized by comprising: A weighting parameter storage unit, which stores the weighting parameters calculated by the machine learning; a vehicle state estimating unit that outputs an estimation result of the state of the suspension based on the sensor information and the weighting parameter; 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 outputted 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 an estimation accuracy of the state of the suspension by the vehicle state estimation unit by comparing the first physical value and the second physical value; and A travel data management unit instructs a learning management server to learn the weighting parameter based on an output result of the estimation accuracy of the estimation accuracy verification unit.

2. The suspension system according to claim 1, characterized in that: The vehicle includes four vehicle state estimating units respectively cooperating with four wheels provided on the vehicle. The vehicle includes a selector for selecting three of the estimation results from among the four estimation results outputted from the four vehicle state estimation units.

3. The suspension system according to claim 2, characterized in that: The first vehicle behavior calculation unit calculates four types of the first physical values ​​based on the three estimation results selected by the selector. When four difference values ​​based on the comparison between the four first physical values ​​and the second physical value are smaller than a predetermined threshold value, the estimation accuracy verification unit adopts the four estimation results output from the four vehicle state estimation units as the state of the suspension, When a difference value based on a comparison between the first physical value of one of the four calculated first physical values ​​and the second physical value is less than a predetermined threshold value, the estimation accuracy verification unit adopts the three estimation results selected by the selector for calculating the first physical value as the state of the suspension, When a difference value based on a comparison between the calculated four first physical values ​​and the second physical value is greater than a predetermined threshold, the first vehicle behavior calculation unit recalculates the four first physical values ​​based on the three estimation results selected by the selector.

4. The suspension system according to claim 3, characterized in that: The estimation accuracy verification unit calculates the estimation result not selected by the selector based on the three estimation results selected by the selector and the second physical value.

5. The suspension system according to claim 1, characterized in that: The estimated accuracy output by the estimated accuracy verification unit is presented to a user via a display unit.

6. The suspension system according to claim 1, characterized in that: The first physical value and the second physical value are a pitch rate or a roll rate of the vehicle.

7. The suspension system according to claim 1, characterized in that: The learning management server learns the weighting parameter based on second sensor information acquired by a second sensor mounted on the vehicle and virtual environment information learned outside the vehicle.

8. The suspension system according to claim 4, characterized in that: The learning management server learns the weighting parameters based on a computer mounted on the vehicle.

Images