control device

Abstract

This control device comprises: an acquisition unit that acquires the acceleration of a vehicle body in a vertical direction of a vehicle; and an estimation unit. The estimation unit estimates the relative speed according to a numerical value obtained by multiplying: an acceleration obtained by the acquisition unit using a neural network that has been trained to estimate the relative speed of the wheels to the vehicle body in the vertical direction of the vehicle in accordance with the input of a numerical value obtained by multiplying the acceleration and the mass of the vehicle; and a mass the size of which is the same or different from that at the time of learning of the neural network. The control device controls a suspension device on the basis of the estimated relative speed.

Description

[Technical Field]

[0001] Embodiments of the present invention relate to a control device. [Background technology]

[0002] Conventionally, a technique is known in which the relative speed of the wheels to the vehicle body in the vertical direction, so-called stroke speed, is estimated by a physical quantity estimation device, and the damping force of the shock absorber of the suspension system is controlled based on the estimated stroke speed. [Prior art documents] [Patent Documents]

[0003] [Patent Document 1] Japanese Patent Application Publication No. 10-913 [Patent Document 2] Japanese Patent Publication No. 2016-117326 [Overview of the project] [Problems that the invention aims to solve]

[0004] When estimating physical quantities such as stroke speed using neural networks, it was time-consuming because a separate neural network had to be created for each vehicle specification.

[0005] Therefore, one of the objectives of the embodiments of the present invention is to reduce the effort required to create a neural network, even when a vehicle has multiple specifications. [Means for solving the problem]

[0006] An embodiment of the present invention is a control device for controlling a suspension system interposed between the vehicle body and the wheels of a vehicle, comprising: an acquisition unit for acquiring the acceleration of the vehicle body in the vertical direction; and an estimation unit for estimating the relative velocity of the wheels with respect to the vehicle body in the vertical direction, using a neural network that has been trained to estimate the relative velocity in response to an input of a numerical value obtained by multiplying the acceleration by the mass, which is the same as or different in magnitude as when the neural network was trained, and controlling the suspension system based on the estimated relative velocity. [Effects of the Invention]

[0007] According to the control device of the present invention, for example, the same neural network can be used for vehicles with multiple specifications having different masses. Therefore, even when a vehicle has multiple specifications, the effort required to create the neural network can be reduced. [Brief explanation of the drawing]

[0008] [Figure 1] Figure 1 is a schematic diagram showing the general configuration of an example of a vehicle according to the first embodiment. [Figure 2] Figure 2 is a schematic diagram showing an example of the configuration of a suspension system in a vehicle according to the first embodiment. [Figure 3] Figure 3 is a functional block diagram of the control device for an example of a vehicle according to the first embodiment. [Figure 4] Figure 4 is an explanatory diagram illustrating the physical quantities of a vehicle according to the first embodiment. [Figure 5] Figure 5 shows an example of a neural network used for analysis. [Figure 6] Figure 6 shows an example of stroke velocity estimation results using a neural network for analysis, specifically the case where the vehicle mass is the reference. [Figure 7]FIG. 7 is a diagram showing an example of the estimated result of the stroke speed by the neural network for examination, and shows the case where the mass of the vehicle is half of the reference. [Figure 8] FIG. 8 is a diagram showing an example of the estimated result of the stroke speed by the neural network for examination, and shows the case where the mass of the vehicle is 1.5 times the reference. [Figure 9] FIG. 9 is a diagram showing an example of the estimated result of the stroke speed by the equation of motion for examination, and shows the case where the mass of the vehicle is the reference. [Figure 10] FIG. 10 is a diagram showing an example of the estimated result of the stroke speed by the equation of motion for examination, and shows the case where the mass of the vehicle is half of the reference. [Figure 11] FIG. 11 is a diagram showing an example of the estimated result of the stroke speed by the equation of motion for examination, and shows the case where the mass of the vehicle is 1.5 times the reference. [Figure 12] FIG. 12 is a diagram for explaining the neural network for examination. [Figure 13] FIG. 13 is a diagram showing the neural network according to the first embodiment. [Figure 14] FIG. 14 is a diagram showing an example of the estimated result of the stroke speed by the neural network according to the first embodiment, and shows the case where the mass of the vehicle is the reference. [Figure 15] FIG. 15 is a diagram showing an example of the estimated result of the stroke speed by the neural network according to the first embodiment, and shows the case where the mass of the vehicle is half of the reference. [Figure 16] FIG. 16 is a diagram showing an example of the estimated result of the stroke speed by the neural network according to the first embodiment, and shows the case where the mass of the vehicle is 1.5 times the reference. [Figure 17] FIG. 17 is a diagram showing an example of the estimated result of the stroke speed by the equation of motion of the comparative example, and shows the case where the mass of the vehicle is the reference. [Figure 18]Figure 18 shows an example of the stroke velocity estimation results using the equation of motion for a comparative example, specifically the case where the vehicle mass is half of the standard mass. [Figure 19] Figure 19 shows an example of the stroke velocity estimation results using the equation of motion for a comparative example, where the vehicle mass is 1.5 times the standard mass. [Figure 20] Figure 20 is a flowchart showing an example of a physical quantity estimation direction executed by the control device according to the first embodiment. [Figure 21] Figure 21 shows a neural network according to the second embodiment. [Figure 22] Figure 22 shows a neural network according to the third embodiment. [Figure 23] Figure 23 shows a neural network according to the fourth embodiment. [Modes for carrying out the invention]

[0009] Embodiments will be described in detail below with reference to the drawings. However, the present invention is not limited by these embodiments. The following embodiments include similar components. These similar components are given common reference numerals, and redundant descriptions are omitted.

[0010] <First Embodiment> Figure 1 is a schematic diagram showing the general configuration of an example of a vehicle 1 according to the first embodiment. In this embodiment, vehicle 1 may be, for example, an automobile (internal combustion engine automobile) driven by an internal combustion engine (engine, not shown), an automobile (electric vehicle, fuel cell vehicle, etc.) driven by an electric motor (motor, not shown), or an automobile (hybrid vehicle) driven by both. Vehicle 1 can also be equipped with various transmissions and various devices (systems, parts, etc.) necessary for driving the internal combustion engine or electric motor. Furthermore, the type, number, layout, etc. of the devices related to driving the wheels 3 in vehicle 1 can be set in various ways. In this embodiment, as an example, vehicle 1 is a four-wheeled vehicle (four-wheeled automobile) and has two front wheels 3F on the left and right, and two rear wheels 3R ​​on the left and right. In Figure 1, the front (direction Fr) in the longitudinal direction of the vehicle is the left side. Vehicle 1 is an example of an object. The object may be a moving body or device other than vehicle 1.

[0011] In this embodiment, as an example, the vehicle control system 100 of vehicle 1 includes a control device 10, a steering device 11, a steering angle sensor 12, a yaw rate sensor 13, a braking system 61, etc. The vehicle control system 100 also includes a suspension system 4, a rotation sensor 5, an acceleration sensor 6, etc. corresponding to each of the two front wheels 3F, and a suspension system 4, a rotation sensor 5, an acceleration sensor 6, etc. corresponding to each of the two rear wheels 3R. Although vehicle 1 has other basic components as vehicle 1 in addition to those shown in Figure 1, only the configuration related to the vehicle control system 100 and the control related to that configuration will be described here. The term "wheels 3" is used as a general term for the two front wheels 3F and the two rear wheels 3R. The relative speed of the wheels 3 of vehicle 1 with respect to the vehicle body 2 of vehicle 1 is also called the stroke speed, the part of vehicle 1 on the vehicle body 2 side with respect to the suspension system 4 is called the sprung mass, and the part of vehicle 1 on the wheel 3 side with respect to the suspension system 4 is called the unsprung mass. Furthermore, unless otherwise specified in the following explanation, relative speed refers to the relative speed of the wheels 3 to the vehicle body 2 in the vertical direction of vehicle 1.

[0012] The control device 10 receives signals and data from various parts of the vehicle control system 100, and performs control and various calculations on the various parts of the vehicle control system 100. The control device 10 is also called a state quantity estimation device, a vehicle state quantity estimation device, or a physical quantity estimation device. The control device 10 is configured as a computer and includes an arithmetic processing unit (microcomputer, ECU (Electronic Control Unit), etc., not shown) and a storage unit 10d (for example, ROM (Read Only Memory), RAM (Random Access Memory), flash memory, etc., see Figure 3). The arithmetic processing unit reads a program stored (installed) in the non-volatile storage unit 10d (for example, ROM or flash memory), performs calculations according to the program, and can function (operate) as the various parts shown in Figure 3. The storage unit 10d can also store data used in various calculations related to control (tables (data groups), functions, etc.), calculation results (including intermediate values), etc. The storage unit 10d also stores a neural network 121. Details of neural network 121 will be described later.

[0013] The steering system 11, for example, includes a steering wheel and steers (turns) the two front wheels 3F. The steering angle sensor 12 detects the steering angle (steering angle, turning angle, steering angle) of the front wheels 3F and outputs steering angle data indicating the detected steering angle.

[0014] The yaw rate sensor 13 detects the yaw rate of the vehicle 1 (vehicle body 2) and outputs yaw rate information indicating the detected yaw rate.

[0015] Figure 2 is a schematic diagram showing the general configuration of an example of a suspension system 4 in a vehicle according to the first embodiment. As shown in Figure 2, the suspension system 4 is interposed between the wheel 3 and the vehicle body 2 to suppress the transmission of vibrations and shocks from the road surface to the vehicle body. The suspension system 4 includes a coil spring 4a and a shock absorber 4b. The damping force (damping characteristics) of the shock absorber 4b can be electrically controlled (adjusted). Specifically, the shock absorber 4b has an actuator 4bb that operates based on the input current. The actuator 4bb can change the opening degree of the orifice provided in the piston of the shock absorber 4b, or change the opening degree between the valve body and the valve seat. This controls the flow rate of lubricating oil circulating between two oil chambers separated by the piston within the shock absorber 4b, thereby adjusting the damping force of the shock absorber 4b. The suspension system 4 is provided on each of the four wheels 3 (two front wheels 3F and two rear wheels 3R), and the control device 10 can control the damping force of each of the four wheels 3. The control device 10 can control the four wheels 3 to have different damping forces from each other. More specifically, in order to control the damping force of the shock absorber 4b (= damping coefficient × stroke speed), the damping coefficient is controlled. More specifically, the damping coefficient is controlled by controlling the current flowing through the actuator 4bb of the shock absorber 4b.

[0016] The rotation sensor 5 can output signals corresponding to the rotational speed (angular velocity, rotational speed, rotational state) of each of the four wheels 3. The control device 10 can calculate the speed of the vehicle 1 from the detection results of the rotation sensor 5. In addition to the rotation sensor 5 for the wheels 3, a rotation sensor (not shown) that detects the rotation of the crankshaft, axle, etc. may be provided, and the control device 10 may obtain the speed of the vehicle 1 from the detection results of this rotation sensor.

[0017] The acceleration sensors 6 are installed on the vehicle body 2, corresponding to each of the four wheels 3 and the suspension system 4, and detect the acceleration of the vehicle body 2 in the vehicle 1. Specifically, each acceleration sensor 6 is installed at an acceleration detection target location on the vehicle body 2, directly above each wheel 3 (directly above the wheel). In other words, in this embodiment, there are four acceleration detection target locations. In this embodiment, as an example, the acceleration sensor 6 can acquire the vertical acceleration of the vehicle 1, the longitudinal acceleration of the vehicle 1, and the widthwise acceleration of the vehicle 1 (width direction, short direction, left-right direction).

[0018] The above-described configuration of the vehicle control system 100 is merely an example and can be modified in various ways. Known devices can be used as the individual components of the vehicle control system 100. Furthermore, each component of the vehicle control system 100 can be shared with other configurations.

[0019] In this embodiment, for example, the control device 10 can function (operate) as an acquisition unit 10a, an estimation unit 10b, an attenuation control unit 10c, etc., as shown in Figure 3, through the cooperation of hardware and software (program). That is, the program may include, for example, modules corresponding to each block except for the storage unit 10d shown in Figure 3.

[0020] The acquisition unit 10a acquires acceleration data indicating the acceleration of the vehicle 1 from each acceleration sensor 6. The acceleration of the vehicle 1 may be acceleration in any direction. Furthermore, the acceleration of the vehicle 1 may be acceleration in multiple directions. That is, the acceleration of the vehicle 1 may be one or more of the following: acceleration in the longitudinal direction of the vehicle 1, acceleration in the lateral direction of the vehicle 1, and acceleration in the vertical direction of the vehicle 1. In addition, the acquisition unit 10a acquires the current value (current value data), which is the value of the current input to the shock absorber 4b, from the damping control unit 10c.

[0021] The estimation unit 10b receives, for example, predetermined data acquired by the acquisition unit 10a (hereinafter also referred to as input data). The estimation unit 10b uses a neural network 121 to estimate the stroke velocity according to the input data. The input data includes at least acceleration data. In addition to acceleration data, the input data may include at least one of the following: yaw rate data, current value data, steering angle data, and wheel speed data. However, the input data is not limited to the above.

[0022] Here, a single vehicle system of Vehicle 1 may have multiple specifications (types, characteristics). For example, multiple specifications could include gasoline-two-wheel drive, hybrid-two-wheel drive, gasoline-four-wheel drive, and hybrid-four-wheel drive. The gasoline-two-wheel drive specification has a gasoline engine as the power source and two drive wheels. The hybrid-two-wheel drive specification has a gasoline engine and electric motor as the power source and two drive wheels. The gasoline-four-wheel drive specification has a gasoline engine as the power source and four drive wheels. The hybrid-four-wheel drive specification has a gasoline engine and electric motor as the power source and four drive wheels. Vehicle 1 with multiple specifications like these may differ in mass, for example. Therefore, conventionally, in order to estimate the stroke speed of each Vehicle 1, a neural network was created for each Vehicle 1, even within a single vehicle system. Creating a neural network requires measuring input data and training data and performing training, and creating multiple neural networks is time-consuming.

[0023] Therefore, the inventors considered a configuration (method) that can reduce the effort required to create a neural network even when vehicle 1 has multiple specifications. The inventors' considerations are described below.

[0024] The inventors compared the neural network with the equations of motion that describe the motion of vehicle 1. Here, Figure 4 is an explanatory diagram illustrating the physical quantities of vehicle 1 according to the first embodiment. As shown in Figure 4, the mass of vehicle 1 is M b Let x be the amount of sprung displacement, which is the amount of vertical displacement of the vehicle body 2. bLet the spring constant of the coil spring be K c Let the damping coefficient of the shock absorber 4b be C x Let the displacement of the wheel 3 in the vertical direction, i.e., the displacement under the spring, be x t Let the spring constant of the tire of the wheel 3 be K t Let the displacement of the lower end of the tire be x r Assume this is the case. In this case, the equation of motion of the vehicle 1 is expressed by the following formula (1). M b (x b )´´=-K c (x b ―x t )-C x ((x b )´-(x t )´)·····(1) In the formula (1), "´" and "´´" represent the first derivative and the second derivative, respectively. Therefore, the estimated stroke speed ((x b )´-(x t )´) is expressed by the formula (2). ((x b )´-(x t )´)=-(M b (x b )´´) / C x -(K c (x b ―x t )) / C x ·····(2)

[0025] Figure 5 is a diagram showing an example of the neural network 200 for consideration. As shown in Figure 5, in the neural network 200 for consideration, the input data is the acceleration above the spring and the current value of the shock absorber 4b, and the output data is the stroke speed.

[0026] The above neural network 200 and formulas (1) and (2) are created based on the reference mass of the vehicle 1.

[0027] The inventors calculated estimated stroke velocity values ​​using the neural network 200 and equation (2) while varying the mass. These results are shown in Figures 6 to 11. Figure 6 shows an example of stroke velocity estimation results using the neural network 121 for consideration, where the mass of vehicle 1 is the standard. Figure 7 shows an example of stroke velocity estimation results using the neural network 121 for consideration according to the first embodiment, where the mass of vehicle 1 is half of the standard. Figure 8 shows an example of stroke velocity estimation results using the neural network 121 for consideration, where the mass of vehicle 1 is 1.5 times the standard. Figure 9 shows an example of stroke velocity estimation results using the equation of motion for consideration, where the mass of vehicle 1 is the standard. Figure 10 shows an example of stroke velocity estimation results using the equation of motion for consideration, where the mass of vehicle 1 is half of the standard. Figure 11 shows an example of stroke velocity estimation results using the equation of motion for consideration, where the mass of vehicle 1 is 1.5 times the standard.

[0028] As shown in Figures 6 and 9, when the input mass was set to the reference mass, both neural network 200 and equation (2) yielded results where the estimated value was close to the true value. As shown in Figures 7 and 10, when the input mass was set to half the reference mass, both neural network 200 and equation (2) yielded results where the deviation from the true value to the estimated value was large. As shown in Figures 8 and 11, when the input mass was set to 1.5 times the reference mass, both neural network 200 and equation (2) yielded results where the deviation from the true value to the estimated value was large. In other words, if a different numerical value is intentionally entered as the input value, the deviation from the true value to the estimated value will be large. This deviation can be considered an error.

[0029] From the above results, the inventor discovered that the way in which the estimated value deviates from the true value, i.e., the way in which the estimated value is incorrect, when the input value is intentionally entered as a different numerical value, is similar in the neural network 200 and equation (2). Then, as shown in Figure 12, the inventor considered that there is a part 200a corresponding to the parameter of mass within the neural network 200, and came to the idea that it is possible to move the parameter of mass outside the neural network 200. Then, the inventor came up with a configuration in the neural network 121 of this embodiment in which the parameter of mass is moved outside.

[0030] The neural network 121 will be described below. Figure 13 shows the neural network 121 according to the first embodiment. The neural network 121 is configured to be pre-trained to estimate the relative velocity of the wheels 3 of vehicle 1 relative to the body 2 of vehicle 1 in the vertical direction, i.e., the stroke velocity, in response to input data. For example, the neural network 121 is configured using LSTM (Long short-term memory). During training of the neural network 121, the input data is the numerical value of the force obtained by multiplying the sprung mass by the mass, and the current value of the shock absorber 4b. The sprung mass is the acceleration detected by the acceleration sensor 6, i.e., the acceleration of the body 2 of vehicle 1 in the vertical direction. The mass is the reference mass. In other words, the mass is a constant. The output data as training data is, for example, the measured value (true value) of the relative velocity of the wheels 3 of vehicle 1 relative to the body 2 of vehicle 1, i.e., the stroke velocity, measured by a stroke velocity sensor (not shown). Sprung acceleration is an example of the first physical quantity, mass is an example of the second physical quantity, and stroke velocity is an example of the third physical quantity.

[0031] Figure 14 shows an example of the stroke velocity estimation results by the neural network 121 according to the first embodiment, where the mass of vehicle 1 is the reference mass. Figure 15 shows an example of the stroke velocity estimation results by the neural network 121 according to the first embodiment, where the mass of vehicle 1 is half the reference mass. Figure 16 shows an example of the stroke velocity estimation results by the neural network 121 according to the first embodiment, where the mass of vehicle 1 is 1.5 times the reference mass.

[0032] As shown in Figures 14 to 16, even when the mass used to calculate the input data was changed for the neural network 121, the estimated values ​​and the true values ​​were found to be in close agreement.

[0033] In this case, if the mass taken outside the neural network 121 is input to the neural network 121 as standalone input data without being multiplied by the sprung acceleration, the neural network 121 will not recognize the mass because it has not learned mass as a parameter. Therefore, even if the mass is changed and input to the neural network 121, the discrepancy between the estimated value and the true value will be large. In contrast, in this embodiment, since the mass is multiplied by the sprung acceleration, the product of the mass and the sprung acceleration is recognized by the neural network 121 as a parameter. Therefore, even if the mass is changed, the discrepancy between the estimated value and the true value will not be large. Consequently, the same neural network 121 can be used for multiple vehicles 1 with different masses. That is, a neural network 121 created for one vehicle 1 among multiple vehicle 1 specifications can be used for other vehicles 1 with different specifications. In such a neural network 121, the mass of vehicle 1 (the second physical quantity) may be the same or different in magnitude as when the neural network 121 was trained. For example, if a neural network 121 is created for a gasoline-powered, two-wheel-drive vehicle 1, the mass of the vehicle 1 used in the neural network 121 is the same as during training. On the other hand, for neural networks 121 installed in other vehicle specifications (hybrid, two-wheel-drive; gasoline, four-wheel-drive; and hybrid, four-wheel-drive), the mass of the vehicle 1 used in the neural network 121 is different from that during training.

[0034] Here, the waveforms shown in Figures 14 to 16 differ from those shown in Figures 6 to 8 because the specifications of Vehicle 1 used in the simulation are different. The waveform of the stroke velocity obtained from the equation of motion (comparative example) created using the specifications of Vehicle 1 for the waveforms shown in Figures 14 to 16 is shown in Figures 17 to 19.

[0035] Returning to Figure 3, the damping control unit 10c controls the damping force of the shock absorber 4b based on the stroke velocity estimated by the estimation unit 10b. Specifically, the damping control unit 10c determines the current value to be input to the shock absorber 4b and inputs this current value to the shock absorber 4b.

[0036] Next, an example of a physical quantity estimation direction executed by the control device 10 will be described with reference to Figure 20. Figure 20 is a flowchart showing an example of a physical quantity estimation direction executed by the control device 10 according to the first embodiment.

[0037] As shown in Figure 20, first, the acquisition unit 10a acquires input data (data) (S1). The input data is, for example, the sprung mass acceleration and the current value of the shock absorber 4b. Next, the estimation unit 10b inputs the numerical value obtained by the acquisition unit 10a multiplying the sprung mass acceleration and the mass of the vehicle 1, and the current value acquired by the acquisition unit 10a, into the neural network 121, and the neural network 121 estimates the physical quantity to be estimated (S2). The physical quantity is, for example, the stroke velocity. The mass is stored in the memory unit 10d. That is, the memory unit 10d stores the second physical quantity that is input into the neural network 121.

[0038] As described above, in this embodiment, the control device 10 comprises an acquisition unit 10a and an estimation unit 10b. The acquisition unit 10a acquires a first physical quantity, which is a variable, relating to the vehicle 1 (object). The estimation unit 10b uses a neural network 121 that has been trained to estimate a third physical quantity relating to the vehicle 1 in response to an input of a numerical value obtained by multiplying the first physical quantity acquired by the acquisition unit 10a by a second physical quantity that is the same or different in magnitude as when the neural network 121 was trained. The third physical quantity corresponds to a numerical value obtained by multiplying the first physical quantity acquired by the acquisition unit 10a by a second physical quantity that is the same or different in magnitude as when the neural network 121 was trained. The first to third physical quantities are, as an example, physical quantities relating to the motion of the vehicle 1.

[0039] With this configuration, for example, the same neural network 121 can be used for multiple vehicles 1 that have different specifications for the second physical quantity. Therefore, even if vehicle 1 has multiple specifications, the effort required to create the neural network 121 can be reduced.

[0040] Furthermore, the first physical quantity is the acceleration of the vehicle body 2 in the vertical direction of vehicle 1, and the second physical quantity is the mass of vehicle 1. The third physical quantity is the relative velocity (stroke velocity) of the wheels 3 with respect to the vehicle body 2 in the vertical direction of vehicle 1.

[0041] With this configuration, even if there are multiple specifications of vehicle 1 with different masses, the effort required to create the neural network 121 can be reduced.

[0042] <Second Embodiment> Figure 21 shows a neural network 121 according to the second embodiment.

[0043] As shown in Figure 21, this embodiment differs from the first embodiment mainly in that the input data input to the neural network 121 is the numerical value of the force obtained by multiplying the sprung mass by the mass of the vehicle 1, and the damping force of the shock absorber 4b.

[0044] The damping force of the shock absorber 4b is calculated based on the current value of the shock absorber 4b and the stroke velocity (third physical quantity) previously estimated by the estimation unit 10b. Specifically, the damping force is calculated based on the damping coefficient of the shock absorber corresponding to the current value of the shock absorber 4b and the stroke velocity previously estimated by the estimation unit 10b. The estimation unit 10b estimates a new stroke velocity (relative velocity) by multiplying the damping coefficient or a quantity related to the damping coefficient of the shock absorber 4b by the previously estimated stroke velocity (relative velocity). The stroke velocity (third physical quantity) previously estimated by the estimation unit 10b corresponds to the first physical quantity in the current estimation of stroke velocity by the estimation unit 10b. The current value of the shock absorber 4b is an example of the second physical quantity.

[0045] As described above, in this embodiment, the estimation unit 10b uses the previously estimated third physical quantity as the first physical quantity when estimating the third physical quantity.

[0046] With this configuration, the estimation unit 10b can estimate the latest third physical quantity using the third physical quantity estimated previously.

[0047] Furthermore, the second physical quantity is a quantity related to the damping coefficient of the shock absorber 4b (current value). The third physical quantity is the relative speed of the wheels 3 to the vehicle body 2 in the vertical direction of the vehicle 1, i.e., the stroke speed.

[0048] With this configuration, even if there are multiple specifications for the shock absorber 4b with different damping coefficients in the vehicle 1, the effort required to create the neural network 121 can be reduced.

[0049] Alternatively, the damping coefficient of shock absorber 4b may be used instead of the current value, which is a quantity related to the damping coefficient of shock absorber 4b.

[0050] <Third Embodiment> Figure 22 shows a neural network 121 according to the third embodiment.

[0051] In this embodiment, as shown in Figure 22, the main difference from the second embodiment is that the input data input to the neural network 121 is the numerical value of the force obtained by multiplying the sprung mass by the mass of the vehicle 1, the damping force of the shock absorber 4b, and the spring force of the coil spring 4a. In this embodiment, the output data is the stroke velocity and the relative displacement of the wheels 3 with respect to the vehicle body 2 in the vertical direction of the vehicle 1 (so-called stroke displacement).

[0052] The spring force of the coil spring 4a is calculated based on the spring constant of the coil spring 4a and the stroke displacement amount previously estimated by the estimation unit 10b. Specifically, the spring force of the coil spring 4a is calculated by multiplying the spring constant of the coil spring 4a by the stroke displacement amount previously estimated by the estimation unit 10b. The estimation unit 10b estimates a new stroke displacement amount by multiplying the spring constant of the coil spring 4a by the previously estimated stroke displacement amount. The stroke displacement amount previously estimated by the estimation unit 10b (the third physical quantity) corresponds to the first physical quantity in the current estimation of the stroke displacement amount by the estimation unit 10b. The spring constant of the coil spring 4a is an example of the second physical quantity.

[0053] As described above, in this embodiment, the second physical quantity is the spring constant of the coil spring 4a (spring). The third physical quantity is the stroke displacement, which is the relative displacement of the wheel 3 with respect to the vehicle body 2 in the vertical direction of the vehicle 1.

[0054] With this configuration, even if there are multiple specifications for the coil spring 4a with different spring constants in the vehicle 1, the effort required to create the neural network 121 can be reduced.

[0055] <Fourth Embodiment> Figure 23 shows a neural network 121 according to the fourth embodiment.

[0056] In this embodiment, the vehicle 1 has a front stabilizer (not shown) and a rear stabilizer (not shown). Also, as shown in Figure 23, in this embodiment, the main difference from the second embodiment is that the input data input to the neural network 121 is the numerical value of the force obtained by multiplying the sprung mass by the mass of the vehicle 1, the spring force of the front stabilizer, and the spring force of the rear stabilizer. In addition, in this embodiment, the output data (third physical quantity) is the stroke velocity, the FR stroke displacement, which is the stroke displacement of the right front of the vehicle 1, the FL stroke displacement, which is the stroke displacement of the left front of the vehicle 1, the RR stroke displacement, which is the stroke displacement of the right rear of the vehicle 1, and the RL stroke displacement, which is the stroke displacement of the right front of the vehicle 1.

[0057] The front stabilizer spring force is calculated by the estimation unit 10b using the difference between the FR stroke displacement and the FL stroke displacement previously estimated, and the spring constant (K) of the front stabilizer. f The rear stabilizer spring force is calculated by multiplying the difference between the RR stroke displacement and RL stroke displacement previously estimated by the estimation unit 10b and the spring constant of the rear stabilizer (K r It is calculated by multiplying by ). The FL stroke displacement, FR stroke displacement, RR stroke displacement, and RR stroke displacement (third physical quantity) previously estimated by the estimation unit 10b correspond to the second physical quantity in the current estimation of stroke displacement by the estimation unit 10b. The spring constant of the front stabilizer and the spring constant of the rear stabilizer are the second physical quantity.

[0058] As described above, in this embodiment, the second physical quantity is the spring constant of the front stabilizer and the spring constant of the rear stabilizer.

[0059] With this configuration, even if there are multiple specifications of vehicle 1 with different spring constants for the front stabilizer and rear stabilizer, the effort required to create the neural network 121 can be reduced.

[0060] In the above embodiment, the first physical quantity may be the longitudinal acceleration of the vehicle body 2, the lateral acceleration of the vehicle body 2, the wheel speed, the yaw rate of the vehicle body 2, etc. The second physical quantity may be the height of the center of gravity of the vehicle 1. The pitch moment may be calculated by multiplying the longitudinal acceleration of the vehicle body 2 by the height of the center of gravity of the vehicle 1, and this pitch moment may be used as input data for the neural network 121. The roll moment may be calculated by multiplying the lateral acceleration of the vehicle body 2 by the height of the center of gravity of the vehicle 1, and this roll moment may be used as input data for the neural network 121.

[0061] Furthermore, in the above embodiment, the second physical quantity, the mass of the vehicle 1, may include the occupants and cargo of the vehicle 1. In this case, the mass of the vehicle 1 including the occupants and cargo can be calculated by providing a detection device that detects the mass of the occupants and cargo of the vehicle 1.

[0062] While embodiments have been described in this invention, these are presented as examples only and are not intended to limit the scope of the invention. These novel embodiments can be implemented in various other forms, and various omissions, substitutions, and modifications can be made without departing from the spirit of the invention. These embodiments and their variations are included in the scope and spirit of the invention, as well as in the claims of the invention and its equivalents.

[0063] [Summary of this embodiment] The control device (10) of this embodiment comprises at least the following configurations. In other words, the control device (10) is a control device (10) that controls the suspension system (4) interposed between the vehicle body (2) and the wheels (3) of the vehicle (1), An acquisition unit (10a) acquires the acceleration of the vehicle body (2) in the vertical direction of the vehicle (1), The system includes a neural network (121) trained to estimate the relative velocity of the wheels (3) to the vehicle body (2) in the vertical direction of the vehicle (1) in response to an input value obtained by multiplying the acceleration by the mass of the vehicle (1), and an estimation unit (10b) that estimates the relative velocity corresponding to a value obtained by multiplying the acceleration acquired by the acquisition unit (10a) by a mass that is the same or different in magnitude as when the neural network (121) was trained. The suspension system (4) is controlled based on the estimated relative speed.

[0064] With this configuration, for example, the same neural network (121) can be used for multiple vehicles (1) with different mass specifications. Therefore, even if a vehicle (1) has multiple specifications, the effort required to create the neural network (121) can be reduced.

[0065] Furthermore, the control device (10) of the embodiment is equipped with a shock absorber (4b) interposed between the vehicle body (2) and the wheels (3), and the estimation unit (10b) preferably estimates a new relative speed by multiplying the damping coefficient or a quantity related to the damping coefficient of the shock absorber (4b) by the previously estimated relative speed.

[0066] This configuration reduces the effort required to create the neural network (121) even when there are multiple specifications for the vehicle (1) with different damping coefficients or quantities related to the damping coefficient of the shock absorber (4b).

[0067] Furthermore, the control device (10) of the embodiment is equipped with a spring (4a) interposed between the vehicle body (2) and the wheel (3). The estimation unit (10b) estimates the relative displacement of the wheel (3) with respect to the vehicle body (2) in the vertical direction of the vehicle from the relative speed, and estimates a new relative displacement by multiplying the spring constant of the spring (4a) by the previously estimated relative displacement.

[0068] This configuration reduces the effort required to create the neural network (121) even when there are multiple vehicle (1) models with different spring constants for the spring (4a). [Explanation of Symbols]

[0069] 1…Vehicle 2… Vehicle body 3...wheels 4…Suspension device 4a... Coil spring (spring) 4b... Shock absorber 10...Control device 10a…Acquisition part 10b…Estimation part 121...Neural Network

Claims

1. A control device for controlling a suspension system interposed between the vehicle body and the wheels of the said vehicle, An acquisition unit that acquires the acceleration of the vehicle body in the vertical direction, An estimation unit estimates the relative velocity of the wheels relative to the vehicle body in the vertical direction of the vehicle, using a neural network that has been trained to estimate the relative velocity of the wheels relative to the vehicle body in the vertical direction of the vehicle in response to the input of a numerical value obtained by multiplying the acceleration acquired by the acquisition unit by the mass, which is the same or different in magnitude as when the neural network was trained. Equipped with, A control device that controls the suspension system based on the estimated relative speed.

2. A shock absorber is interposed between the vehicle body and the wheel, The neural network also estimates the relative velocity in response to input of a numerical value obtained by multiplying the damping coefficient of the shock absorber or a quantity related to the damping coefficient by the previously estimated relative velocity. The control device according to claim 1.

3. A spring is interposed between the vehicle body and the wheel. The neural network estimates the relative displacement of the wheels with respect to the vehicle body in the vertical direction from the relative velocity. The neural network also estimates the relative displacement in response to input values ​​obtained by multiplying the spring constant of the spring by the previously estimated relative displacement. The control device according to claim 1.

Images