Vehicle state quantity estimation device

Abstract

A vehicle state quantity estimation device of an embodiment has a data acquisition section and a vehicle state estimation section. The data acquisition section acquires first data relating to a speed of a vehicle. The vehicle state estimation section estimates at least one of a relative speed of a wheel with respect to a vehicle body of the vehicle in a vertical direction of the vehicle and a vehicle body speed of the vehicle body in the vertical direction, using a neural network that is learned to estimate at least one of the relative speed and the vehicle body speed from an input of the first data, in correspondence with the first data acquired by the data acquisition section.

Description

Technical Field

[0001] Embodiments of the present invention relate to a vehicle state quantity estimation device. Background Technology

[0002] Previously, a technique was known that, as a vehicle state, the relative speed of the wheels to the vehicle body in the vertical direction (so-called travel speed) or the speed of the vehicle body in the vertical direction (so-called sprung speed) was estimated, and the damping force of the suspension shock absorbers was controlled based on the estimated relative speed or vehicle body speed.

[0003] Existing technical documents

[0004] Patent documents

[0005] Patent Document 1: Japanese Patent Application Publication No. 10-913

[0006] Patent Document 2: Japanese Patent Application Publication No. 2016-117326 Summary of the Invention

[0007] The problem that the invention aims to solve

[0008] In this technology, it would be meaningful to improve the accuracy of vehicle state estimation.

[0009] Technical solutions to the problem

[0010] The vehicle state estimation device according to an embodiment of the present invention includes: a data acquisition unit for acquiring data related to the speed of a vehicle, i.e., first data; and a vehicle state estimation unit for estimating at least one of the relative speed of the wheels of the vehicle relative to the vehicle body in the vertical direction and the speed of the vehicle body in the vertical direction, i.e., the vehicle body speed, based on the input of the first data, using a first neural network, and estimating at least one of the relative speed and the vehicle body speed corresponding to the first data acquired by the data acquisition unit.

[0011] According to this structure, the vehicle state estimation unit utilizes a first neural network that has been trained to estimate at least one of the relative speed of the wheel relative to the vehicle body in the vertical direction of the vehicle and the speed of the vehicle body in the vertical direction of the vehicle, i.e., the vehicle body speed, based on the input of the first data. It estimates at least one of the relative speed and the vehicle body speed corresponding to the first data acquired by the data acquisition unit, thereby improving the estimation accuracy of the vehicle state (at least one of the relative speed and the vehicle body speed).

[0012] In the vehicle state quantity estimation device, for example, the first data is wheel speed data representing the rotational speed of the wheel.

[0013] Based on this structure, the vehicle state estimation unit utilizes a first neural network that has been trained to estimate at least one of the relative speed of the wheel relative to the vehicle body in the vertical direction and the speed of the vehicle body in the vertical direction, i.e., the vehicle body speed, based on the input of wheel speed data. It estimates at least one of the relative speed and the vehicle body speed corresponding to the wheel speed data acquired by the data acquisition unit, thereby improving the estimation accuracy of the vehicle state (at least one of the relative speed and the vehicle body speed).

[0014] In the vehicle state estimation device, for example, the data acquisition unit acquires yaw rate data representing the yaw rate of the vehicle, and the first neural network learns to estimate at least one of the relative speed and the vehicle body speed based on the wheel speed data and the yaw rate input. The vehicle state estimation unit uses the first neural network to estimate at least one of the relative speed and the vehicle body speed corresponding to the wheel speed data and the yaw rate data acquired by the data acquisition unit.

[0015] Based on this structure, the accuracy of estimating vehicle state (at least one of relative speed and vehicle body speed) can be further improved.

[0016] In the vehicle state estimation device, for example, the vehicle has a shock absorber located between the vehicle body and the wheels, which can change the damping force according to the input current. The data acquisition unit acquires current value data representing the value of the current. The first neural network learns to estimate at least one of the relative speed and the vehicle body speed based on the input of the wheel speed data and the current value data. The vehicle state estimation unit uses the first neural network to estimate at least one of the relative speed and the vehicle body speed corresponding to the wheel speed data and the current value data acquired by the data acquisition unit.

[0017] Based on this structure, the accuracy of estimating vehicle state (at least one of relative speed and vehicle body speed) can be further improved.

[0018] In the vehicle state estimation device, for example, the data acquisition unit acquires steering angle data representing the steering angle of the vehicle, the first neural network learns to estimate at least one of the relative speed and the vehicle body speed based on the input of the wheel speed data and the steering angle data, and the vehicle state estimation unit uses the first neural network to estimate at least one of the relative speed and the vehicle body speed corresponding to the wheel speed data and the steering angle data acquired by the data acquisition unit.

[0019] Based on this structure, the accuracy of estimating vehicle state (at least one of relative speed and vehicle body speed) can be further improved.

[0020] Furthermore, in the vehicle state estimation device, the first data is acceleration data representing the acceleration of the vehicle, the first neural network is learned to estimate the relative speed, and the vehicle state estimation unit uses the first neural network to estimate the relative speed corresponding to the acceleration data acquired by the data acquisition unit.

[0021] Based on this structure, the relative velocity estimation unit utilizes a first neural network that has been trained to estimate the relative velocity of the vehicle's wheels in the vertical direction relative to the vehicle body based on the input of acceleration data. This estimates the relative velocity corresponding to the acceleration data acquired by the data acquisition unit, thereby improving the estimation accuracy of the relative velocity of the wheels relative to the vehicle body.

[0022] In the vehicle state estimation device, for example, the data acquisition unit acquires yaw rate data representing the yaw rate of the vehicle, the first neural network learns to estimate the relative speed based on the input of the acceleration data and the yaw rate data, and the vehicle state estimation unit uses the first neural network to estimate the relative speed corresponding to the acceleration data and the yaw rate data acquired by the data acquisition unit.

[0023] This structure allows for further improvement in the accuracy of estimating the relative speed of the wheels to the vehicle body.

[0024] In the vehicle state estimation device, for example, the vehicle has a shock absorber located between the vehicle body and the wheels, which can change the damping force according to the input current. The data acquisition unit acquires current value data representing the value of the current. The first neural network learns to estimate the relative speed based on the input of the acceleration data and the current value data. The vehicle state estimation unit uses the first neural network to estimate the relative speed corresponding to the acceleration data and the current value data acquired by the data acquisition unit.

[0025] This structure allows for further improvement in the accuracy of estimating the relative speed of the wheels to the vehicle body.

[0026] In the vehicle state estimation device, for example, the data acquisition unit acquires steering angle data representing the steering angle of the vehicle, the first neural network learns to estimate the relative speed based on the input of the acceleration data and the steering angle data, and the vehicle state estimation unit uses the first neural network to estimate the relative speed corresponding to the acceleration data and the steering angle data acquired by the data acquisition unit.

[0027] This structure allows for further improvement in the accuracy of estimating the relative speed of the wheels to the vehicle body.

[0028] In the vehicle state estimation device, for example, the data acquisition unit acquires wheel speed data representing the rotational speed of the wheel, the first neural network learns to estimate the relative speed based on the input of the acceleration data and the wheel speed data, and the vehicle state estimation unit uses the first neural network to estimate the relative speed corresponding to the acceleration data and the wheel speed data acquired by the data acquisition unit.

[0029] This structure allows for further improvement in the accuracy of estimating the relative speed of the wheels to the vehicle body.

[0030] In the vehicle state quantity estimation device, for example, the data acquisition unit includes: an acquisition unit that acquires actual acceleration data, representing the actual acceleration of the acceleration detection target parts, from acceleration sensors that are only installed on a portion of multiple acceleration detection target parts located at different positions on the vehicle; and an estimation unit that uses a second neural network that has been learned to estimate the actual acceleration of the acceleration detection target parts detected by each acceleration sensor when the multiple acceleration detection target parts are equipped with the acceleration sensors, based on the input of the actual acceleration data, to estimate all the actual accelerations of the multiple acceleration detection target parts corresponding to the input of the actual acceleration data.

[0031] This structure can improve the accuracy of acceleration data and further improve the estimation accuracy of the relative velocity of the wheels relative to the vehicle body.

[0032] In the vehicle state estimation device, for example, the first neural network is an LSTM (Long Short-Term Memory) network.

[0033] This structure can improve the accuracy of estimating vehicle state (at least one of relative speed and vehicle speed).

[0034] The vehicle state quantity estimation device according to an embodiment of the present invention includes: an acquisition unit that acquires actual acceleration data, i.e., actual acceleration, representing the actual acceleration of a plurality of acceleration detection target parts that are only installed on a vehicle at different locations, from acceleration sensors. The estimation unit uses a neural network that has been learned to estimate the actual acceleration of the acceleration detection target parts detected by each acceleration sensor when the acceleration sensors are installed at all of the plurality of acceleration detection target parts. The estimation unit estimates the actual acceleration of all of the plurality of acceleration detection target parts corresponding to the input of the actual acceleration data.

[0035] Based on this structure, the accuracy of vehicle acceleration estimation can be improved. Attached Figure Description

[0036] Figure 1 This is a schematic diagram showing a schematic structure of an example of a vehicle according to the first embodiment.

[0037] Figure 2 This is a schematic diagram showing a schematic structure of an example of the suspension system of a vehicle according to the first embodiment.

[0038] Figure 3 This is a functional block diagram of a control device for an example of a vehicle according to the first embodiment.

[0039] Figure 4 This is an exemplary and schematic block diagram illustrating the structure of the vehicle state estimation network of the first embodiment.

[0040] Figure 5 This is an exemplary and schematic block diagram showing an example of the structure of the encoder section of the vehicle state estimation network of the first embodiment.

[0041] Figure 6 This is an exemplary and schematic block diagram illustrating an example of the structure of the decoder section of the vehicle state estimation network according to the first embodiment.

[0042] Figure 7 This is an exemplary and schematic block diagram illustrating an example of the structure of an LSTM block of the vehicle state estimation network of the first embodiment.

[0043] Figure 8 This is a flowchart illustrating an example of a vehicle state estimation method executed by the control device of the first embodiment.

[0044] Figure 9 This is an exemplary and schematic diagram illustrating an example of the estimated result of the travel speed based on the technology of the first embodiment.

[0045] Figure 10 This is a schematic diagram showing a schematic structure of an example of a vehicle according to the second embodiment.

[0046] Figure 11 This is a functional block diagram of a control device for an example of a vehicle according to the second embodiment.

[0047] Figure 12 This is an exemplary and schematic block diagram illustrating the structure of the travel speed estimation network of the second embodiment.

[0048] Figure 13 This is a flowchart illustrating an example of a relative speed estimation method executed by the control device of the second embodiment.

[0049] Figure 14 This is an exemplary and schematic diagram illustrating an example of the estimated result of the travel speed based on the technology of the second embodiment.

[0050] Figure 15 This is an exemplary and schematic diagram illustrating one example of the estimated result of the travel speed based on the comparative example technology.

[0051] Figure 16 This is an exemplary and schematic diagram illustrating an example of the estimated error of the travel speed based on the technology of the second embodiment.

[0052] Figure 17 This is an exemplary and schematic diagram illustrating an example of the estimated error of the travel speed based on the technology of the second embodiment.

[0053] Figure 18 This is a schematic diagram showing a schematic structure of an example of a vehicle according to the third embodiment.

[0054] Figure 19 This is a functional block diagram of a control device for an example of a vehicle according to the third embodiment.

[0055] Figure 20 This is an exemplary and schematic block diagram illustrating the structure of the stroke speed estimation network and the sprung acceleration estimation network of the third embodiment.

[0056] Figure 21 This is an exemplary and schematic diagram illustrating an example of the estimated result of sprung acceleration based on the technology of the third embodiment.

[0057] Figure 22 This is an exemplary and schematic diagram illustrating one example of the estimated result of sprung acceleration based on the comparative example technique. Detailed Implementation

[0058] Hereinafter, embodiments will be described in detail with reference to the accompanying drawings. However, the present invention is not limited to these embodiments. The same structural components are included in several of the following embodiments. These same structural components are labeled with the same reference numerals, and repeated descriptions are omitted.

[0059] <First Implementation>

[0060] Figure 1 This is a schematic diagram showing a schematic structure of an example of the vehicle 1 according to the first embodiment. In this embodiment, the vehicle 1 may be, for example, a car driven by an internal combustion engine (engine, not shown) (internal combustion engine car), a car driven by an electric motor (motor, not shown) (electric car, fuel cell car, etc.), or a car driven by both (hybrid car). Furthermore, the vehicle 1 may be equipped with various transmission devices, and may also be equipped with various devices (systems, components, etc.) required to drive the internal combustion engine or the electric motor. Additionally, the manner, number, and layout of the devices related to the drive of the wheels 3 in the vehicle 1 can be configured in various ways. Furthermore, in this embodiment, as an example, the vehicle 1 is a four-wheeled vehicle (four-wheeled car) having two front wheels 3F and two rear wheels 3R. Moreover, in... Figure 1 In the middle, the front (direction Fr) of the vehicle in the front-to-back direction is the left side.

[0061] 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, and a braking system 61. Furthermore, the vehicle control system 100 includes suspension devices 4, rotation sensors 5, and braking devices 6 corresponding to the two front wheels 3F, and suspension devices 4 and rotation sensors 5 corresponding to the two rear wheels 3R. In addition, vehicle 1, besides... Figure 1 In addition to the structural components, the vehicle 1 also has basic structural components, but here, only the structures related to the vehicle control system 100 and the controls related to these structures will be described. Furthermore, the term "wheel 3" is used collectively for the two front wheels 3F and the two rear wheels 3R. The relative speed of the wheel 3 of the vehicle 1 with respect to the vehicle body 2 is called the travel speed; the portion of the vehicle 1 located on the side of the vehicle body 2 relative to the suspension device 4 is called the unsprung speed; and the portion of the vehicle 1 located on the side of the wheel 3 relative to the suspension device 4 is called the unsprung speed. In the following description, unless otherwise specified, relative speed refers to the relative speed of the wheel 3 with respect to the vehicle body 2 in the vertical direction of the vehicle 1.

[0062] The control device 10 receives signals and data from various parts of the vehicle control system 100 and executes control and various calculations for each part of the vehicle control system 100. In this embodiment, the control device 10 is an example of a vehicle state quantity estimation device. Furthermore, the control device 10 is composed of a computer and includes a processing unit (microcomputer, ECU (Electronic Control Unit), etc., not shown), a storage unit 10d (e.g., ROM (Read Only Memory), RAM (Random Access Memory), flash memory, etc., see reference 10). Figure 3 The arithmetic processing unit can read the program stored (installed) in the non-volatile storage unit 10d (e.g., ROM, flash memory, etc.), and perform arithmetic processing according to the program, as... Figure 3 Each of the shown parts performs its function (operates). Additionally, the storage unit 10d can store data (tables (data sets), functions, etc.) used in various control-related calculations, as well as calculation results (including values ​​during the calculation process). Furthermore, the storage unit 10d stores the vehicle state estimation network 121. The vehicle state estimation network 121 will be described in detail later.

[0063] The steering device 11 includes, for example, a steering wheel, which steers (rotates) the two front wheels 3F. The steering angle sensor 12 detects the steering angle (rudder angle, steering angle, steering angle) of the front wheels 3F and outputs steering angle data representing the detected steering angle.

[0064] Yaw rate sensor 13 detects the yaw rate of vehicle 1 (vehicle body 2) and outputs yaw rate information representing the detected yaw rate.

[0065] Figure 2 This is a schematic diagram showing a schematic structure of an example of the suspension device 4 of the vehicle according to the first embodiment. Figure 2As shown, the suspension device 4 is located between the wheels 3 and the vehicle body 2, suppressing the transmission of vibrations or impacts from the road surface to the vehicle body. The suspension device 4 has a coil spring 4a and a shock absorber 4b. The shock absorber 4b can electrically control (adjust) the damping force (damping characteristics). Specifically, the shock absorber 4b has an actuator 4bb that operates based on the input current. The actuator 4bb can change the opening of the throttle orifice of the piston provided in the shock absorber 4b, or change the opening between the valve body and the valve seat. Thus, the flow of lubricating oil between the two oil chambers separated by the piston in the shock absorber 4b is controlled, and the damping force of the shock absorber 4b is adjusted. The suspension device 4 is respectively provided on 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. In detail, in order to control the damping force of the shock absorber 4b (=damping coefficient C×stroke speed V), the damping coefficient C is controlled. More specifically, the damping coefficient C is controlled by controlling the current flowing through the actuator 4bb of the shock absorber 4b. The shock absorber 4b is also known as a damper.

[0066] The rotation sensor 5 can output signals corresponding to the rotational speed (angular velocity, revolutions, rotational state) of each of the four wheels 3. The control device 10 can calculate the speed of the vehicle 1 based on 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. can also be installed, and the control device 10 can also obtain the speed of the vehicle 1 based on the detection results of the rotation sensor.

[0067] Furthermore, the structure of the vehicle control system 100 described above is merely an example, and various modifications can be made to implement it. Known devices can be used as the various components constituting the vehicle control system 100. Additionally, the various structures of the vehicle control system 100 can be shared with other structures.

[0068] Furthermore, in this embodiment, as an example, the control device 10, through the cooperation of hardware and software (program), can function as such Figure 3 The data acquisition unit 10a, vehicle state estimation unit 10b, and attenuation control unit 10c shown perform their functions (operate). That is, in the program, as an example, it may include... Figure 3 The modules shown are those corresponding to each block except for the storage section 10d.

[0069] The data acquisition unit 10a acquires wheel speed data, representing the rotational speed of the wheel 3, from each rotation sensor 5. Additionally, the data acquisition unit 10a acquires yaw rate data, representing the yaw rate of the vehicle 1, from the yaw rate sensor 13. Furthermore, the data acquisition unit 10a acquires current value data, representing the value of the current input to the shock absorber 4b, from the damping control unit 10c. Finally, the data acquisition unit 10a acquires steering angle data, representing the steering angle of the vehicle 1, from the steering angle sensor 12.

[0070] As an example, the prescribed data (hereinafter referred to as input data) acquired by the data acquisition unit 10a is input to the vehicle state estimation unit 10b. The vehicle state estimation unit 10b uses the vehicle state estimation network 121 to estimate the relative speed corresponding to the input data. The input data includes at least wheel speed data. In addition to wheel speed data, the input data may also include at least one of yaw rate data, current value data, and steering angle data. Furthermore, the input data is not limited to the data described above. Wheel speed data is an example of the first type of data.

[0071] The vehicle state estimation network 121 is configured as a recurrent neural network (RNN) that has been pre-learned to estimate at least one of the relative velocities of the wheels 3 of vehicle 1 in the vertical direction relative to the body 2 of vehicle 1 and the velocity of the body 2 in the vertical direction of vehicle 1 (i.e., the vehicle body velocity), based on the input data. During the learning process of the vehicle state estimation network 121, the input data is the aforementioned input data acquired by the data acquisition unit 10a. The output data, serving as teacher data, is, for example, at least one of the measured values ​​(true values) of the relative velocities of the wheels 3 of vehicle 1 in the vertical direction relative to the body 2 of vehicle 1 measured by a relative speed sensor (not shown) and the measured values ​​of the velocity of the body 2 in the vertical direction of vehicle 1 measured by a speed sensor (not shown) (i.e., both). Alternatively, the vehicle state estimation network 121 may also be configured as a recurrent neural network (RNN) that has been pre-learned to estimate either the relative velocity or the vehicle body velocity.

[0072] Figure 4 This is an exemplary and schematic block diagram illustrating the structure of the vehicle state estimation network 121 of the first embodiment. Figure 5 This is an exemplary and schematic block diagram showing an example of the structure of the encoder section 121a of the vehicle state estimation network 121 of the first embodiment. Figure 6 This is an exemplary and schematic block diagram showing an example of the structure of the decoder unit 121b of the vehicle state estimation network 121 of the first embodiment. Figure 7 This is an exemplary and schematic block diagram illustrating an example of the structure of an LSTM block of the vehicle state estimation network 121 of the first embodiment.

[0073] like Figure 4 As shown, the vehicle state estimation network 121 of the embodiment is configured as a Seq2Seq (Sequence to Sequence) model. More specifically, the vehicle state estimation network 121 has: an encoder unit 121a, which receives input data and performs encoding processing; and a decoder unit 121b, which performs decoding processing based on the encoder result of the encoder unit 121a and outputs the data that has undergone decoding processing (hereinafter also referred to as output data).

[0074] exist Figure 4 In the example shown, as input to encoder unit 121a, input data x1, x2, x3, x4, x5, ... are shown input at predetermined time intervals.

[0075] On the other hand, Figure 5 In the example shown, output data y1, y2, y3, y4, y5, ... are shown as outputs from the decoder unit 121b, output at predetermined time intervals.

[0076] The vehicle state estimation network 121 of the implementation method is pre-trained by machine learning to receive input time series data such as the input data x1, x2, x3, x4, x5, ... as described above, and output time series data representing the output at each time point such as the output data y1, y2, y3, y4, y5, ... as described above.

[0077] The following is for reference Figures 5-7 The specific structures of the encoder section 121a and decoder section 121b of the vehicle state estimation network 121 in the first embodiment will be described. However, the following... Figure 5 and Figure 6 The structure shown is just one example.

[0078] like Figure 5 As shown, the encoder unit 121a of the first embodiment is constructed based on LSTM (Long Short-Term Memory) network. That is, the encoder unit 121a uses multiple (e.g., N) LSTM blocks B. 11 B 12 ...B 1N And thus constituted. Later, as LSTM block B 11 B 12 ...B 1N The collective term, sometimes referred to as LSTM block B1. Each LSTM block B... 11 B 12 ...B 1NThe structure is a general structure with input gates, output gates, and forget gates. As an example, such as... Figure 7 As shown, LSTM block B1 includes matrix multiplication units 201a and 201b, addition units 201c, 201d, and 201k, multiplication units 201j and 201p, a slice unit 201e, sigmoid function units 201f, 201h, and 201i, and a tanh (hyperbolic tangent) unit 201g. Operations are performed by these units. The addition units 201c, 201d, and 201k, and the multiplication units 201j and 201p perform operations (Hadamard operations) on each matrix element. Furthermore, Figure 7 In this context, Wh represents the weights assigned to the hidden layer, and Wx represents the weights assigned to the input values.

[0079] LSTM Block B 11 Receive input data x1, and pass the data h1 representing the output corresponding to the input and the data c1 representing the storage unit to the next LSTM block B. 12 LSTM block B 12 Subsequent blocks follow the same procedure, with the Nth LSTM block B... 1N Received data x N The input is given, and the data h representing the output corresponding to the input is given. N and the data c representing the storage unit N The signal is transmitted to the outside of the encoder unit 121a (i.e., the decoder unit 121b).

[0080] like Figure 6 As shown, similar to the encoder unit 121a described above, the decoder unit 121b in this embodiment is also based on LSTM. That is, the decoder unit 121b also uses multiple (e.g., N) LSTM blocks B. 21 B 22 ...B 2N And constitutes it. Similar to the aforementioned LSTM block B. 11 B 12 ...B 1N The same applies to each LSTM block B. 21 B 22 ...B 2N The structure is a general structure with input gates, output gates, and forget gates.

[0081] However, unlike the encoder section 121a described above, the decoder section 121b has a connection with the LSTM block B. 21 B 22 ...B 2N The corresponding multiple (e.g., N) transformation layers L1, L2, ... L N These transition layers L1, L2, ... L NRepresenting the source from LSTM block B respectively 21 B 22 ...B 2N The output data H1, H2, ..., H N Converted into output data y1, y2, ... y corresponding to the sensor signals N Here, the output data y = (data H) × (weight W) + (bias b). The weights W and bias b are obtained through machine learning.

[0082] LSTM Block B 21 Receive data h from encoder unit 121a N and c N The input is passed to the transformation layer L1, which transmits the data H1 representing the output corresponding to the input, and then passes the data H1 and the data C1 representing the storage unit to the next LSTM block B. 22 LSTM block B 22 Subsequent blocks operate in the same way, the Nth LSTM block B 2N The data H represents the output corresponding to the input. N Passed to the transformation layer L N .

[0083] Thus, the vehicle state estimation network 121 in this implementation is based on a recurrent neural network, which is constructed from an LSTM-based Seq2Seq model. More specifically, in this implementation, execution should be performed in LSTM block B. 11 B 12 ...B 1N LSTM block B 21 B 22 ...B 2N Transition layers L1, L2, ... L N The training of the recurrent neural network with set weights W and biases b, enabling it to learn the correspondence between input and output data, constitutes the vehicle state estimation network 121. Furthermore, the vehicle state estimation network 121 is not limited to the above. For example, in addition to the above, the vehicle state estimation network 121 can also employ Attention, RNN (Recurrent Neural Network), bidirectional LSTM, skip connections, Transformer (a sequence model based on attention mechanisms), etc.

[0084] Return to Figure 3The damping control unit 10c controls the damping force of the shock absorber 4b based on at least one of the relative speed and the vehicle speed (for example, both) estimated by the vehicle state estimation unit 10b. Specifically, the damping control unit 10c determines the current value input to the shock absorber 4b and inputs the current value into the shock absorber 4b.

[0085] Next, refer to Figure 8 This describes an example of a vehicle state estimation method executed by control device 10. Figure 8 This is a flowchart illustrating an example of a vehicle state estimation method executed by the control device 10 of the first embodiment. Furthermore, Figure 8 An example of using wheel speed data as input data is shown, but the input data is not limited to this.

[0086] like Figure 8 As shown, firstly, the data acquisition unit 10a acquires wheel speed data (step S1). Next, the vehicle state estimation unit 10b estimates at least one of the travel speed and the vehicle body speed (for example, both) as relative speeds based on the wheel speed data acquired by the data acquisition unit 10a and the vehicle state estimation network 121 (step S2).

[0087] Figure 9 This is an exemplary and schematic diagram illustrating an example of the estimated travel speed based on the technology of the first embodiment. Figure 9 The diagram shows the estimated stroke speed (estimated value) and the actual measured stroke speed (true value) using the technique of this embodiment. In this embodiment, it is possible to estimate the stroke speed of a waveform of 10Hz or higher that is effective in controlling the damping force of the shock absorber 4b.

[0088] Furthermore, the control device 10 (damping control unit 10c) of this embodiment controls the shock absorber 4b in the following manner: The control device 10 controls the damping force of the shock absorber 4b based on the relative speed and vehicle speed (sprung speed) estimated using the vehicle state estimation network 121. More specifically, the control device 10 obtains the target damping force by performing known ride comfort controls such as skyhook control based on the vehicle speed. Then, the control device 10 calculates the target damping coefficient of the shock absorber 4b based on the relative speed estimated using the vehicle state estimation network 121 and the target damping force. Next, the control device 10 inputs a current corresponding to the target damping coefficient to the shock absorber 4b.

[0089] As described above, in this embodiment, the control device 10 (vehicle state quantity estimation device) includes a data acquisition unit 10a and a vehicle state estimation unit 10b (vehicle state estimation unit). The data acquisition unit 10a acquires wheel speed data representing the rotational speed of the wheels 3 of the vehicle 1. The vehicle state estimation unit 10b utilizes a vehicle state estimation network 121 (neural network) that has been trained to estimate at least one of the relative speed (travel speed) of the wheels 3 of the vehicle 1 relative to the body 2 of the vehicle 1 in the vertical direction and the speed of the body 2 in the vertical direction of the vehicle 1, i.e., the speed of the body, based on the input of the wheel speed data, to estimate at least one of the relative speed (travel speed) and the speed of the body corresponding to the wheel speed data acquired by the data acquisition unit 10a.

[0090] According to this structure, since the vehicle state estimation unit 10b utilizes a vehicle state estimation network 121 (neural network) that has been trained to estimate at least one of the relative speed of the wheel 3 of vehicle 1 relative to the body 2 of vehicle 1 in the vertical direction of vehicle 1 and the speed of the body 2 (i.e., the speed of the body), based on the input wheel speed data, it estimates at least one of the relative speed (travel speed) and the speed of the body corresponding to the wheel speed data acquired by the data acquisition unit 10a. Therefore, the estimation accuracy of the vehicle state (at least one of the relative speed and the speed of the body) can be improved. In addition, during training, the vehicle state estimation network 121 also learns the input delay (time lag) of the input data caused by communication delay as a feature. Therefore, in the vehicle state estimation network 121, for example, it is not necessary to remove a portion of the input data by means of filters or the like to suppress the influence of communication delay in the estimation of the vehicle state. Therefore, the vehicle state estimation network 121 can estimate the vehicle state even at time intervals corresponding to the resonant frequency of the unsprung surface.

[0091] In addition, in this embodiment, the vehicle state estimation network 121 is an LSTM.

[0092] This structure can improve the accuracy of estimating vehicle state (at least one of relative speed and vehicle speed).

[0093] In this embodiment, the data acquisition unit 10a acquires yaw rate data representing the yaw rate of the vehicle 1. The vehicle state estimation network 121 learns to estimate at least one of relative speed and vehicle body speed based on the input wheel speed data and yaw rate data. The vehicle state estimation unit 10b uses the vehicle state estimation network 121 to estimate at least one of relative speed and vehicle body speed corresponding to the wheel speed data and yaw rate data acquired by the data acquisition unit 10a.

[0094] Based on this structure, the accuracy of estimating vehicle state (at least one of relative speed and vehicle body speed) can be further improved.

[0095] In this embodiment, vehicle 1 has a shock absorber 4b located between vehicle body 2 and wheels 3, which can change its damping force according to the input current. Data acquisition unit 10a acquires current value data representing the current value. Vehicle state estimation network 121 learns to estimate at least one of relative speed and vehicle body speed based on the input wheel speed data and current value data. Vehicle state estimation unit 10b uses vehicle state estimation network 121 to estimate at least one of relative speed and vehicle body speed corresponding to the wheel speed data and current value data acquired by data acquisition unit 10a.

[0096] Based on this structure, the accuracy of estimating vehicle state (at least one of relative speed and vehicle body speed) can be further improved.

[0097] In this embodiment, the data acquisition unit 10a acquires steering angle data representing the steering angle of the vehicle 1. The vehicle state estimation network 121 learns to estimate at least one of relative speed and vehicle body speed based on the input wheel speed data and steering angle data. The vehicle state estimation unit 10b uses the vehicle state estimation network 121 to estimate at least one of relative speed and vehicle body speed corresponding to the wheel speed data and steering angle data acquired by the data acquisition unit 10a.

[0098] Based on this structure, the accuracy of estimating vehicle state (at least one of relative speed and vehicle body speed) can be further improved.

[0099] <Second Implementation>

[0100] Figure 10 This is a schematic diagram showing a schematic structure of an example of the vehicle 1 according to the second embodiment. The difference between this embodiment and the first embodiment is that it includes an acceleration sensor 6 and a control device 10 for processing.

[0101] 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. Furthermore, the vehicle control system 100 includes suspension devices 4, rotation sensors 5, acceleration sensors 6, etc., corresponding to the two front wheels 3F respectively, and suspension devices 4, rotation sensors 5, acceleration sensors 6, etc., corresponding to the two rear wheels 3R ​​respectively.

[0102] Accelerometers 6 are respectively installed corresponding to the two front wheels 3F and the two rear wheels 3R. That is, acceleration sensors 6 are respectively installed on the vehicle body 2 corresponding to the four wheels 3 and the suspension device 4. Accelerometers 6 detect the acceleration of the vehicle body 2 in the vehicle 1. Specifically, each acceleration sensor 6 is installed on the vehicle body 2 at an acceleration detection target location located directly above each wheel 3 (directly above the wheel hub). That is, in this embodiment, there are four acceleration detection target locations. In this embodiment, as an example, acceleration sensors 6 can acquire the vertical acceleration of the vehicle 1, the longitudinal acceleration of the vehicle 1 (length direction), and the width acceleration of the vehicle 1 (vehicle width direction, width direction, left and right direction).

[0103] Furthermore, in this embodiment, as an example, the control device 10, through the cooperation of hardware and software (program), can function as such Figure 11 The data acquisition unit 10a, stroke speed estimation unit 10bA, and attenuation control unit 10c shown perform their functions (operate). That is, in the program, as an example, it may include... Figure 11 The modules shown are those corresponding to each block except for the storage section 10d.

[0104] The data acquisition unit 10a acquires acceleration data representing the acceleration of vehicle 1 from each acceleration sensor 6. The acceleration of vehicle 1 can be in any direction. Furthermore, the acceleration of vehicle 1 can be in multiple directions. That is, the acceleration of vehicle 1 can be one or more of the following: acceleration in the forward / backward direction, acceleration in the left / right direction, and acceleration in the up / down direction. Additionally, the data acquisition unit 10a acquires yaw rate data representing the yaw rate of vehicle 1 from the yaw rate sensor 13. Additionally, the data acquisition unit 10a acquires current value data representing the value of the current input to the shock absorber 4b from the damping control unit 10c. Additionally, the data acquisition unit 10a acquires steering angle data representing the steering angle of vehicle 1 from the steering angle sensor 12. Additionally, the data acquisition unit 10a acquires wheel speed data representing the rotational speed of wheel 3 from the rotation sensor 5. The acceleration data is an example of the first type of data related to the speed of vehicle 1.

[0105] As an example, the prescribed data (hereinafter referred to as input data) acquired by the data acquisition unit 10a is input to the stroke speed estimation unit 10bA. The stroke speed estimation unit 10bA uses the stroke speed estimation network 121A to estimate the relative speed corresponding to the input data. The input data includes at least acceleration data. In addition to acceleration data, the input data may also include at least one of yaw rate data, current value data, steering angle data, and wheel speed data. Furthermore, the input data is not limited to the data described above.

[0106] The travel speed estimation network 121A is configured as a recurrent neural network (RNN) that has been pre-learned to estimate the relative speed of the wheels 3 of vehicle 1 in the vertical direction relative to the body 2 of vehicle 1 based on the input data. During the learning process of the travel speed estimation network 121A, the input data is the aforementioned input data acquired by the data acquisition unit 10a, and the output data, serving as teacher data, is, for example, the measured value (true value) of the relative speed of the wheels 3 of vehicle 1 relative to the body 2 of vehicle 1, measured using a relative speed sensor (not shown). The travel speed estimation network 121A is an example of a first neural network.

[0107] Figure 12 This is an exemplary and schematic block diagram illustrating the structure of the travel speed estimation network 121A according to the second embodiment.

[0108] like Figure 12 As shown, the travel speed estimation network 121A of this embodiment is configured as a Seq2Seq (Sequence to Sequence) model. More specifically, the travel speed estimation network 121A has the same structure as the vehicle state estimation network 121 of the first embodiment, namely the encoder unit 121a and the decoder unit 121b.

[0109] Return to Figure 11 The damping control unit 10c controls the damping force of the shock absorber 4b based on the relative speed estimated by the stroke speed estimation unit 10bA. Specifically, the damping control unit 10c determines the current value input to the shock absorber 4b and inputs the current value into the shock absorber 4b.

[0110] Next, refer to Figure 13 This illustrates an example of a relative velocity estimation method executed by the control device 10. Figure 13 This is a flowchart illustrating an example of the relative speed estimation method executed by the control device 10 of the second embodiment. Furthermore, Figure 13 An example of using acceleration data as input data is shown, but the input data is not limited to this.

[0111] like Figure 13 As shown, firstly, the data acquisition unit 10a acquires acceleration data (S11). Next, the stroke speed estimation unit 10bA estimates the stroke speed as a relative speed based on the acceleration data acquired by the data acquisition unit 10a and the stroke speed estimation network 121A (S12).

[0112] Figure 14 This is an exemplary and schematic diagram illustrating an example of the estimated travel speed based on the technology of the second embodiment. Figure 14The diagram shows the estimated stroke speed (estimated value) and the actual measured stroke speed (true value) using the technique of this embodiment.

[0113] Figure 15 This is an exemplary and schematic diagram illustrating one example of the estimated stroke speed result based on the comparative example technology. Figure 15 The diagram shows the estimated stroke speed (estimated value) and the actually measured stroke speed (true value) using the comparative example technique. The comparative example technique estimates the stroke speed based on the generated equations of motion. Figure 14 and Figure 15 It can be seen that, compared with the technology of the comparative example, the error between the estimated stroke speed and the actual measured stroke speed is smaller in this embodiment.

[0114] Figure 16 This is an exemplary and schematic diagram illustrating an example of the estimated error of the travel speed based on the technology of the second embodiment. Figure 16 The horizontal axis represents the number of learning iterations. Figure 16 The vertical axis represents the estimation error of the relative velocity relative to the true value. Figure 16 The line Q1 represents the estimated error when the input is the vertical acceleration corresponding to a wheel 3. Figure 16 The line Q2 represents the estimated error when the acceleration in the up and down directions corresponding to the four wheels 3 is input. Figure 16 Line Q4 represents the estimation error given the input of the vertical acceleration corresponding to the four wheels 3, and the longitudinal and lateral acceleration corresponding to one wheel 3. Furthermore, Figure 16 The acceleration in the example is just one example; the input acceleration is not limited to this.

[0115] Figure 17 This is an exemplary and schematic diagram illustrating an example of the estimated error in the stroke speed based on the technology of the second embodiment. Figure 17 The figure shows the estimated error of the travel speed corresponding to a single wheel (for example, the front wheel). Figure 17 The horizontal axis represents the number of learning iterations. Figure 17 The vertical axis represents the estimation error of the relative velocity relative to the true value. Figure 17 The line Q5 represents the estimated error when the input is the vertical acceleration corresponding to a wheel 3. Figure 17 The line Q6 represents the estimated error given the input of the vertical acceleration corresponding to a wheel 3 and the input current value of the shock absorber 4b to a wheel 3. Figure 17 Line Q7 represents the estimated error assuming the input of the vertical acceleration corresponding to the four wheels 3 and the input current value to the shock absorbers 4b of the four wheels 3. Furthermore, Figure 17The acceleration in the example is just one example; the input acceleration is not limited to this.

[0116] Furthermore, the control device 10 (damping control unit 10c) of this embodiment controls the shock absorber 4b in the following manner: The control device 10 controls the damping force of the shock absorber 4b based on the relative speed estimated using the travel speed estimation network 121A. More specifically, the control device 10 calculates the speed of the vehicle body 2 (hereinafter also referred to as sprung speed) by integrating the acceleration data from the acceleration sensor 6. Next, a target damping force is obtained by performing known ride comfort controls such as ceiling control based on the speed of the vehicle body 2. Then, the control device 10 calculates the target damping coefficient of the shock absorber 4b based on the relative speed estimated using the travel speed estimation network 121A and the target damping force. Finally, the control device 10 inputs a current corresponding to the target damping coefficient to the shock absorber 4b.

[0117] As described above, in this embodiment, the control device 10 (vehicle state quantity estimation device) includes a data acquisition unit 10a and a travel speed estimation unit 10bA (relative speed estimation unit, vehicle state estimation unit). The data acquisition unit 10a acquires acceleration data representing the acceleration of the vehicle 1. The travel speed estimation unit 10bA uses a travel speed estimation network 121A (first neural network) that has been learned to estimate the relative speed of the wheels 3 of the vehicle 1 relative to the body 2 of the vehicle 1 in the vertical direction based on the input of the acceleration data to estimate the relative speed (travel speed) corresponding to the acceleration data acquired by the data acquisition unit 10a.

[0118] According to this structure, since the stroke speed estimation unit 10bA uses the stroke speed estimation network 121A, which has been trained to estimate the relative speed of the wheel 3 of the vehicle 1 in the vertical direction of the vehicle 1 relative to the body 2 of the vehicle 1 based on the input of acceleration data, the relative speed corresponding to the acceleration data acquired by the data acquisition unit 10a can be estimated, thus improving the estimation accuracy of the relative speed of the wheel 3 relative to the body 2.

[0119] In addition, in this embodiment, the travel speed estimation network 121A is an LSTM.

[0120] Based on this structure, the estimation accuracy of the relative speed of wheel 3 relative to vehicle body 2 can be improved.

[0121] In this embodiment, the data acquisition unit 10a acquires yaw rate data representing the yaw rate of the vehicle 1. The travel speed estimation network 121A learns to estimate the relative speed based on the input acceleration data and yaw rate data. The travel speed estimation unit 10bA uses the travel speed estimation network 121A to estimate the relative speed corresponding to the acceleration data and yaw rate data acquired by the data acquisition unit 10a.

[0122] Based on this structure, the estimation accuracy of the relative speed of wheel 3 relative to vehicle body 2 can be further improved.

[0123] In this embodiment, vehicle 1 has a shock absorber 4b located between vehicle body 2 and wheel 3, which can change its damping force according to the input current. Data acquisition unit 10a acquires current value data representing the current value. Stroke speed estimation network 121A learns to estimate relative speed based on the input acceleration data and current value data. Stroke speed estimation unit 10bA uses stroke speed estimation network 121A to estimate the relative speed corresponding to the acceleration data and current value data acquired by data acquisition unit 10a.

[0124] Based on this structure, the estimation accuracy of the relative speed of wheel 3 relative to vehicle body 2 can be further improved.

[0125] In this embodiment, the data acquisition unit 10a acquires steering angle data representing the steering angle of the vehicle 1. The travel speed estimation network 121A learns to estimate the relative speed based on the input acceleration data and steering angle data. The travel speed estimation unit 10bA uses the travel speed estimation network 121A to estimate the relative speed corresponding to the acceleration data and steering angle data acquired by the data acquisition unit 10a.

[0126] Based on this structure, the estimation accuracy of the relative speed of wheel 3 relative to vehicle body 2 can be further improved.

[0127] In this embodiment, the data acquisition unit 10a acquires wheel speed data representing the rotational speed of the wheel 3. The stroke speed estimation network 121A learns to estimate the relative speed based on the input acceleration data and wheel speed data. The stroke speed estimation unit 10bA uses the stroke speed estimation network 121A to estimate the relative speed corresponding to the acceleration data and wheel speed data acquired by the data acquisition unit 10a.

[0128] Based on this structure, the estimation accuracy of the relative speed of wheel 3 relative to vehicle body 2 can be further improved.

[0129] Furthermore, in the above embodiment, an example is shown where one acceleration sensor 6 is provided for each of the four wheels 3, but the number and mounting position of the acceleration sensors 6 are not limited to this. For example, the acceleration sensors 6 may also be mounted on only three of the four wheels 3, excluding one designated wheel (e.g., the left rear wheel 3R). In this case, for example, it can be assumed that the vehicle body 2 is a rigid body, and the acceleration of the wheel without an acceleration sensor 6 can be geometrically estimated based on the output values ​​of the acceleration sensors 6 provided on the three wheels. In addition, the acceleration sensors 6 are preferably mounted directly above the wheel 3 (directly above the wheel hub), but if the acceleration sensors 6 are mounted at a position slightly away from directly above the wheel 3, the acceleration directly above the wheel 3 can also be estimated based on the output values ​​of the acceleration sensors 6.

[0130] <Third Implementation Method>

[0131] Figure 18 This is a schematic diagram showing a schematic structure of an example of the vehicle 1 according to the third embodiment. The difference between this embodiment and the second embodiment lies in the number of acceleration sensors 6 and the processing performed by the control device 10.

[0132] like Figure 18 As shown, the acceleration sensor 6 is installed on only three of the four wheels 3. For example, the acceleration sensor 6 is installed on the two front wheels 3F and the left rear wheel 3L. That is, no acceleration sensor 6 is installed on the right rear wheel 3R. In this embodiment, the acceleration sensor 6 is installed on only three of the four acceleration detection target locations on the vehicle body 2, located directly above each wheel 3 (directly above the wheel hub). The acceleration sensor 6 detects the actual acceleration at the acceleration detection target location, i.e., the actual acceleration. Furthermore, the acceleration of the vehicle body is also referred to as sprung acceleration. That is, the acceleration sensor 6 detects sprung acceleration.

[0133] Figure 19 This is a functional block diagram of the control device 10 of an example of the vehicle 1 in the third embodiment. Figure 19 As shown, the data acquisition unit 10a has an acquisition unit 10aa and an estimation unit 10ab.

[0134] The acquisition unit 10aa acquires actual acceleration data, representing the actual acceleration of the object to be detected, from the acceleration sensor 6.

[0135] The actual acceleration data of the three acceleration detection target parts of the accelerometer 6 acquired by the acquisition unit 10aa are input to the estimation unit 10ab. The estimation unit 10ab uses the sprung acceleration estimation network 122 (second neural network) to estimate the actual acceleration of all (for example, four) acceleration detection target parts corresponding to the actual acceleration data of the three input acceleration detection target parts.

[0136] like Figure 19 As shown, in addition to the stroke speed estimation network 121A, the storage unit 10d also stores the sprung acceleration estimation network 122.

[0137] Figure 20 This is an exemplary and schematic block diagram showing the structure of the stroke speed estimation network 121A and the sprung acceleration estimation network 122 of the third embodiment.

[0138] The sprung acceleration estimation network 122 is configured as a recurrent neural network (RNN) that has been pre-learned to estimate the actual acceleration data (output) of all (for example, four) sprung acceleration detection target parts based on the input of actual acceleration data (input data) of a subset (three) of the multiple sprung acceleration detection target parts. During the learning of the travel speed estimation network 121A, the input data is the actual acceleration data of a subset (for example, three) of the multiple sprung acceleration detection target parts. Furthermore, the output data, serving as teacher data, is the actual acceleration of each sprung acceleration detection target part detected by each acceleration sensor 6 when acceleration sensors 6 are installed at all multiple sprung acceleration detection target parts. Moreover, the mounting of each acceleration sensor 6 relative to the vehicle body 2 can differ between the vehicle 1 during the learning of the travel speed estimation network 121A (for example, a test vehicle for learning) and the vehicle 1 that actually uses the sprung acceleration estimation network 122 (for example, a production vehicle). For example, compared to a vehicle 1 that actually uses the sprung acceleration estimation network 122, in vehicle 1 during the learning of the travel speed estimation network 121A, each acceleration sensor 6 can be more securely mounted on the vehicle body 2. This is achieved, for example, by making the rigidity of the brackets, etc., that mount the acceleration sensors 6 to the vehicle body 2 relatively high. Therefore, by suppressing vibrations of the acceleration sensors 6 during learning, the accuracy of acceleration detection can be further improved. The sprung acceleration estimation network 122 is an example of a neural network and a second neural network.

[0139] Similar to the stroke speed estimation network 121A, the sprung acceleration estimation network 122 is configured as a Seq2Seq model. More specifically, similar to the stroke speed estimation network 121A, the sprung acceleration estimation network 122 has: an encoder unit 121a that receives input data and performs encoding processing; and a decoder unit 121b that performs decoding processing based on the encoder result of the encoder unit 121a and outputs the data that has undergone decoding processing.

[0140] The acceleration data estimated by the data acquisition unit 10a is input to the stroke speed estimation unit 10bA. The stroke speed estimation unit 10bA takes the acceleration data estimated by the data acquisition unit 10a as input data and estimates the relative speed, i.e., the stroke speed, through the stroke speed estimation network 121A.

[0141] Figure 21 This is an exemplary and schematic diagram illustrating an example of the estimated result of sprung acceleration based on the technology of the third embodiment. Figure 21 The diagram shows the estimated sprung acceleration (estimated value) and the actual measured sprung acceleration (true value) using the technique of the third embodiment.

[0142] Figure 22 This is an exemplary and schematic diagram illustrating one example of the estimated result of sprung acceleration based on a comparative example technique. Figure 22 The diagram shows the estimated sprung acceleration (estimated value) and the actual measured sprung acceleration (true value) using the comparative example technique. The comparative example technique geometrically calculates the acceleration of the acceleration detection object without acceleration sensor 6 based on the acceleration of the object equipped with acceleration sensor 6. Figure 21 and Figure 22 It can be seen that, compared with the technology of the comparative example, the technology of this embodiment has a smaller error between the estimated sprung acceleration and the actual measured sprung acceleration.

[0143] As described above, in this embodiment, the data acquisition unit 10a includes an acquisition unit 10aa and an estimation unit 10ab. The acquisition unit 10aa acquires actual acceleration data, representing the actual acceleration of the acceleration detection target parts, from acceleration sensors 6, which are installed only on a portion of multiple acceleration detection target parts at different positions on the vehicle 1. The estimation unit 10ab uses a sprung acceleration estimation network 122 (second neural network) that has been trained to estimate the actual acceleration of the acceleration detection target parts detected by each acceleration sensor when acceleration sensors are installed on multiple acceleration detection target parts, based on the input of the actual acceleration data, to estimate the actual acceleration of all multiple acceleration detection target parts corresponding to the input of the actual acceleration data.

[0144] With this structure, since the estimation unit 10ab uses the sprung acceleration estimation network 122 to estimate the actual acceleration of all multiple acceleration detection target parts, including those without acceleration sensors 6, the accuracy of acceleration data can be improved. Therefore, the estimation accuracy of the relative velocity of the wheel relative to the vehicle body can be further improved.

[0145] Furthermore, the estimation unit 10ab can also utilize the sprung acceleration estimation network 122 to estimate the actual acceleration of only the acceleration detection target parts among the plurality of acceleration detection target parts that do not have the acceleration sensor 6 installed. Additionally, the estimation unit 10ab can also utilize the sprung acceleration estimation network 122 to estimate the actual acceleration of at least the acceleration detection target parts among the plurality of acceleration detection target parts that do not have the acceleration sensor 6 installed. In these cases, the actual acceleration of the acceleration detection target parts that are not estimated by the estimation unit 10ab can be acquired by the acquisition unit 10aa from the acceleration sensor 6 and input to the stroke speed estimation unit 10bA.

[0146] Embodiments and variations of the present invention have been described, but these are merely exemplary and not intended to limit the scope of the invention. These new embodiments can be implemented in various other ways, and various omissions, substitutions, and modifications can be made without departing from the spirit of the invention. These embodiments and variations are included within the scope and spirit of the invention, and are included within the scope of the invention as set forth in the claims and its equivalents.

[0147] Explanation of reference numerals in the attached figures

[0148] 1: Vehicle

[0149] 2: Vehicle body

[0150] 3: Wheels

[0151] 4b: Shock absorber

[0152] 10: Control device (vehicle state quantity estimation device)

[0153] 10a: Data Acquisition Department

[0154] 10aa: Acquisition Department

[0155] 10ab: Presumption Department

[0156] 10b: Vehicle Condition Prediction Section

[0157] 10bA: Stroke speed estimation section (relative speed estimation section)

[0158] 121: Vehicle State Estimation Network (First Neural Network)

[0159] 121A: Travel Velocity Estimation Network (First Neural Network)

[0160] 122: Sprout acceleration estimation network (neural network, second neural network)

Claims

1. A vehicle state quantity estimation device, wherein, include: The data acquisition department acquires data related to the vehicle's speed, i.e., the first data. as well as The vehicle state estimation unit utilizes a first neural network that has been trained to estimate at least one of the relative velocity of the vehicle's wheels relative to the vehicle body in the vertical direction and the velocity of the vehicle body in the vertical direction, i.e., the vehicle body velocity, based on the input of the first data. The estimation unit estimates at least one of the relative velocity and the vehicle body velocity corresponding to the first data acquired by the data acquisition unit. The first data is wheel speed data, representing the rotational speed of the wheel. The vehicle has shock absorbers located between the vehicle body and the wheels, which can adjust the damping force according to the input current. The data acquisition unit acquires current value data representing the value of the current. The first neural network learns to estimate at least one of the relative speed and the vehicle speed based on the input of the wheel speed data and the current value data. The vehicle state estimation unit uses the first neural network to estimate at least one of the relative speed and the vehicle body speed corresponding to the wheel speed data and the current value data acquired by the data acquisition unit.

2. A vehicle state quantity estimation device, wherein, include: The data acquisition department acquires data related to the vehicle's speed, i.e., the first data. as well as The vehicle state estimation unit utilizes a first neural network that has been trained to estimate at least one of the relative velocity of the vehicle's wheels relative to the vehicle body in the vertical direction and the velocity of the vehicle body in the vertical direction, i.e., the vehicle body velocity, based on the input of the first data. The estimation unit estimates at least one of the relative velocity and the vehicle body velocity corresponding to the first data acquired by the data acquisition unit. The first data is acceleration data representing the acceleration of the vehicle. The first neural network learns to estimate the relative speed. The vehicle state estimation unit uses the first neural network to estimate the relative velocity corresponding to the acceleration data acquired by the data acquisition unit. The vehicle has shock absorbers located between the vehicle body and the wheels, which can adjust the damping force according to the input current. The data acquisition unit acquires current value data representing the value of the current. The first neural network learns to estimate the relative velocity based on the input of the acceleration data and the current value data. The vehicle state estimation unit uses the first neural network to estimate the relative speed corresponding to the acceleration data and the current value data acquired by the data acquisition unit.

3. A vehicle state quantity estimation device, wherein, include: The data acquisition department acquires data related to the vehicle's speed, i.e., the first data. as well as The vehicle state estimation unit utilizes a first neural network that has been trained to estimate at least one of the relative velocity of the vehicle's wheels relative to the vehicle body in the vertical direction and the velocity of the vehicle body in the vertical direction, i.e., the vehicle body velocity, based on the input of the first data. The estimation unit estimates at least one of the relative velocity and the vehicle body velocity corresponding to the first data acquired by the data acquisition unit. The data acquisition unit includes: The acquisition unit acquires actual acceleration data, representing the actual acceleration of a given acceleration detection target part, from acceleration sensors that are installed only at a portion of multiple acceleration detection target parts located at different positions on the vehicle. as well as The estimation unit utilizes a second neural network that has been trained to estimate the actual acceleration of the acceleration detection target location detected by each of the acceleration sensors when the acceleration sensors are installed at multiple acceleration detection target locations, based on the input of the actual acceleration data, to estimate all the actual accelerations of the multiple acceleration detection target locations corresponding to the input of the actual acceleration data.

4. The vehicle state quantity estimation device according to any one of claims 1 to 3, wherein, The first neural network is a long short-term memory network.

5. A vehicle state quantity estimation device, wherein, include: The acquisition unit acquires actual acceleration data, representing the actual acceleration of a given acceleration detection target part, from acceleration sensors that are installed only at a portion of multiple acceleration detection target parts located at different positions on the vehicle. as well as The estimation unit utilizes a neural network that has been trained to estimate the actual acceleration of the acceleration detection target location detected by each of the acceleration sensors when the acceleration sensors are installed at multiple acceleration detection target locations, based on the input of the actual acceleration data, to estimate all the actual accelerations of the multiple acceleration detection target locations corresponding to the input of the actual acceleration data.

Images