Weight estimation device and weight estimation method

The weight estimation device employs dual learning models to estimate vehicle weight based on suspension displacement, correcting for vehicle inclination and individual variations, addressing inaccuracies from temperature and tilt measurements to enhance precision.

JP2026182011APending Publication Date: 2026-11-19ISUZU MOTORS LTD
View PDF 1 Cites 0 Cited by

Patent Information

Application Number
JP2025077893
Authority / Receiving Office
JP · JP
Patent Type
Applications
Current Assignee / Owner
Filing Date
2025-05-08
Publication Date
2026-11-19

AI Technical Summary

Technical Problem

Existing weight estimation systems using strain sensors are prone to inaccuracies due to measurement errors in temperature and vehicle tilt, which complicate the learning model and reduce the accuracy of load estimation.

Method used

A weight estimation device that utilizes two learning models: a first model to estimate load based on suspension displacement and a second model to correct for vehicle inclination and individual suspension variations, without relying on temperature and tilt measurements.

Benefits of technology

Improves the accuracy of vehicle weight estimation by reducing model complexity and minimizing measurement errors, thereby enhancing the precision of load calculations.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 2026182011000001_ABST
    Figure 2026182011000001_ABST
Patent Text Reader

Abstract

To improve the accuracy of vehicle weight estimation using a learning model. [Solution] The weight estimation device 30 includes a first estimation unit 323 that estimates the load acting on each wheel W corresponding to each suspension S by inputting the displacement amount for each suspension S identified by the identification unit 322 to a first learning model Ms which has been trained to output the load acting on the wheel W corresponding to the suspension S when the displacement amount of the suspension S is input, and a second estimation unit 324 that estimates the correction value of the load acting on each wheel W corresponding to each suspension S by inputting the difference value and displacement amount identified by the identification unit 322 to a second learning model Mp which has been trained to output a correction value for correcting the load when the difference value and displacement amount of each suspension S are input.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to a weight estimation device and a weight estimation method.

Background Art

[0002] The loaded weight estimation system of Patent Document 1 acquires measurement data related to a displacement amount including a strain amount detected by a strain sensor attached to an axle, and estimates the loaded weight by inputting the measurement data into a learned model that has learned the relationship between the measurement data and the loaded weight.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of Invention

Problems to be Solved by the Invention

[0004] The strain amount detected by the strain sensor includes at least strain caused by the elongation amount of the member corresponding to the temperature of the member to which the strain sensor is attached, in addition to strain caused by the load acting on the suspension connected to the axle. Furthermore, the load acting on the suspension varies depending on the attitude (inclination) of the vehicle. For this reason, even if measurement data including the strain amount is input into the learned model, there is a risk that the weight corresponding to the load may be erroneously estimated.

[0005] Therefore, one approach is to input measurement data, including strain, component temperature, and vehicle tilt, into the learning model. However, in this case, the learning model represents the relationship between strain, component temperature, vehicle tilt, and weight, making it more complex than a learning model that represents the relationship between strain and weight. Furthermore, the component temperature measured by the temperature sensor includes measurement errors, and there is a correlation between this temperature and the strain. In addition, the vehicle tilt measured by the tilt angle sensor includes measurement errors, and there is a correlation between this tilt and the vehicle's weight. As a result, a learning model that represents the relationship between strain, component temperature, vehicle tilt, and weight is more susceptible to reduced accuracy in estimating weight due to measurement errors than a learning model that represents the relationship between strain and weight. In other words, in a learning model, the more each data point correlates with other data points, the more likely it is that the accuracy of the estimated weight will decrease due to measurement errors.

[0006] Therefore, the present invention has been made in view of these points, and aims to improve the accuracy of the vehicle weight estimated using a learning model. [Means for solving the problem]

[0007] A weight estimation device according to a first aspect of the present invention includes: an acquisition unit that acquires a measurement value corresponding to the vertical distance between the end of the axle to which the suspension is attached and the vehicle frame, measured for each suspension of the vehicle; an identification unit that identifies the displacement amount of the suspension based on the measurement value and then identifies the difference in the displacement amounts of two of the suspensions among a plurality of suspensions; and a first learning model that has been machine-trained to output the load acting on the wheel corresponding to the suspension when the displacement amount of the suspension is input, which identifies the load for each suspension identified by the identification unit. The system includes: a first estimation unit that estimates the load acting on each wheel corresponding to each suspension by inputting the displacement amount; a second estimation unit that estimates the correction value for the load acting on each wheel corresponding to each suspension by inputting the difference value and the displacement amount identified by the identification unit to a second learning model that has been trained to output a correction value for correcting the load when the difference value and the displacement amount for each suspension are input; and an output unit that outputs an integrated value obtained by integrating the load estimated by the first estimation unit and the correction value estimated by the second estimation unit as the weight of the vehicle.

[0008] The system may further include a storage unit that stores the first learning model for each wheel, which is learned using first training data that includes the vertical load applied to each wheel corresponding to each suspension and the measured displacement, which is the displacement of each suspension measured while the load is applied.

[0009] The measured displacement amount includes a first measured displacement amount measured for each suspension while the load on the wheel is increasing, and a second measured displacement amount measured while the load on the wheel is decreasing. The storage unit may store a first learning model that has been trained to output the median of a first load corresponding to the first measured displacement amount indicating the displacement amount, and a second load corresponding to the second measured displacement amount indicating the displacement amount, when the displacement amount identified by the identification unit is input.

[0010] The system may further include a storage unit that stores a second learning model for each wheel, which is learned using second training data that includes the displacement of each suspension measured when the load acting on at least one of the wheels is different from the load acting on the other wheels, and the load acting on each wheel.

[0011] The system may further include a storage unit that stores a second learning model, which, upon input of the difference value and the displacement amount for each of the suspensions, outputs a correction value including a load corresponding to the displacement amount of the suspension that changes due to the inclination of the vehicle.

[0012] The specified unit may specify, for each suspension, the difference between the displacement of a first other suspension located closest to the first suspension in the longitudinal direction of the vehicle and the displacement of the first suspension, and the difference between the displacement of a second other suspension located closest to the first suspension in the width direction of the vehicle and the displacement of the first suspension.

[0013] The vehicle further comprises a link mechanism formed by connecting a plurality of rod-shaped members so as to be rotatable in the longitudinal direction, wherein the end of the rod-shaped member located at one end of the link mechanism is connected to the end of the axle, and the end of the rod-shaped member located at the other end of the link mechanism is connected to an angle sensor provided on the bottom surface of the vehicle frame facing the road surface, and the acquisition unit may acquire the angle between the rod-shaped member connected to the angle sensor and the vehicle frame, as measured by the angle sensor, as the measured value.

[0014] A second aspect of the present invention relates to a weight estimation method, which includes: an acquisition step, performed by a computer, to acquire measurement values ​​corresponding to the vertical distance between the end of the axle to which the suspension is attached and the vehicle frame, measured for each suspension of the vehicle; an identification step, after identifying the displacement amount of the suspension based on the measurement values, to identify the difference in the displacement amounts of two of the suspensions among a plurality of suspensions; and a first learning model, which has been trained to output the load acting on the wheel corresponding to the suspension when the displacement amount of the suspension is input, to output the load acting on the wheel corresponding to the suspension identified in the identification step for each suspension The system includes: a first estimation step of inputting the displacement amount to estimate the load acting on each wheel corresponding to each suspension; a second estimation step of inputting the difference value and displacement amount identified in the first estimation step to a second learning model that has been trained to output a correction value for correcting the load when the difference value and displacement amount for each suspension are input to estimate the correction value for the load acting on each wheel corresponding to each suspension; and an output step of outputting an integrated value obtained by integrating the load estimated in the first estimation step and the correction value estimated in the second estimation step as the weight of the vehicle. [Effects of the Invention]

[0015] According to the present invention, the accuracy of the vehicle weight estimated using a learning model is improved. [Brief explanation of the drawing]

[0016] [Figure 1] This is a diagram illustrating the overview of weight estimation system 1. [Figure 2] This is a schematic diagram showing an example of a truck. [Figure 3] This is a diagram illustrating the angle measured by the angle sensor 11. [Figure 4] This diagram shows an overview of the process by which the weight estimation device 30 estimates the vehicle weight of vehicle V. [Figure 5]This is a diagram for explaining the output of the first learning model Ms. [Figure 6] This is a diagram showing an example of a processing sequence in the weight estimation device 30. [Mode for Carrying Out the Invention]

[0017] <Overview of Weight Estimation System 1> Figure 1 is a diagram showing an overview of the weight estimation system 1. The weight estimation system 1 shown in Figure 1 includes a sensor group 10, a display device 20, and a weight estimation device 30. The weight estimation system 1 is mounted on a vehicle, and has a function of estimating a vehicle weight (hereinafter referred to as "vehicle gross weight") including the weight of the cargo (loaded object) loaded on the vehicle and the weight of the vehicle itself. The vehicle is a vehicle for transporting cargo, and is, for example, a freight vehicle (truck).

[0018] Figure 2 is a schematic diagram showing an example of a truck. Figure 2 is a bottom view of the vehicle V which is a truck. Figure 2 shows a cargo bed 1a, a cab 1b, two axles A (A1, A2), four suspensions S (S1 to S4), and four wheels W (W1 to W4). The suspension S is a leaf suspension formed by stacking leaf springs. The axle A is an axle housing that covers the axle shaft.

[0019] The sensor group 10 includes an angle sensor 11, a linear sensor 12, or a laser displacement sensor 13 for each suspension S provided in the vehicle V. The angle sensor 11 is, for example, a potentiometer for measuring an angle, and measures the angle between a link connected to the end of the axle A and the vehicle body frame of the vehicle V.

[0020] Figure 3 is a diagram for explaining the angle measured by the angle sensor 11. FIG. 3 shows the left side surface of the vehicle V and an enlarged view in which a portion E of the left side surface is enlarged. The enlarged view shows the portion E in a state where the wheel W1 is removed, and shows the vehicle body frame F, the suspension S1, the axle A1, the link mechanism K1, and the angle sensor 11. In FIG. 3, for simplicity of explanation, a single leaf spring is shown as the suspension S1, but the actual suspension S1 is formed by stacking a plurality of leaf springs.

[0021] The link mechanism K1 is a link mechanism formed by connecting each of a plurality of rod-shaped links (B1 to B3) so as to be rotatable in the longitudinal direction. The end of the link B1 located at one end of the link mechanism K1 is connected to the end of the axle A1, and the end of the link B3 located at the other end of the link mechanism K1 is connected to the angle sensor 11 provided on the bottom surface of the vehicle body frame F facing the road surface.

[0022] As shown in FIG. 3, the angle sensor 11 measures an angle θ between the link B3 and the vehicle body frame F. By measuring the angle θ in this manner, the weight estimation device 30 can specify the distance G between the vehicle body frame F and the axle A1 corresponding to the displacement amount of the suspension S1, based on the angle θ. Furthermore, since the sensor group 10 has the angle sensor 11 for each suspension S, the weight estimation device 30 can specify the distance between the vehicle body frame F and the axle A for each suspension S.

[0023] Returning to Figure 1, the linear sensor 12 is, for example, a linear transducer or linear potentiometer for measuring distance. The linear sensor 12 is mounted on the bottom surface of the vehicle frame F facing the road surface and measures the distance between the damper included in the suspension S and the vehicle frame F. By operating in this manner, the weight estimation device 30 can determine the difference in damper length (so-called damper stroke) between the damper's most compressed state and the damper's most extended state, based on the distance measured by the linear sensor 12. Then, based on the determined damper stroke, the weight estimation device 30 can determine the displacement of the suspension S.

[0024] The laser displacement sensor 13 is installed on the bottom surface of the vehicle frame F facing the road surface and measures the distance between the axle A and the vehicle frame F by irradiating a laser beam onto a component connected to the axle A or the axle A itself. By operating in this manner, the weight estimation device 30 can determine the distance between the vehicle frame F and the axle A1 that corresponds to the displacement of the suspension S1, based on the distance measured by the laser displacement sensor 13.

[0025] The display device 20 is a display included in the instrument panel installed in the passenger compartment 1b. The display device 20 may be a display on an information terminal (e.g., a tablet or mobile phone) used by the driver of the vehicle V, or a display included in the car navigation system installed in the vehicle V. The display device 20 obtains information indicating the vehicle weight of the vehicle V from the weight estimation device 30 and displays text, images, or moving images indicating the vehicle weight.

[0026] The weight estimation device 30 is a device for estimating the vehicle weight of a vehicle V. The weight estimation device 30 identifies the displacement of each suspension S based on the measured values ​​(for example, the angle θ measured by the angle sensor 11) of the sensors included in the sensor group 10, and estimates the vehicle weight based on the said displacement. The weight estimation device 30 has a machine learning model that, for example, outputs the load acting on the wheel W corresponding to the suspension S when the displacement of the suspension S is input, and estimates the vehicle weight by inputting the identified displacement into the learning model. The weight estimation device 30 is, for example, a housing containing electronic components or a printed circuit board on which electronic components are mounted.

[0027] The load acting on the wheels W changes not only with the weight of the cargo on the loading platform 1a, but also with the posture of the vehicle V (tilt in the longitudinal and widthwise directions relative to the horizontal plane). Therefore, the weight estimation device 30 can improve the accuracy of estimating the vehicle weight by estimating the vehicle weight of the vehicle V based on the displacement of the suspension S and the tilt of the vehicle V. However, the learning model that represents the relationship between the change in the suspension S, the tilt of the vehicle V, and the vehicle weight of the vehicle V is more complex than the learning model that represents the relationship between the change in the suspension S and the vehicle weight of the vehicle V, so the accuracy of estimating the vehicle weight may decrease. Furthermore, the tilt of the vehicle V measured using measuring instruments such as tilt angle sensors (not shown) contains measurement errors, which reduces the accuracy of the learning model in estimating the vehicle weight.

[0028] Therefore, the weight estimation device 30 has a first learning model and a second learning model for each wheel W. The first learning model is a learning model that represents the relationship between the load acting on the wheel W and the displacement of the suspension S (so-called basic characteristics of the suspension). The second learning model is a learning model that represents elements different from the basic characteristics of the suspension, for example, a learning model that represents the relationship between the load acting on the wheel W and the inclination of the vehicle V and the individual differences (so-called variability) in the displacement of each suspension S. Furthermore, the weight estimation device 30 identifies the difference in the displacement of each suspension S as the inclination of the vehicle V and inputs it into the second learning model.

[0029] Figure 4 shows an overview of the process by which the weight estimation device 30 estimates the vehicle weight of vehicle V. The process shown in Figure 4 is the process for estimating the vehicle weight of vehicle V shown in Figure 2. Figure 4 shows the same number of processing blocks W1L to W4L as the number of wheels W. For example, processing block W1L estimates the load acting on wheel W1, and processing block W2L estimates the load acting on wheel W2. The weight estimation device 30 estimates that the vehicle weight is the sum of the loads estimated by each processing block (i.e., the loads acting on each wheel W).

[0030] For example, in processing block W1L, the load acting on the suspension S1 corresponding to wheel W1 is estimated by inputting the displacement amount of each suspension S into the first learning model Ms. Then, by inputting the difference in the displacement amounts of each suspension S and the said displacement amount into the second learning model Mp, a correction value for the load acting on the suspension S1 corresponding to wheel W1 is estimated. The correction value is a value that may arise based on, for example, the inclination of the vehicle V and the individual differences of each suspension S. Subsequently, in processing block W1L, the integrated value obtained by multiplying the load output by the first learning model Ms and the correction value output by the second learning model Mp is estimated to be the load acting on wheel W1.

[0031] As described above, by having two learning models for each wheel W, the weight estimation device 30 can suppress the complexity of the relationship between the input and output of the learning models. Furthermore, by using the difference in displacement of each suspension S instead of the inclination of the vehicle V, the weight estimation device 30 can estimate the vehicle weight of the vehicle V without using measuring instruments (sensors) to measure the inclination, thereby suppressing measurement errors included in the measured values. As a result, the weight estimation device 30 can estimate the vehicle weight of the vehicle V with high accuracy. The configuration and operation of the weight estimation device 30 will be described in detail below.

[0032] <Configuration of the weight estimation device 30> As shown in Figure 1, the weight estimation device 30 has a storage unit 31 and a control unit 32. The control unit 32 has an acquisition unit 321, a identification unit 322, a first estimation unit 323, a second estimation unit 324, and an output unit 325.

[0033] The memory unit 31 has a storage medium such as ROM (Read Only Memory), RAM (Random Access Memory), HDD (Hard Disk Drive), or SSD (Solid State Drive). The memory unit 31 stores a program for the control unit 32 to execute. The memory unit 31 stores the first learning model Ms and the second learning model Mp shown in Figure 4 for each wheel W.

[0034] The first learning model Ms is a pre-trained model that has been machine-trained using first training data, which includes the load applied to the wheels W during a bench test of the vehicle V and the displacement of the suspension S measured under the applied load. The first learning model Ms is, for example, a learning model that has been trained using first training data, which includes the vertical load applied to each wheel W corresponding to each suspension S and the measured displacement, which is the displacement of each suspension S measured under the applied load.

[0035] The second learning model Mp is a trained model that has been trained using second training data, which includes the change in the suspension S measured when the inclination of the vehicle V is changed by applying a predetermined load to a predetermined position on the vehicle V, and the load acting on the wheels W. The second learning model Mp is, for example, a learning model that has been trained using second training data, which includes the displacement of each suspension S measured when the load acting on at least one of the wheels W is different from the load acting on the other wheels W, and the load acting on each wheel W.

[0036] The control unit 32 is, for example, a processor such as a CPU (Central Processing Unit) or an ECU (Electronic Control Unit). By executing a program stored in the storage unit 31, the control unit 32 functions as an acquisition unit 321, a identification unit 322, a first estimation unit 323, a second estimation unit 324, and an output unit 325. The control unit 32 may consist of a single processor, or it may consist of multiple processors or a combination of one or more processors and electronic circuits.

[0037] The acquisition unit 321 acquires a measurement value corresponding to the vertical distance between the end of the axle A to which the suspension S is attached and the vehicle frame F, measured for each suspension S provided by the vehicle V. For example, if the sensor group 10 has an angle sensor 11, the acquisition unit 321 acquires the angle θ between the link B3 connected to the angle sensor 11 and the vehicle frame F as a measurement value. The acquisition unit 321 acquires the angle θ measured by the angle sensor 11 for each wheel W.

[0038] If the sensor group 10 includes a linear sensor 12, the acquisition unit 321 acquires the distance between the damper and the vehicle frame F for each wheel W as a measured value. If the sensor group 10 includes a laser displacement sensor 13, the acquisition unit 321 acquires the distance between the axle A and the vehicle frame F for each wheel W as a measured value.

[0039] The identification unit 322 identifies the displacement amount of the suspension S based on the measured value. For each wheel W, the identification unit 322 identifies the displacement amount of the suspension S corresponding to the wheel W. For example, if the sensor group 10 has an angle sensor 11, the identification unit 322 identifies the displacement amount of the suspension S corresponding to the angle θ detected by the angle sensor 11 for each wheel W by referring to the first displacement amount table stored in the storage unit 31. The first displacement amount table is a table showing the displacement amount corresponding to the link angle.

[0040] The identification unit 322 calculates the damper stroke based on the distance between the damper and the vehicle frame F acquired by the acquisition unit 321, for example, when the sensor group 10 has a linear sensor 12. The identification unit 322 then identifies the displacement amount of the suspension S corresponding to the calculated damper stroke for each wheel W by referring to, for example, the second displacement amount table stored in the storage unit 31. The second displacement amount table is a table that shows the displacement amount corresponding to the damper stroke.

[0041] The identification unit 322, for example, when the sensor group 10 has a laser displacement sensor 13, identifies the displacement amount of the suspension S corresponding to the distance acquired by the acquisition unit 321 for each wheel W by referring to the third displacement amount table stored in the storage unit 31. The third displacement amount table is a table showing the displacement amount corresponding to the distance between the axle A and the vehicle frame F.

[0042] The identification unit 322 identifies the displacement amount of each suspension S for each wheel W, and then identifies the difference in displacement amounts between two of the multiple suspensions S. For example, the identification unit 322 identifies the difference by calculating the difference between the displacement amount of the first other suspension S located closest to the suspension in the longitudinal direction of the vehicle V and the displacement amount of the first suspension S for each suspension S. Then, the identification unit 322 identifies the difference by calculating the difference between the displacement amount of the first suspension S and the displacement amount of the second other suspension S located closest to the suspension S in the width direction of the vehicle V for each suspension S.

[0043] In other words, the identification unit 322 identifies the difference in displacement between two suspensions S on one side (right side or left side) of the vehicle V, and the difference in displacement between two suspensions S provided on the same axle A (A1 or A2), for each suspension S. As an example, the identification unit 322 calculates the difference in displacement between one suspension S1 and a first other suspension S2 on the left side of the vehicle V shown in Figure 2. Then, the identification unit 322 calculates the difference in displacement between one suspension S1 and a second other suspension S3 provided on axle A1. The identification unit 322 also calculates the difference for each suspension S, even for other suspensions S different from suspension S1.

[0044] The first estimation unit 323 estimates the load acting on each wheel W using a first learning model Ms, which is stored in the memory unit 31 and has been trained to output the load acting on the wheel W corresponding to the suspension S when the displacement amount of the suspension S is input. The first estimation unit 323 estimates the load acting on each wheel W corresponding to each suspension S by inputting the displacement amount of each suspension S identified by the identification unit 322 for each first learning model Ms.

[0045] To estimate the load acting on the wheel W, one could consider a method of estimating the load acting on the wheel W by inputting the amount of strain of a member such as the vehicle frame F or axle A into the first learning model Ms, instead of the amount of displacement of the suspension S. However, since the amount of strain includes strain caused by the amount of elongation of the member corresponding to the temperature of the member to which the strain sensor is attached, the first learning model Ms needs to represent the relationship between the amount of strain, temperature, and load. This increases the complexity of the first learning model Ms, and the accuracy of the load on the wheel W estimated by the first estimation unit 323 decreases. Furthermore, since the measured values ​​measured by equipment for measuring the temperature of the member (e.g., a temperature sensor) contain measurement errors, the accuracy of the load output by the first learning model Ms decreases.

[0046] In contrast, the first estimation unit 323 inputs the displacement of the suspension S, which is less affected by the temperature of the member, to the first learning model Ms, thus enabling the estimation of the load on the wheel W without inputting temperature to the first learning model Ms. As a result, the first estimation unit 323 can suppress the increasing complexity of the first learning model Ms and reduce measurement errors included in the data input to the first learning model Ms, thereby improving the accuracy of the load output by the first learning model Ms.

[0047] The first training data used in training the first learning model Ms includes measured displacement, which is the displacement of the suspension S when a vertical load is applied to the wheel W. The measured displacement includes, for each suspension S, a first measured displacement measured while the load on the wheel W is increasing (while the load is being applied) and a second measured displacement measured while the load on the wheel W is decreasing (while the load is being removed). Due to hysteresis, the displacement corresponding to the load differs between the first and second measured displacements.

[0048] Therefore, the first estimation unit 323 uses a first learning model Ms that has been trained to output the median value of a first load corresponding to a first measured displacement amount indicating the said displacement amount, and a second load corresponding to a second measured displacement amount indicating the said displacement amount, when a displacement amount is input to it. The first estimation unit 323 outputs the median value of the first load and the second load corresponding to the displacement amount by inputting the displacement amount identified by the identification unit 322 to the first learning model Ms. Then, the first estimation unit 323 estimates, for example, that the median value output by the first learning model Ms is the load acting on the wheel W.

[0049] Figure 5 is a diagram illustrating the output of the first learning model Ms. The horizontal axis of Figure 5 represents the measured displacement, and the vertical axis represents the load. Figure 5 shows the first measured displacement LO measured while the wheel W was being loaded, the second measured displacement UL measured while the wheel W was being unloaded, and the median C between the load corresponding to the first measured displacement LO and the load corresponding to the second measured displacement UL. For example, if the displacement input by the first estimation unit 323 to the learning model Ms represents a displacement H, the learning model Ms outputs the median La of the first load L1 corresponding to the first measured displacement representing displacement H and the second load L2 corresponding to the second measured displacement representing displacement H as the load on the wheel W.

[0050] By operating as described above, the first estimation unit 323 can minimize the difference between the identified load and the true value, regardless of whether the timing of generating the measured value coincides with the timing of applying a load to the wheel W or the timing of removing the load from the wheel W. As a result, the weight estimation device 30 can improve the accuracy of estimating the vehicle weight of the vehicle V.

[0051] The second estimation unit 324 estimates the correction value of the load acting on each wheel W using a second learning model Mp, which has been trained to output a correction value for correcting the load when the difference value and displacement amount for each suspension S stored in the memory unit 31 are input. The correction value includes the load corresponding to the displacement amount of the suspension S that changes due to the inclination of the vehicle V, and the load corresponding to the individual differences in the displacement amount of the suspension S. The second estimation unit 324 estimates the correction value of the load acting on each wheel W corresponding to each suspension S by inputting the difference value and displacement amount of each suspension S identified by the identification unit 322 for each second learning model Mp.

[0052] In this way, the second estimation unit 324 estimates a correction value, allowing the weight estimation device 30 to correct the load acting on the wheels W according to the inclination of the vehicle V and the variation of each suspension S, thereby improving the accuracy of estimating the vehicle weight of the vehicle V. In addition, since the second learning model Mp outputs a correction value according to the inclination of the vehicle V based on the difference in the displacement of each suspension, the weight estimation device 30 can estimate the vehicle weight according to the inclination of the vehicle V without using measuring instruments such as inclination angle sensors. This reduces the measurement error included in the sensor readings, thereby improving the accuracy of estimating the vehicle weight of the vehicle V.

[0053] Furthermore, by having the first estimation unit 323 use the first learning model Ms and the second estimation unit 324 use the second learning model Mp, the weight estimation device 30 can reduce the complexity of each learning model compared to estimating vehicle weight using only one learning model. As a result, it is possible to prevent a decrease in the prediction accuracy of the learning models due to an increase in complexity, and thus the weight estimation device 30 can improve the accuracy of estimating vehicle weight.

[0054] The output unit 325 outputs the integrated value obtained by combining the load estimated by the first estimation unit 323 and the correction value estimated by the second estimation unit 324 as the vehicle weight of the vehicle V. As an example, Figure 5 shows the integrated value D of one wheel W among several wheels W. For example, the output unit 325 identifies the integrated value Lb, obtained by combining the load La estimated by the first estimation unit 323 and the correction value Lb-La estimated by the second estimation unit 324, as the load acting on one wheel W. The output unit 325 outputs the sum of the integrated values ​​identified for each wheel W as the vehicle weight of the vehicle V.

[0055] <Processing sequence in weight estimation device 30> Figure 6 shows an example of a processing sequence in the weight estimation device 30. The processing sequence shown in Figure 6 is a processing sequence that shows the process by which the weight estimation device 30 estimates the vehicle weight of vehicle V.

[0056] The acquisition unit 321 acquires the measured values ​​from the sensors included in the sensor group 10 (step S11). For example, if the sensor group 10 includes an angle sensor 11, the acquisition unit 321 acquires the angle θ between link B3 and vehicle frame F measured by the angle sensor 11 as a measured value. The acquisition unit 321 also acquires the angle θ for each wheel W as a measured value.

[0057] The identification unit 322 identifies the displacement amount of the suspension S corresponding to the wheel W based on the acquired measurement values ​​(step S12). For example, if the acquisition unit 321 acquires the angle θ for each wheel W, the identification unit 322 identifies the displacement amount of the suspension S corresponding to the angle θ for each wheel W by referring to the first displacement amount table stored in the storage unit 31, which shows the displacement amount corresponding to the link angle.

[0058] The identification unit 322 identifies the difference in displacement between two of the multiple suspensions S based on the displacement of the suspension S identified in step S12 (step S13). The identification unit 322 identifies, for example, one or more difference values. As an example, the identification unit 322 identifies the difference between the displacement of suspension S1 and suspension S2, and the difference between the displacement of suspension S1 and suspension S3, as the difference in displacement of suspension S1, as shown in Figure 2. In other words, the identification unit 322 identifies two difference values ​​as the difference in displacement of suspension S1. The identification unit 322 then identifies two difference values ​​for each suspension S, even for other suspensions S different from suspension S1.

[0059] The first estimation unit 323 estimates the load acting on each wheel W based on the displacement amount of each suspension S identified by the identification unit 322 (step S14). The first estimation unit 323 estimates the load acting on each wheel W by, for example, inputting the displacement amount of each suspension S into each of the first learning models Ms for each wheel W.

[0060] The second estimation unit 324 estimates a correction value to correct the load estimated by the first estimation unit 323 based on the displacement amount and difference value of each suspension S identified by the identification unit 322 (step S15). The second estimation unit 324 estimates the correction value of the load acting on each wheel W by inputting the displacement amount and difference value of each suspension S into each of the second learning models Mp for each wheel W.

[0061] The output unit 325 outputs an integrated value obtained by multiplying the load estimated by the first estimation unit 323 and the correction value estimated by the second estimation unit 324 as the vehicle weight of the vehicle V (step S16). For example, the output unit 325 outputs an integrated value obtained by multiplying the load of each wheel W and the correction value of each wheel W as the vehicle weight of the vehicle V to the display device 20.

[0062] <First variation> In the above explanation, the operation is illustrated as calculating an integrated value for each wheel W by integrating the load estimated by the first estimation unit 323 and the correction value estimated by the second estimation unit 324, and then calculating the sum of the integrated values ​​for each wheel W. However, the operation is not limited to this. The weight estimation device 30 may also calculate a corrected integrated value by integrating the loads of each wheel W and the correction value for each wheel W, and then output the sum of the load integrated value and the corrected integrated value as the vehicle weight of the vehicle V.

[0063] <Second variation> In the above explanation, the example given is that the first learning model Ms and the second learning model Mp stored in the memory unit 31 are trained models that have been machine-trained using training data, but the explanation is not limited to this. The first learning model Ms and the second learning model Mp may also be models generated by deep learning.

[0064] <Effects of the weight estimation device 30> As described above, the weight estimation device 30 includes an acquisition unit 321 that acquires a measurement value corresponding to the vertical distance between the end of the axle A to which the suspension S is attached and the vehicle frame F, measured for each suspension S provided by the vehicle V; an identification unit 322 that identifies the displacement amount of the suspension S based on the measurement value and then identifies the difference in displacement values ​​of two of the multiple suspension S; and a first learning model Ms that has been trained to output the load acting on the wheel W corresponding to the suspension S when the displacement amount of the suspension S is input, and the suspension identified by the identification unit 322. The system includes: a first estimation unit 323 that estimates the load acting on each wheel W corresponding to each suspension S by inputting the displacement amount for each S; a second estimation unit 324 that estimates the correction value for the load acting on each wheel W corresponding to each suspension S by inputting the difference value and displacement amount identified by the identification unit 322 into a second learning model Mp that has been trained to output a correction value for correcting the load when the difference value and displacement amount for each suspension S are input; and an output unit 325 that outputs an integrated value obtained by integrating the load estimated by the first estimation unit 323 and the correction value estimated by the second estimation unit 324 as the weight of the vehicle V.

[0065] With this configuration, the weight estimation device 30 has two learning models to estimate the weight of the vehicle V, thus suppressing the complexity of the relationship between the inputs and outputs of the learning models. Furthermore, since the weight estimation device 30 estimates the weight of the vehicle V using the difference in the displacement of each suspension S, it can estimate the weight without using measuring equipment to measure the inclination of the vehicle V. This reduces the measurement error of measuring equipment included in the input data to the learning model. As a result, the weight estimation device 30 can improve the accuracy of estimating the weight of the vehicle V.

[0066] Although the present invention has been described above using embodiments, the technical scope of the present invention is not limited to the scope described in the above embodiments, and various modifications and changes are possible within the scope of its gist. For example, all or part of the apparatus can be configured by functionally or physically distributing and integrating in any unit. Furthermore, new embodiments resulting from any combination of multiple embodiments are also included in the embodiments of the present invention. The effects of the new embodiments resulting from the combinations are combined with the effects of the original embodiments. [Explanation of Symbols]

[0067] 1. Weight Estimation System 10 Sensor Groups 11 Angle sensor 12 Linear Sensors 13. Laser displacement sensor 20 Display device 30 Weight estimation device 31 Storage section 32 Control Unit 321 Acquisition Department 322 Specific part 323 1st estimation part 324 Second estimation part 325 Output section

Claims

1. An acquisition unit that acquires a measurement value corresponding to the vertical distance between the end of the axle to which the suspension is attached and the vehicle frame, measured for each suspension equipped on the vehicle, A determination unit that, after determining the displacement amount of the suspension based on the measured value, determines the difference in the displacement amounts of two of the suspensions among the plurality of suspensions, A first estimation unit estimates the load acting on each wheel corresponding to each suspension by inputting the displacement amount for each suspension identified by the identification unit into a first learning model that has been trained to output the load acting on the wheel corresponding to the suspension when the displacement amount of the suspension is input. A second estimation unit estimates the correction value of the load acting on each wheel corresponding to each suspension by inputting the difference value and displacement amount identified by the identification unit into a second learning model which has been trained to output a correction value for correcting the load when the difference value and displacement amount for each suspension are input. An output unit outputs an integrated value obtained by integrating the load estimated by the first estimation unit and the correction value estimated by the second estimation unit as the weight of the vehicle. A weight estimation device having the following features.

2. The system further includes a storage unit that stores the first learning model for each wheel, which is learned using first training data that includes the vertical load applied to each wheel corresponding to each suspension and the measured displacement, which is the displacement of each suspension measured while the load is applied. The weight estimation device according to claim 1.

3. The measured displacement amount includes a first measured displacement amount measured for each suspension while the load on the wheel is increasing, and a second measured displacement amount measured while the load on the wheel is decreasing. The memory unit stores a first learning model that has been trained to output the median value of a first load corresponding to the first measured displacement amount indicating the displacement amount, and a second load corresponding to the second measured displacement amount indicating the displacement amount, when the displacement amount identified by the identification unit is input. The weight estimation device according to claim 2.

4. The system further includes a storage unit that stores a second learning model for each wheel, which is learned using second training data that includes the displacement of each suspension measured when the load acting on at least one of the wheels is different from the load acting on the other wheels, and the load acting on each wheel. The weight estimation device according to claim 1.

5. The system further includes a storage unit that stores a second learning model, which, upon input of the difference value and the displacement amount for each of the suspensions, outputs a correction value including a load corresponding to the displacement amount of the suspension that changes due to the inclination of the vehicle. The weight estimation device according to claim 1.

6. The specified unit identifies, for each suspension, the difference between the displacement of a first other suspension located closest to the first suspension in the longitudinal direction of the vehicle and the displacement of the first suspension, and the difference between the displacement of a second other suspension located closest to the first suspension in the width direction of the vehicle and the displacement of the first suspension. The weight estimation device according to claim 1.

7. The vehicle further comprises a link mechanism formed by connecting a plurality of rod-shaped members so as to be rotatable in the longitudinal direction, wherein the end of the rod-shaped member located at one end of the link mechanism is connected to the end of the axle, and the end of the rod-shaped member located at the other end of the link mechanism is connected to an angle sensor provided on the bottom surface of the vehicle frame facing the road surface. The acquisition unit acquires the angle between the rod-shaped member connected to the angle sensor and the vehicle frame, as measured by the angle sensor, as the measured value. The weight estimation device according to claim 1.

8. A computer executes An acquisition step is to obtain a measurement value corresponding to the vertical distance between the end of the axle to which the suspension is attached and the vehicle frame, measured for each suspension equipped on the vehicle. After determining the displacement amount of the suspension based on the measured values, the process involves determining the difference in the displacement amounts of two of the suspensions among the plurality of suspensions. A first estimation step involves inputting the displacement amount for each suspension identified in the specific step into a first learning model that has been trained to output the load acting on the wheel corresponding to the suspension, when the displacement amount of the suspension is input, thereby estimating the load acting on each wheel corresponding to each suspension. A second estimation step involves inputting the difference value and displacement amount identified in the specific step into a second learning model, which has been trained to output a correction value for correcting the load, when the difference value and displacement amount for each suspension are input, thereby estimating the correction value for the load acting on each wheel corresponding to each suspension. An output step which outputs an integrated value obtained by integrating the load estimated in the first estimation step and the correction value estimated in the second estimation step as the weight of the vehicle, A weight estimation method having the following characteristics.

Citation Information

Patent Citations

  • Loading weight estimation system of specially-equipped vehicle

    JP2021103166A