Vehicle mass estimation method and vehicle mass estimation system

Abstract

This vehicle mass estimation method involves: detecting a front-rear acceleration of a vehicle by means of an acceleration sensor; inputting a driving force of the vehicle into a vehicle model to calculate an estimated value of the front-rear acceleration that is expected to be detected by the front-rear acceleration sensor, and prior estimated values of state variables of an equation of the vehicle model, the equation representing a relationship between the front-rear acceleration, the driving force, the vehicle mass, and an acceleration error occurring in the front-rear acceleration sensor, and the state variables being the vehicle mass and the acceleration error; calculating estimated values of the state variables on the basis of the prior estimated values and a value obtained by multiplying the difference between the estimated value of the front-rear acceleration and the detected value of the front-rear acceleration detected by the front-rear acceleration sensor by a correction gain; updating the state variables of the equation on the basis of the estimated values of the state variables; calculating the correction gain via a Kalman filter algorithm that incorporates system noise and that uses the driving force as an input value; and changing the system noise on the basis of a variable that is correlated with time.

Description

Vehicle mass estimation method and vehicle mass estimation system

[0001] The present invention relates to a vehicle mass estimation method and a vehicle mass estimation system.

[0002] JP 4926258B discloses a technique for calculating rotational acceleration from the rotational speed of a vehicle's wheels, and estimating vehicle mass as a regression coefficient when linear regression of rotational acceleration and driving force is performed based on a Kalman filter from the vehicle's driving force and rotational angular velocity in accordance with the equation of motion of vehicle driving force = tire rotational acceleration x vehicle mass. JP 4583028B also discloses a similar technique.

[0003] However, in the case of the above-described prior art, the responsiveness of the vehicle mass estimation is always constant, and changing the constants included in the Kalman filter to improve the estimated speed increases the fluctuation of the estimated value in the steady state, while changing the constants included in the Kalman filter to reduce the fluctuation of the estimated value in the steady state slows down the rate of change, making it difficult to achieve both steady-state fluctuation and estimated speed.

[0004] Therefore, an object of the present invention is to provide a vehicle mass estimation method and a vehicle mass estimation system that can achieve both steady-state fluctuations in the estimated value of the vehicle mass and an estimated speed.

[0005] According to one aspect of the present invention, the longitudinal acceleration of a vehicle is detected by a longitudinal acceleration sensor, and the driving force is input into a vehicle model having an equation expressing the relationship between the longitudinal acceleration, the driving force of the vehicle, the vehicle mass, and the acceleration error occurring in the longitudinal acceleration sensor, the equation having the vehicle mass and the acceleration error as state variables, thereby calculating an estimated value of the longitudinal acceleration expected to be detected by the longitudinal acceleration sensor and a priori estimated value of the state variable. The estimated value of the state variable is calculated based on the prior estimated value and a value obtained by multiplying the difference between the estimated value of the longitudinal acceleration and the detected value of the longitudinal acceleration detected by the longitudinal acceleration sensor by a correction gain. The state variable in the equation is updated using the estimated value of the state variable. The correction gain is calculated using a Kalman filter algorithm that incorporates system noise and uses the driving force as an input value. The system noise is changed based on a variable correlated with time.

[0006] FIG. 1 is a block diagram showing the basic configuration of an electric vehicle system to which the vehicle mass estimation method of this embodiment is applied. FIG. 2 is a flowchart showing the flow of processing performed by a motor controller. FIG. 3 is a diagram showing an example of an accelerator opening-torque table. FIG. 4 is a block diagram for explaining the vehicle mass estimation processing. FIG. 5 is a block diagram for explaining an input correction unit. FIG. 6 is a block diagram for explaining a system noise calculation unit. FIG. 7 is a block diagram for explaining a mass estimation calculation unit. FIG. 8 is a time chart when the vehicle mass estimation method of the comparative example is applied. FIG. 9 is a time chart when the vehicle mass estimation method of this embodiment is applied. FIG. 10 is a time chart of the reset processing of the system noise component in the system noise calculation unit.

[0007] Hereinafter, an embodiment of the present invention will be described with reference to the drawings.

[0008] [System Configuration of Electric Vehicle 10] Figure 1 is a block diagram showing the basic configuration of an electric vehicle system to which the vehicle mass estimation method of this embodiment is applied. The electric vehicle 10 is a vehicle that has a motor 18 as part or all of the vehicle's drive source and can run using the driving force of the motor 18, and includes electric vehicles and hybrid vehicles. Note that the electric vehicle 10 shown in Figure 1 is a 2WD vehicle using the motor 18 as its drive source, but the invention can also be applied to 2WD vehicles using an engine as their drive source, and 4WD vehicles using either the motor 18 or an engine as their drive source.

[0009] Digital signals indicating vehicle conditions such as the longitudinal acceleration of the vehicle, vehicle speed V, accelerator opening APO, rotor phase α of motor 18, and three-phase AC currents iu, iv, and iw of motor 18 are input to motor controller 12 (control device). Based on the input signals, motor controller 12 generates PWM signals tu, tv, and tw for controlling motor 18. In addition, motor controller 12 generates drive signals for inverter 16 in accordance with the generated PWM signals tu, tv, and tw.

[0010] The inverter 16 converts the direct current supplied from the battery 14 into alternating current by turning on / off two switching elements (e.g., power semiconductor elements such as IGBTs and MOS-FETs) provided for each phase, and supplies the desired current to the motor 18.

[0011] The motor 18 (three-phase AC motor) generates driving force using the AC current supplied from the inverter 16, and transmits the driving force to left and right drive wheels 24 a, 24 b via a reducer 20 and a drive shaft 22. When the motor 18 is rotated by the drive wheels 24 a, 24 b while the vehicle is running, it generates regenerative driving force, thereby recovering the kinetic energy of the vehicle as electrical energy. In this case, the inverter 16 converts the AC current generated during regenerative operation of the motor 18 into DC current and supplies it to the battery 14.

[0012] The current sensor 26 detects three-phase AC currents iu, iv, and iw flowing through the motor 18. However, since the sum of the three-phase AC currents iu, iv, and iw is zero, the currents of any two phases may be detected and the current of the remaining phase may be calculated.

[0013] The rotation sensor 28 is, for example, a resolver or an encoder, and detects the rotor phase α of the motor 18 .

[0014] 2 is a flowchart showing the flow of processing performed by the motor controller 12. The processing from S201 to S206 is constantly executed at regular intervals while the electric vehicle system is running.

[0015] In the input process of S201, signals required for the control calculations described below are input to the motor controller 12. Here, the longitudinal acceleration (m / s ２ ), vehicle speed V (km / h), accelerator opening APO (%), and rotor phase α (rad) of the motor 18. Also input are the rotation speed N (rpm) of the motor 18, three-phase AC currents iu, iv, and iw flowing through the motor 18, and the DC voltage value Vdc (V) of the battery 14.

[0016] Vehicle longitudinal acceleration sensor value a [m / s ２] (detected value) is acquired by a longitudinal acceleration sensor (not shown) that detects longitudinal acceleration occurring in the vehicle.

[0017] The vehicle speed V (km / h) is acquired from a vehicle speed sensor (not shown) or another controller via communication. Alternatively, the motor controller 12 obtains the vehicle speed V (m / s) by multiplying the motor angular velocity detection value ωm by the tire dynamic radius r and dividing the result by the gear ratio of the final gear, and then multiplying the result by 3600 / 1000 for unit conversion to obtain the vehicle speed V (km / h).

[0018] The accelerator opening APO (%) is obtained from an accelerator opening sensor (not shown) or is obtained by communication from another controller (not shown) such as a vehicle controller.

[0019] The rotor phase α (rad) of the motor 18 is acquired from the rotation sensor 28. The motor angular velocity detection value ωm, which is the mechanical angular velocity of the motor 18, is obtained by dividing the rotor angular velocity ω (electrical angle) by the number p of pole pairs of the motor 18. The rotation speed N (rpm) of the motor 18 is obtained by multiplying the obtained motor angular velocity detection value ωm by 60 / (2π). The rotor angular velocity ω is obtained by differentiating the rotor phase α.

[0020] The three-phase AC currents iu, iv, and iw (A) flowing through the motor 18 are acquired from a current sensor 26 .

[0021] The DC voltage value Vdc (V) is detected by a voltage sensor (not shown) provided on a DC power supply line between the battery 14 and the inverter 16. Note that the DC voltage value Vdc (V) may also be detected from a signal related to the power supply voltage value transmitted from a battery controller (not shown).

[0022] In the basic target torque calculation process of S202, the motor controller 12 calculates the basic motor torque command value Tm based on the accelerator opening APO and the motor angular velocity detection value ωm using the accelerator opening-torque table shown in FIG. １ ＊ Set.

[0023] In the vibration suppression control process of S203, the basic motor torque command value Tm calculated in S202 １＊ and the motor angular velocity detection value ωm are input, and the final motor torque command value Tm is calculated to suppress torque transmission system vibration (torsional vibration of the drive shaft, etc.) without sacrificing the response of the drive shaft torque. ２ ＊ Calculate.

[0024] In the vehicle mass estimation process in S204, the final motor torque command value Tm calculated in S203 is used. ２ ＊ and the longitudinal acceleration sensor value a, the vehicle mass M is estimated. The vehicle mass estimation process will be described in detail later.

[0025] In the current command value calculation process of S205, the motor controller 12 calculates the final motor torque command value Tm ２ ＊ , the motor angular velocity detection value ωm, and the d-axis current command value id based on the DC voltage value Vdc. ＊ , q-axis current command value iq ＊ For example, the final motor torque command value Tm ２ ＊ , the rotation speed N of the motor 18, the DC voltage value Vdc, and the d-axis current command value id ＊ and the q-axis current command value iq ＊ A table is prepared in advance that defines the relationship between the d-axis current command value id and the ＊ , q-axis current command value iq ＊ Ask for.

[0026] In the current control of S206, the motor controller 12 controls the d-axis current id and the q-axis current iq by the d-axis current command value id calculated in S205. ＊ and the q-axis current command value iq ＊ To this end, first, the d-axis current id and the q-axis current iq are calculated based on the three-phase AC currents iu, iv, and iw input in S201 and the rotor phase α of the motor 18. Next, the d-axis current command value id ＊ and the d-axis current id, the d-axis voltage command value vd is calculated, and the q-axis current command value iq ＊and the q-axis current iq, a q-axis voltage command value vq is calculated. Note that a decoupling voltage required to cancel out the interference voltage between the d- and q-axis orthogonal coordinate axes may be added to the calculated d-axis voltage command value vd and q-axis voltage command value vq.

[0027] Next, three-phase AC voltage command values ​​vu, vv, and vw are calculated from the d-axis voltage command value vd, the q-axis voltage command value vq, and the rotor phase α of the motor 18. Then, PWM signals tu(%), tv(%), and tw(%) are calculated from the calculated three-phase AC voltage command values ​​vu, vv, and vw and the DC voltage value Vdc. The PWM signals tu, tv, and tw thus calculated are used to open and close the switching elements of the inverter 16, thereby controlling the motor 18 to the final motor torque command value Tm ２ ＊ The motor can be driven at a desired torque indicated by the arrow.

[0028] [Vehicle Model] The vehicle model of this embodiment will be described below. The equation of the vehicle model is shown in equation (1).

[0029] where a is the longitudinal acceleration sensor value, M is the vehicle mass, u is the driving force, and e is the acceleration error.

[0030] The longitudinal acceleration sensor is a sensor that can detect longitudinal acceleration occurring in the vehicle, and therefore the longitudinal acceleration sensor value also includes a gravitational acceleration component due to the gradient.

[0031] The acceleration error e includes an error in the installation of the longitudinal acceleration sensor and an error in the acceleration caused by the tilt of the vehicle due to loading.

[0032] Here, when the transformation of equation (2) is applied to equation (1), equation (3) is obtained.

[0033] When equation (3) is discretized, equation (4) is obtained.

[0034] Reciprocal M of vehicle mass M ｉｎｖ and the acceleration error e, the state variable x ｋ is the inverse of the vehicle mass M ｉｎｖWhen the acceleration error e is taken as the acceleration error, and the observed value y is taken as the longitudinal acceleration sensor value a, the state equation becomes equation (5) and the observation equation becomes equation (6).

[0035] Furthermore, the mean is 0 and the variance σ ｗ ２ The system noise w is a normal white noise with mean 0 and variance σ ｖ ２ Considering the observation noise v, which is normal white noise, equation (5) becomes equation (7), and equation (6) becomes equation (8).

[0036] However, in equation (7), equations (9) and (10) hold, and in equation (8), equation (11) holds.

[0037] [Vehicle Mass Estimation Process] Fig. 4 is a block diagram for explaining the vehicle mass estimation process. The motor controller 12 (Fig. 1) includes an input correction unit (S401), a system noise calculation unit (S402), and a mass estimation calculation unit (S403).

[0038] The input correction unit (S401) calculates the vehicle longitudinal acceleration sensor value, the final motor torque command value Tm ２ ＊ , driving wheel speed ω ｗ＿ｄｒｉｖｅ , driven wheel speed ω ｗ＿ｄｒｉｖｅｎ , air resistance / rolling resistance F ｌｏａｄ The vehicle longitudinal acceleration sensor value (corrected) a and the driving force u, which are input signals to the mass estimation calculation unit (S402), are calculated based on the above. The details of the input correction unit (S401) will be described with reference to FIG.

[0039] The system noise calculation unit (S402) calculates the system noise Q (system noise component q Ｍｉｎｖ The system noise calculation unit (S402) will be described in detail with reference to FIG.

[0040] The mass estimation calculation unit (S403) estimates the vehicle mass M and acceleration error e defined in equation (1) based on the driving force u and the vehicle longitudinal acceleration sensor value (after correction) a. Details of the mass estimation calculation unit (S403) will be described using FIG. 7.

[0041] <Input Correction Unit> Fig. 5 is a block diagram for explaining the input correction unit. The input correction unit (S401 in Fig. 4) executes the following steps S501 to S511.

[0042] In S501, the input correction unit calculates the driving wheel speed ω ｗ＿ｄｒｉｖｅ Approximate differentiation is performed to obtain the driving wheel angular acceleration ω ・ ｗ＿ｄｒｉｖｅ Calculate.

[0043] In S502, the input correction unit calculates the driving wheel angular acceleration ω ・ ｗ＿ｄｒｉｖｅ Wheel inertia J ｗ , motor / differential inertia (motor / differential gear inertia) J ｍｅ , wheel radius R ａ The driving force equivalent of the drive wheel inertia is calculated by multiplying the drive wheel inertia by a gain configured as follows:

[0044] In S503, the input correction unit calculates the driven wheel speed ω ｗ＿ｄｒｉｖｅｎ Approximate differentiation is performed to obtain the driven wheel angular acceleration ω ・ ｗ＿ｄｒｉｖｅｎ Calculate.

[0045] In S504, the input correction unit calculates the driven wheel angular acceleration ω ・ ｗ＿ｄｒｉｖｅｎ Wheel inertia J ｗ , wheel radius R ａ The driving force equivalent of the driven wheel inertia is calculated by multiplying the driving force by a gain composed of the following:

[0046] In S505, the input correction unit adds the driving force equivalent value of the driving wheel inertia output in S502 to the driving force equivalent value of the driven wheel inertia output in S504 to calculate the driving force equivalent value of the driving wheel / driven wheel inertia.

[0047] In S506, the input correction unit calculates the air resistance / rolling resistance F ｌｏａｄA low-pass filter is applied to adjust the phase of the air resistance / rolling resistance F ｌｏａｄ is calculated using equation (12) using the vehicle speed V, and the parameter A ０ , A １ , A ２ The design value or the value identified by experiment is used.

[0048] In S507, the input correction unit corrects the final motor torque command value Tm ２ ＊ A low-pass filter process is performed on the signal for phase adjustment.

[0049] In S508, the input correction unit calculates the final motor torque command value Tm ２ ＊ The gain K, which is the efficiency of the drive source and the drive force transmission system, ｅ Multiply by.

[0050] In S509, the input correction unit calculates the gear ratio N ａｌ and wheel radius R ａ By multiplying the gain composed of the above, the unit is converted into the dimension of driving force.

[0051] In S510, the input correction unit calculates the driving force u by subtracting the output of S505 and the output of S506 from the output of S509.

[0052] In S511, the input correction unit performs low-pass filtering for phase adjustment on the vehicle longitudinal acceleration sensor value, and calculates a vehicle longitudinal acceleration sensor value (corrected) a.

[0053] The time constants τ used in the filter processes of S501, S503, S506, S507, and S511 are all set to the same value to align the phases.

[0054] <System Noise Calculation Unit> Fig. 6 is a block diagram for explaining the system noise calculation unit. The system noise calculation unit (S402 in Fig. 4) executes S601 to S605 shown below.

[0055] In S601, the system noise calculation unit outputs a system noise reset signal as "HIGH" when the ignition switch is switched from "OFF" to "ON" or when the door state is switched from "open" to "closed", and outputs the reset signal as "LOW" in all other cases.

[0056] In S602, the system noise calculation unit calculates a preset initial value qMinv_start and a preset final value q Ｍｉｎｖ＿ｅｎｄ The system noise calculation unit outputs either of the initial value qMinv_start when the reset signal is "HIGH" and the final value qMinv_start when the reset signal is "LOW". Ｍｉｎｖ＿ｅｎｄ Output.

[0057] In S603, the system noise calculation unit calculates the difference between the initial value qMinv_start and the output of S602.

[0058] In S604, the system noise calculation unit calculates the time constant τ ｑＭｉｎｖ low-pass filtering is performed.

[0059] In S605, the output of S602 is added to the output of S604 to change the value from "OFF" to "ON" or the door state changes from "open" to "closed" to change the value from the initial value qMinv_start to the final value q Ｍｉｎｖ＿ｅｎｄ time constant τ ｑＭｉｎｖ and the system noise component q related to the vehicle mass M. Ｍｉｎｖ can be output.

[0060] <Mass Estimation Calculation Unit> Figure 7 is a block diagram for explaining the mass estimation calculation unit. The mass estimation calculation unit (S403 in Figure 4) executes S701 to S705 shown below.

[0061] In S701, the mass estimation calculation unit calculates the state variable x by processing the equations (5) and (6) (or the equations (7) and (8)) using the driving force u as an input value. ｋ The prior estimate of x^ ｋ｜ｋ－１ and calculate the estimated value a^ of the longitudinal acceleration that is expected to be detected by the vehicle longitudinal acceleration sensor.ｋ｜ｋ－１ is the state variable x in equations (5) and (7). ｋ＋１ The estimated value a^ of the longitudinal acceleration detected by the vehicle longitudinal acceleration sensor is calculated using the observed value y ｋ This becomes:

[0062] In S702, the mass estimation calculation unit outputs the difference obtained by subtracting the estimated value a^ of the longitudinal acceleration detected by the vehicle longitudinal acceleration sensor from the vehicle longitudinal acceleration sensor value (after correction) a.

[0063] In S703, the mass estimation calculation unit adds the driving force u and the system noise component q to the output value of S702. Ｍｉｎｖ The correction gain K Ｋ By multiplying by ｋ The prior estimate of x^ ｋ｜ｋ－１ A correction amount for correcting the above is calculated.

[0064] Correction gain K Ｋ When the Kalman filter algorithm is used, it is found by successively calculating the equations (13), (14), and (15).

[0065] where P is the error covariance matrix, K Ｋ is the Kalman gain, Q is the covariance matrix for the system noise, and R is the covariance matrix for the observation noise. Ｋ is updated within the algorithm using the driving force u as an input value, but Q and R are set arbitrarily, so they are set taking into consideration noise contained in the sensors used for the input and observed values ​​and any errors that may occur. The initial value of P is also set taking into consideration the magnitude of the initial estimation error.

[0066] A covariance matrix is ​​a matrix of covariances between vector elements. The larger the covariance, the larger the error involved in the state transition and observation. In this case, the autocovariance (the diagonal components of the covariance matrix) is important among the covariances. This is because the acceleration error e, which is a state variable, and the reciprocal M of the vehicle mass M ｉｎｖ are independent of each other. That is, Q is given by equation (16) and R is given by equation (17).

[0067] Here, q ｅ is the autocovariance of the acceleration error e, q Ｍｉｎｖ is the autocovariance of the vehicle mass M (system noise component), r ａ is the autocovariance of the longitudinal acceleration sensor value a. ｅ , q Ｍｉｎｖ , and r ａ is estimated by experiment or prior knowledge. For example, Q depends on the sampling rate, and R depends on the noise characteristics of the sensor.

[0068] In addition, the correction gain K Ｋ When the algorithm of the recursive least squares method with forgetting factor is used, it can be found by sequentially calculating the equations (18), (19), and (20).

[0069] where λ is the forgetting factor.

[0070] In S704, the mass estimation calculation unit calculates the state variable x ｋ The prior estimate of x^ ｋ｜ｋ－１ is corrected with the output value of S603, the state variable x ｋ The estimated value of x^ ｋ｜ｋ Here, the state variable x ｋ The component of the acceleration error e and the vehicle mass M are ｉｎｖ Therefore, the estimated value of the vehicle mass M can be calculated using equation (2).

[0071] In S705, the mass estimation calculation unit calculates the state variable x ｋ The estimated value of x^ ｋ｜ｋ Then, the past value of the state variable x is sampled, and the process returns to S701. ｋ The estimated value of x^ ｋ｜ｋ The past value of is the state variable x in equations (5)-(8). ｋ This becomes:

[0072] By sequentially repeating the processes of S701 to S705, the state variable x ｋ The estimated value of x^ ｋ｜ｋBy updating the above, it is possible to calculate an accurate estimate of the vehicle mass M.

[0073] [Time Chart of Vehicle Mass Estimation] Fig. 8 is a time chart when the vehicle mass estimation method of the comparative example is applied, and Fig. 9 is a time chart when the vehicle mass estimation method of this embodiment is applied.

[0074] The time charts shown in Figures 8 and 9 are based on the assumption that the vehicle starts traveling at time t0, reaches a predetermined vehicle speed immediately after time t0, and then applies a step function change to the longitudinal acceleration sensor value at time t1 (for example, when a positional deviation or an angular deviation occurs in the longitudinal acceleration sensor). Here, the final target torque shown in Figures 8 and 9 is the final motor torque command value Tm given to the motor 18 when the vehicle to be used in this embodiment is a vehicle that travels by the motor 18. ２ ＊ In the case of a vehicle that runs on an engine, this corresponds to an engine torque command value that is issued to the engine.

[0075] 8 and 9, in order to always be able to calculate the estimated vehicle mass M, the final target torque (final motor torque command value Tm ２ ＊ ) is superimposed with a pulsating component of a predetermined amplitude and period, but in actual situations, the final target torque (final motor torque command value Tm ２ ＊ ) changes. ２ ＊ 8 and 9, at time t0, the calculation is started with the initial value of the estimated vehicle mass M intentionally deviated from the true value of the vehicle mass M.

[0076] As shown in Fig. 7, in the comparative example, the system noise component related to the vehicle mass is set to a predetermined fixed value. Therefore, as shown in Fig. 7, in the comparative example, the set value is set low, so the estimated speed is slow, and the estimated value converges to the true value at time ta, just before time t1. On the other hand, in the comparative example, because the fixed value is set low, fluctuations in the estimated value are kept low even when a step function-like change is applied to the longitudinal acceleration sensor value at time t1.

[0077] As shown in FIG. 8, in this embodiment, the system noise component q Ｍｉｎｖ is set to an initial value qMinv_start that is higher than the fixed value in the comparative example, and the final value q Ｍｉｎｖ＿ｅｎｄ is set to a value lower than the fixed value in the comparative example (e.g., zero), and the system noise component q Ｍｉｎｖ is the time constant τ from the initial value ｑＭｉｎｖ It decays monotonically to the final value q Ｍｉｎｖ＿ｅｎｄ Therefore, immediately after time t0, the estimated speed is faster than in the comparative example, and the estimated value converges to the true value at time tb, which is shorter than time ta. Note that (time tb - time t0) / (time ta - time t0)=approximately 1 / 10.

[0078] Then, the system noise component q Ｍｉｎｖ becomes lower than the fixed value in the comparative example between time tb and time ta. Therefore, even if a step function change is applied to the longitudinal acceleration sensor value at time t1, the fluctuation in the estimated value can be suppressed lower than in the comparative example. Therefore, even if noise is superimposed on the longitudinal acceleration sensor value or when traveling on a rough road surface, the fluctuation in the estimated value of vehicle mass M can be reduced.

[0079] [Time chart of system noise component reset process] FIG. 10 shows the time chart of the system noise component q Ｍｉｎｖ 1 is a time chart of the reset process. Here, it is assumed that the ignition switch is switched from "OFF" to "ON" at time t0, and the door state is switched from "open" to "closed" at time t1 after time t0. Therefore, the system noise component q Ｍｉｎｖis reset to the initial value qMinv_start at time t0 and time t1. When the driver switches the ignition switch from "OFF" to "ON" or when the door is closed, passengers and luggage are expected to get on and off during that time. Therefore, the system noise component q Ｍｉｎｖ The value of is reset. Note that the time constant τ ｑＭｉｎｖ At time t, the system noise component q Ｍｉｎｖ The value of is the final value q Ｍｉｎｖ＿ｅｎｄ Even if the value qMinv_start has not converged to the initial value qMinv_start, it is reset to the initial value qMinv_start at time t1.

[0080] As described above, according to this embodiment, the vehicle mass M can be estimated more quickly and accurately.

[0081] [Effects of this embodiment] The vehicle mass estimation method of this embodiment detects the longitudinal acceleration of the vehicle using a longitudinal acceleration sensor (not shown), and calculates the relationship between the longitudinal acceleration (longitudinal acceleration sensor value a), the driving force u of the vehicle, the vehicle mass M, and the acceleration error e generated in the longitudinal acceleration sensor (not shown) using equations (equations (5) and (6), or equations (7) and (8)). ｋ By inputting the driving force u into a vehicle model (S701) having an equation, an estimated value a^ of longitudinal acceleration that is assumed to be detected by a longitudinal acceleration sensor (not shown) and a state variable x ｋ The prior estimate of x^ ｋ｜ｋ－１ and calculate the prior estimate x^ ｋ｜ｋ－１ and a correction gain K is added to the difference between the estimated value a^ of the longitudinal acceleration and the detected value a of the longitudinal acceleration detected by the longitudinal acceleration sensor. Ｋ and the state variable x ｋ The estimated value of x^ ｋ｜ｋ and calculate the state variable x ｋ The estimated value of x^ ｋ｜ｋ The state variable x in the equation ｋ is updated, and the correction gain K Ｋis calculated using the Kalman filter algorithm (equations (13)-(15)) that includes the system noise Q and uses the driving force u as an input value, and the system noise Q (system noise component q Ｍｉｎｖ ) is varied based on variables that are correlated with time.

[0082] In the Kalman filter, the system noise Q (system noise component q Ｍｉｎｖ ) is a value related to the responsiveness of the estimated value and fluctuations in the steady state. The larger the value, the higher the response but the larger the fluctuations in the steady state. The smaller the value, the lower the response but the smaller the fluctuations in the steady state. Therefore, by the above method, the system noise Q (system noise component q Ｍｉｎｖ ) over time, the response of the estimated value of the vehicle mass M can be made faster, and the fluctuation of the estimated value of the vehicle mass M in the steady state can be adjusted to be small.

[0083] In this embodiment, the driving force u is sequentially input to the vehicle model (S701) to obtain the estimated value a^ of the longitudinal acceleration and the preliminary estimated value x^ ｋ｜ｋ－１ and are sequentially calculated, and the difference (between the estimated value a^ of the longitudinal acceleration and the detected value of the longitudinal acceleration (longitudinal acceleration sensor value a)) is sequentially calculated based on the sequentially calculated estimated value a^ of the longitudinal acceleration and the detected value of the longitudinal acceleration (longitudinal acceleration sensor value a), and the driving force u is sequentially input into the Kalman filter algorithm (equation (13)-equation (15)) to calculate the correction gain K Ｋ is calculated sequentially, and the calculated prior estimate x^ ｋ｜ｋ－１ , the difference, and the correction gain K Ｋ Based on the state variable x ｋ The estimated value of x^ ｋ｜ｋ The state variables x ｋ The estimated value of x^ ｋ｜ｋ The state variable x in the equation ｋ is updated sequentially.

[0084] By using the above method, it is possible to successively reduce the modeling error in the vehicle model (S701) to accurately estimate the vehicle mass M, to speed up the estimation response of the estimated value of the vehicle mass M, and to adjust the steady-state fluctuation of the estimated value of the vehicle mass M to be small.

[0085] In this embodiment, when the vehicle state changes, the system noise Q (system noise component q Ｍｉｎｖ ) is set to an initial value qMinv_start, and then the system noise Q (system noise component q Ｍｉｎｖ ) is varied based on variables that are correlated with time.

[0086] By using the above method, when the vehicle state changes, the system noise Q (system noise component q Ｍｉｎｖ ) is set to the initial value qMinv_start, so that the estimated speed of the vehicle mass M when the vehicle state changes is increased, and then the system noise Q (system noise component q Ｍｉｎｖ ) based on a time-correlated variable can be adjusted to reduce steady-state fluctuations in the estimate of vehicle mass M.

[0087] In this embodiment, the time-correlated variable is the system noise Q (system noise component q Ｍｉｎｖ ) is set to the initial value qMinv_start, or the system noise Q (system noise component q Ｍｉｎｖ ) is set to the distance traveled by the vehicle since it was set to the initial value qMinv_start.

[0088] When the vehicle state changes, the state variable x ｋ The estimated value of x^ ｋ｜ｋ When the estimated value of the vehicle mass M included in the equation deviates from the true value of the vehicle mass M, the estimated value x^ ｋ｜ｋ It is assumed that at the start of the estimation calculation, the difference between the estimated value and the true value is the largest, which means that responsiveness is required, and that the difference from the true value will become smaller as the estimation calculation is carried out. Then, by using the above method, the system noise Q (system noise component qＭｉｎｖ ) is gradually set to a smaller value, the system noise Q (system noise component q Ｍｉｎｖ ) is reduced, it is possible to estimate the vehicle mass with high response and little fluctuation in the steady state.

[0089] In this embodiment, the system noise Q (system noise component q Ｍｉｎｖ ) is the first component related to the vehicle mass M (system noise component q Ｍｉｎｖ ) and the first component (system noise component q Ｍｉｎｖ ) different from the second component (q ｅ ), and the first component (system noise component q Ｍｉｎｖ ) is varied by a variable correlated with time, and the second component (q ｅ ) is set to a predetermined value regardless of the time-correlated variables.

[0090] By the above method, the first component (system noise component q Ｍｉｎｖ Even if the second component (q ｅ ) can be prevented from adversely affecting the response of the state variable (acceleration error) corresponding to the

[0091] In this embodiment, when the vehicle state changes, the system noise Q (system noise component q Ｍｉｎｖ ) is set to an initial value qMinv_start, and then the system noise Q (system noise component q Ｍｉｎｖ ) is the change in the time-correlated variable (time constant τ ｑＭｉｎｖ ) to set the final value qMinv_start, which is lower than the initial value qMinv_start. Ｍｉｎｖ＿ｅｎｄ Converge to.

[0092] As described above, the highest response is required immediately after the start of the estimation calculation, and the system noise Q (system noise component q Ｍｉｎｖ ) is required to change. Therefore, by the above method, the system noise Q (system noise component q Ｍｉｎｖ ) is largest immediately after the start of the estimation calculation, and then gradually becomes smaller, thereby achieving both high responsiveness and reduced steady-state fluctuations.

[0093] In this embodiment, when a change in the vehicle state occurs multiple times, when a previous change in the vehicle state occurs (for example, when the ignition switch is switched from "OFF" to "ON"), the system noise Q (system noise component q Ｍｉｎｖ ) is set to the initial value qMinv_start, and then when a subsequent vehicle state occurs (for example, the door state changes from "open" to "closed"), the system noise (system noise component q Ｍｉｎｖ ) is the system noise Q (system noise component q Ｍｉｎｖ ) to the initial value qMinv_start.

[0094] By using the above method, it is determined that there is a possibility that a change in vehicle mass occurs whenever the vehicle state changes, and the system noise Q (system noise component q Ｍｉｎｖ ) to the initial value qMinv_start, the vehicle mass can be estimated with a fast response speed.

[0095] Vehicle state changes include the transition of the vehicle ignition from an off state ("OFF") to an on state ("ON") and the transition of a vehicle door from an open state to a closed state.

[0096] By using the above method, the system noise Q (system noise component q Ｍｉｎｖ ) to the initial value qMinv_start, the vehicle mass can be estimated with a fast response speed.

[0097] The vehicle mass estimation system of this embodiment includes a longitudinal acceleration sensor (not shown) for detecting longitudinal acceleration of the vehicle, a driving force calculation unit (S510) for calculating a driving force u of the vehicle, and equations (5) and (6), or (7) and (8) that represent the relationship between the longitudinal acceleration (longitudinal acceleration sensor value a), the driving force u, the vehicle mass M, and an acceleration error e generated in the longitudinal acceleration sensor (not shown), in which the vehicle mass M and the acceleration error e are expressed as a state variable x ｋThe equation is expressed as follows: the estimated value a^ of longitudinal acceleration that is assumed to be detected by a longitudinal acceleration sensor (not shown) when a driving force u is input; and the state variable x ｋ The prior estimate of x^ ｋ｜ｋ－１ and a vehicle model (S701) for calculating the prior estimated value x^ ｋ｜ｋ－１ and a correction gain K is calculated based on the difference between the estimated value a^ of the longitudinal acceleration and the detected value of the longitudinal acceleration (longitudinal acceleration sensor value a) detected by a longitudinal acceleration sensor (not shown). Ｋ and the state variable x ｋ The estimated value of x^ ｋ｜ｋ and a state variable estimation unit (S703, S704) for calculating the state variable x ｋ The estimated value of x^ ｋ｜ｋ The state variable x in the equation ｋ The state variable estimation unit (S703, S704) updates the correction gain K Ｋ , the system noise Q (system noise component q Ｍｉｎｖ ) and is calculated using the algorithm of the Kalman filter (equations (13)-(15)) with the driving force u as an input value, and the system noise Q (system noise component q Ｍｉｎｖ ) varies based on time-correlated variables.

[0098] With the above configuration, the system noise Q (system noise component q Ｍｉｎｖ ) over time, the response of the estimated value of the vehicle mass M can be made faster, and the fluctuation of the estimated value of the vehicle mass M in the steady state can be adjusted to be small.

[0099] Although the embodiments of the present invention have been described above, the above embodiments merely illustrate some of the application examples of the present invention, and it is not intended that the technical scope of the present invention be limited to the specific configurations of the above embodiments.

Claims

1. A vehicle mass estimation method comprising: detecting the longitudinal acceleration of a vehicle using a longitudinal acceleration sensor; inputting the driving force into a vehicle model having an equation representing the relationship between the longitudinal acceleration, the driving force of the vehicle, the vehicle mass, and the acceleration error generated in the longitudinal acceleration sensor, the equation having the vehicle mass and the acceleration error as state variables; calculating an estimate of the longitudinal acceleration expected to be detected by the longitudinal acceleration sensor and a prior estimate of the state variable; calculating an estimate of the state variable based on the prior estimate and a value obtained by multiplying a correction gain by the difference between the estimated longitudinal acceleration and the detected value of the longitudinal acceleration detected by the longitudinal acceleration sensor; updating the state variable in the equation using the estimate of the state variable; calculating the correction gain via a Kalman filter algorithm that incorporates system noise and has the driving force as an input value; and varying the system noise based on a variable correlated with time.

2. A vehicle mass estimation method as set forth in claim 1, comprising the steps of: sequentially inputting the driving force into the vehicle model to sequentially calculate the estimated value of the longitudinal acceleration and the prior estimated value; sequentially calculating the difference based on the sequentially calculated estimated value of the longitudinal acceleration and the detected value of the longitudinal acceleration, and sequentially inputting the driving force into the algorithm of the Kalman filter to sequentially calculate the correction gain; sequentially calculating the estimated value of the state variable based on the sequentially calculated prior estimated value, the difference, and the correction gain; and sequentially updating the state variable in the equation using the sequentially calculated estimated value of the state variable.

3. A method for estimating vehicle mass according to claim 1 or claim 2, wherein the system noise is set to an initial value when the vehicle state changes, and then the system noise is changed based on the time-correlated variable.

4. The vehicle mass estimation method according to claim 3, wherein the time-correlated variable is set to the elapsed time since the system noise was set to the initial value, or the distance traveled by the vehicle since the system noise was set to the initial value.

5. A vehicle mass estimation method according to claim 1 or claim 2, wherein the system noise includes a first component related to the vehicle mass and a second component different from the first component, the first component is varied by a variable correlated with the time, and the second component is set to a predetermined value regardless of the variable correlated with the time.

6. A vehicle mass estimation method according to claim 1 or claim 2, wherein the system noise is set to an initial value when the vehicle state changes, and then the system noise is caused to converge to a final value lower than the initial value according to changes in the time-correlated variable.

7. A vehicle mass estimation method according to claim 6, wherein, in the case where the vehicle state changes multiple times, the system noise is set to the initial value when a previous change in the vehicle state occurs, and then, when a subsequent change in the vehicle state occurs, the system noise is reset to the initial value from the value of the system noise immediately before the subsequent change in the vehicle state occurs.

8. The vehicle mass estimation method according to claim 7, wherein the change in the vehicle state includes a transition of the vehicle ignition from an off state to an on state, and a transition of the vehicle door from an open state to a closed state.

9. A vehicle mass estimation system comprising: a longitudinal acceleration sensor that detects longitudinal acceleration of a vehicle; a driving force calculation unit that calculates a driving force of the vehicle; a vehicle model having an equation that represents the relationship between the longitudinal acceleration, the driving force, vehicle mass, and an acceleration error generated in the longitudinal acceleration sensor, the equation having the vehicle mass and the acceleration error as state variables, the vehicle model calculating an estimated value of the longitudinal acceleration that is expected to be detected by the longitudinal acceleration sensor when the driving force is input, and a priori estimated value of the state variable; and a state variable estimation unit that calculates the estimated value of the state variable based on the priori estimated value and a value obtained by multiplying a correction gain by the difference between the estimated value of the longitudinal acceleration and the detected value of the longitudinal acceleration detected by the longitudinal acceleration sensor, wherein the vehicle model updates the state variable in the equation using the estimated value of the state variable, and the state variable estimation unit calculates the correction gain via a Kalman filter algorithm that incorporates system noise and has the driving force as an input value, and the system noise changes based on a variable that is correlated with time.

Images