JP2009280097A - Steering parameter optimization system - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To attain extra seasoning of a steering system suited to characteristics of individual drivers actually driving vehicles in a market. <P>SOLUTION: A vehicle characteristics identification part 10 identifies a parameter of a vehicle model and a driver characteristics identification part 20 identifies a driver parameter expressing driving characteristics of a present driver. The vehicle parameter and the driver parameter are output to a correction steering amount estimation part 30, simulation by a model of a driver vehicle system is executed, and a correction steering amount relative to a target running path of the driver is calculated. The estimation result of the correction steering amount estimation part 30 is learned by a parameter learning part 40. The correction steering amount optimized by the learning is output to a parameter adaption part 50 to determine a change amount of the steering parameter. Thus, extra seasoning of the steering system suited to characteristics of individual drivers actually driving vehicles in a market can be attained. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

  The present invention relates to a vehicle steering parameter optimization system capable of changing a steering characteristic.

  In recent years, in vehicles such as automobiles, a steer-by-wire (SBW) or electronic power steering (EPS) system without mechanical coupling between a front wheel and a steering wheel is being developed in order to increase the degree of freedom in designing a steering system.

  Against this background, various proposals have been made so far to produce a steering feeling that feels good for the driver. For example, Patent Document 1 proposes an EPS system that enables a desired steering feeling setting even if the vehicle type is different.

Further, Patent Document 2 discloses a driver for the absence of mechanical coupling by estimating a self-aligning torque and the like from information such as a vehicle speed, a steering angle sensor, and a tie rod displacement sensor and realizing a steering reaction force with a motor. An SBW system that reduces the sense of discomfort has been proposed.
JP 2005-59688 A JP-A-10-226346

  By the way, originally, the optimum characteristics differ greatly from the steering characteristics depending on the driving characteristics of the individual drivers. For example, due to the concept of the car model, it is common to produce a quick and heavy steering feeling by reducing the steering gear ratio in a sports car and seasoning the relationship between the steering angle and the front wheels to a certain linear gear ratio. Thus, if it is an expert driver, it becomes possible to extract the maximum vehicle movement performance.

  However, for beginner drivers, corrective steering is frequently performed on highways, etc., due to the sensitivity of vehicle responses, and for female drivers, the weight of the steering wheel does not match the sensitivity, and steering cannot be performed as desired. The phenomenon occurs.

  The present invention has been made in view of the above circumstances, and an object of the present invention is to provide a steering parameter optimization system capable of realizing the seasoning of the steering in accordance with the characteristics of individual drivers who actually drive the vehicle in the market. Yes.

  In order to achieve the above object, a steering parameter optimization system according to the present invention is a vehicle steering parameter optimization system capable of changing a steering characteristic, and is based on a history of a state quantity of a running vehicle and a driving characteristic of a driver. A driver characteristic identification unit for identifying the driver, and a steering parameter for estimating the corrected steering amount with respect to the target travel locus of the driver based on the driving characteristic of the driver and the transmission characteristic of the running vehicle and minimizing the corrected steering amount And a steering parameter optimizing unit that learns and optimizes.

  According to the present invention, it is possible to optimize the steering parameters so as to minimize the correction steering amount of the driver who actually drives the vehicle in the market, and to realize the seasoning of the steering in accordance with the characteristics of each driver.

  Embodiments of the present invention will be described below with reference to the drawings. 1 to 5 relate to an embodiment of the present invention, FIG. 1 is an overall configuration diagram of a steering parameter optimization system, FIG. 2 is a functional block diagram of a driver-car model, and FIG. 3 is an overall operation of the system. FIG. 4 is an explanatory diagram showing characteristics of a steer-by-wire vehicle steering system, and FIG. 5 is an explanatory diagram showing steering angle waveforms before and after optimization.

  The steering parameter optimization system according to the present invention is mounted on a vehicle having a steering system by steer-by-wire (SBW) or electronic power steering (EPS), and controls the motion of the steering system so as to minimize the driver's steering correction amount. Optimize parameters (steering parameters; for example, steering gear ratio, power assist gain, torsion bar rigidity / viscosity, steering wheel damping, etc.).

  In the present embodiment, as shown in FIG. 1, the steering parameter optimization system 1 includes a vehicle characteristic identification unit 10, a driver characteristic identification unit 20, a corrected steering amount estimation unit 30, a parameter learning unit 40, and a parameter adaptation unit 50. As a main component, it is configured by a single computer system or a plurality of computer systems connected via a network or the like.

  Data from in-vehicle sensors such as the yaw rate sensor 2, the steering torque sensor 3, and the wheel speed sensor 4 are input to the vehicle characteristic identification unit 10 and the driver characteristic identification unit 20 via an in-vehicle communication network such as CAN (Controller Area Network). Is done. Furthermore, in addition to the data from these in-vehicle sensors, the driver characteristic identification unit 20 receives a recognition result by the external environment recognition device 5 that recognizes the traveling environment of the vehicle.

  The external environment recognition device 5 is a device for recognizing a traveling environment outside the own vehicle using external environment recognition information by a camera, millimeter wave radar, laser radar, etc., own vehicle position information / map information by a car navigation system, etc. Based on the information on the lane and the road white line recognized from the captured image outside the vehicle, the road information acquired by the vehicle or acquired by external communication, etc., the target driving locus is output by the driver.

  In addition, as a target driving | running | working locus | trajectory of a driver | operator, you may employ | adopt the locus | trajectory which the external environment risk minimizes as proposed by the present applicant in Japanese Patent Application No. 2007-221695. This technology uses intelligent technology to calculate a target trajectory within a certain time that minimizes the external environmental risk ahead of the vehicle behavior.

  Further, in this system, the external environment recognition device 5 is not necessarily an essential component, and without using the external environment recognition device 5, the travel locus actually traveled by the vehicle can be geometrically determined from the yaw rate, the vehicle body speed, and the like. The target trajectory may be obtained by applying a low-pass filter to the travel trajectory obtained geometrically. The low-pass filter is a filter for removing the influence of the driver's correction steering.

  The steering parameter optimization system 1 according to the present embodiment schematically identifies the parameters of a vehicle model that expresses the transfer characteristics of a running vehicle based on the data of the in-vehicle sensor in the vehicle characteristic identification unit 10, The driver characteristic identifying unit 20 identifies a driver parameter that represents the current driving characteristic of the driver based on the history of the state quantity of the vehicle that is running. These vehicle parameters and driver parameters are output to the corrected steering amount estimation unit 30, and the steering parameters are optimized by a steering parameter optimization unit formed by the corrected steering amount estimation unit 30 and the parameter learning unit 40.

  That is, the corrected steering amount estimation unit 30 performs a model simulation in which the driver-steering system-vehicle system is integrated using the vehicle parameters and the driver parameters, and calculates a corrected steering amount for the target travel locus of the driver. The estimation result of the corrected steering amount estimating unit 30 is learned by the parameter learning unit 40 and fed back to the corrected steering amount estimating unit 30. Then, the corrected steering amount optimized by this learning is output to the parameter adaptation unit 50, the change amount of the steering parameter is determined, and is output to the control target (steering control system) 60 that is the actual target.

  Hereinafter, the function of each part of the steering parameter optimization system 1 will be described in detail.

  The vehicle characteristic identification unit 10 calculates unknown vehicle parameters such as cornering power, vehicle weight, and moment of inertia around the z-axis of a vehicle model that expresses the transfer characteristics of a traveling vehicle. This vehicle parameter is a steering that is an input to the vehicle with respect to a vehicle steering system model (which depends on the dynamics of the applicable vehicle type and to which the current vehicle steering parameters are applied) and a vehicle system model connected. Simulation is performed with torque and body speed set, and repeated using the gradient method or genetic algorithm (GA) so that the error between the actual vehicle yaw rate, which is the vehicle response, and the simulation value is sufficiently small Decide while performing calculations.

As a vehicle model, for example, a simulation is performed in which the steering torque and the vehicle body speed are input using the equations of motion expressed by the following equations (1) to (5), and the simulation calculation value of the yaw rate, which is the vehicle response, The calculation is repeated so that the error from the measured value is minimized.
mV (dβ / dt + γ) = − 2K _f β _f −2K _r β _r (1)
Idγ / dt = −2L _f K _f β _f + 2L _r K _r β _r (2)
β _f = β + L _f γ / V−δ _f (3)
β _r = β−L _r γ / V−δ _r (4)
SAT = (ζ _n + ζ _c ) K _f β _f (5)
Where m: mass
I: Moment of inertia around the z axis
V: Body speed
K _f : Front wheel cornering stiffness
K _r : Rear wheel cornering stiffness
L _f : Distance from the center of gravity to the front wheels
L _r : Distance from the center of gravity to the rear wheel
β: Body slip angle
β _f : Front wheel skid angle
β _r : Rear wheel skid angle
γ: Yaw rate
δ _f : Front wheel actual rudder angle
δ _r : rear wheel actual rudder angle
SAT: Self-aligning torque
ζ _n : pneumatic trail
ζ _c : caster trail

The above vehicle model defines a front wheel actual steering angle δ _f and a rear wheel actual steering angle δ _r on the assumption that a four-wheel steering vehicle is used. The actual wheel steering angle δ _r is always 0. In this embodiment, the pneumatic trail ζ _n and caster trail ζ _c in the equation (6) for calculating the self-aligning torque SAT are treated as constants. This is because, in actual vehicle motion, it varies slightly depending on the running state, but in this system, even if it is treated as a constant, the operation is not affected.

  Further, the vehicle specifications during traveling usually vary somewhat depending on the road surface and the number of passengers, but by setting the vehicle specifications (vehicle parameters) at the time of development as vehicle model parameters in advance, the vehicle characteristic identification unit 10 can be used. It can be omitted.

  On the other hand, the driver characteristic identification unit 20 calculates a driver parameter expressing the current driving characteristic of the driver using a driver-vehicle system transfer function model. Specifically, the driving operation of the driver is considered to be a proportional operation with a reaction delay (dead time) plus a differential operation that predicts changes in input and an integration operation that corrects stationary deviations. As shown in FIG. 2, modeling is performed by a feedforward steering unit 21 based on target course information (target travel locus information) and a feedback steering unit 22 based on relative information between the target course and the vehicle.

In this driver model, the target course curvature φ _R and the target course position information Y _R obtained from the external environment recognition device 5 are input to the feedforward steering unit 21 and the feedback steering unit 22, respectively. The model output obtained by adding the feedforward steering component from 21 and the feedback steering component from the feedback steering unit 22 is the steering torque Td of the driver.

  The “corrected steering amount” to be optimized in this system means a feedback steering component by this driver model. The transfer function gain K 0 and dead time t 0 in the feedforward steering unit 21, and the transmission in the feedback steering unit 22. The gains K1, K2, K3 of the function and the dead time t1 are model parameters to be identified.

  Below, the parameter derivation method in each of the feedforward steering unit 21 and the feedback steering unit 22 will be described.

  In the driver model shown in FIG. 2, the driver output is expressed by the steering torque, but the steering angle may be used as the driver output instead of the steering torque in accordance with the steering system of the vehicle. In the driver model based on the transfer function shown in FIG. 2, it is essential to separate steering into feed forward and feedback. However, the type and number of input information and the number of differentiation / integration are determined by the designer. For example, the feedback steering may be changed to a general forward-gaze secondary prediction model in which steering is performed according to the error between the target value and the predicted value at the forward gaze point.

<Feed forward parameter>
In the feedforward steering unit 21, the cross-correlation function between the steering torque value Td measured by the steering torque sensor 3 and the course shape (curvature etc.) measured by the external environment recognition device 5 is maximized in the traveling vehicle. The dead time t0 is determined, and then the gain K0 is obtained by the method of least squares. The cross-correlation function is a function used in a general mathematical method, and shows the temporal transition of the correlation coefficient calculated while slightly shifting the time of one of the two variables.

In this system, when the steering torque value Td with respect to the curvature φ _R of the target course has no delay / advance time, the dead time t0 is 0 and the maximum matching is shown. Due to the operation, for example, a delay or advance of about 0.5 to 3 seconds occurs.

<Feedback parameter>
The feedback steering unit 22 compensates the steering torque value Td for an error that cannot be expressed by feedforward steering using the above-described parameters. Specifically, the gains K1 to K3 are obtained using multiple regression analysis, which is a general technique of multivariate analysis for estimating an objective variable from two or more explanatory variables.

In the multiple regression analysis, a relational expression between the objective variable Y and two or more explanatory variables xi (i = 1, 2, 3,..., N) affecting the objective variable Y is expressed by the following expression (6). This is a method of setting and estimating coefficients (partial regression coefficients) a i (i = 1, 2, 3,..., N) using observation data of the variable Y. For example, when there are three explanatory variables xi, the objective variable Y corresponds to the error between the steering torque and the feedforward steering torque, the partial regression coefficient ai corresponds to the gains K1 to K3, and the explanatory variable xi is the target course. Difference between position and vehicle position (YR-YV), integrated value of difference between target course position and vehicle position (YR-YV) / s, differential value of difference between target course position and vehicle position (YR-YV) It corresponds to s. For the dead time t1, the delay or the lead time at which the correlation between the error and the result of the multiple regression analysis is maximized is obtained by a cross-correlation function.
Y = a1x1 + a2x2 + a3x3 + ... + anxn (6)

  In the above, the feedforward parameter and the feedback parameter are determined by different methods, but by GA (genetic algorithm), by optimizing generational changes using these parameter groups as genes, The feedforward parameter and the feedback parameter may be obtained together.

  Next, the corrected steering amount estimation unit 30 uses the vehicle model parameters obtained by the vehicle characteristic identification unit 10 and the driver parameters obtained by the driver characteristic identification unit 20 to perform a simulation using the driver-vehicle system model of FIG. To calculate and output the corrected steering amount at that time. This simulation is performed every time the corrected steering amount is learned and changed by the parameter learning unit 40, and the calculation is repeated until the ratio of the corrected steering amount (feedback steering amount) to the output of the driver model is minimized.

  The driver-car model shown in FIG. 2 includes a steering system model and a vehicle model as the car model. The steering system model receives the steering torque Td from the driver model and the self-aligning torque SAT from the vehicle body model, and outputs the tire steering angle δ to the vehicle body model. The vehicle model calculates the self-aligning torque SAT and outputs it to the steering system model, and outputs information on the vehicle position Yv and the traveling course curvature Φν. The vehicle position Yv is output to the driver model, and the deviation between the target course position YR and the vehicle position YV is input to the feedback steering unit 22.

  The parameter learning unit 40 repeatedly learns and changes the steering parameter by GA (genetic algorithm) based on the corrected steering amount calculated by the simulation in the corrected steering amount estimation unit 30, and outputs the learned steering parameter. .

The parameter adaptation unit 50 compares the corrected steering amount calculated in the simulation with a preset threshold value. When it is determined that the corrected steering amount is less than the threshold value, it is assumed that the optimization of the steering parameter by learning has been completed. The change amount DP _n of the steering parameter is determined according to the equation (7).
_{DP n = Kp (1-e} -t / T) (Pnew n -Pold n) ... (7)
Where n: parameter number
t: Time from the start of parameter change
Pnew: Changed vehicle parameters
Pold: Current vehicle parameters
Kp, T: Coefficient that determines the parameter change speed
(The change speed increases as Kp increases and T decreases)

  The optimum steering parameter by the above processing is not continuously calculated in real time, but is calculated at a predetermined frequency, for example, once every several days or several hours as background processing during traveling. The optimum steering parameter calculated by the background process is adapted to the vehicle while the parameter adaptation unit 50 takes a sufficient amount of time.

  The overall operation of this system is as shown in the flowchart of FIG. That is, when the operation of the present system is started, first, in the first step S1, measurement data for a predetermined time, for example, 1 to 2 minutes (data from the in-vehicle sensor and the target course and vehicle state of the vehicle based on the data) Quantity) is stored in the memory, and in steps S2A and S2B, processing for identifying the parameters of the vehicle model and the parameters of the driver model is executed in parallel, respectively.

  Next, after identifying the parameters of the vehicle model and the parameters of the driver model, a driver-vehicle system simulation is executed in step S3 to calculate a corrected steering amount. In step S4, the corrected steering amount becomes less than the threshold value. Check if it was. As a result, if the corrected steering amount is greater than or equal to the threshold value, the steering parameter is changed in step S5, and the simulation is executed again in step S3. Then, after changing the steering parameter and repeating the simulation, when the corrected steering amount becomes less than the threshold, the process proceeds from step S4 to step S6, and the steering parameter obtained by the simulation is applied to the vehicle.

[Example of determining steering parameters]
Next, an example in which the present system is applied to a steer-by-wire system vehicle will be described. The steer-by-wire system vehicle does not have a mechanical connection between the front wheels and the steering, and the steering reaction force and the front wheel steering angle are independently controlled by an actuator.

As shown in FIG. 4, the steering reaction force and the front wheel rudder angle are dynamic movements in which a torsion bar system 8 and a steering rack system 9 are equivalently interposed between the steering wheel 6 and the left and right front wheels 7a and 7b. It is calculated and controlled based on a steering system model expressed by an equation. The equation of motion of such a steering system model is illustrated in the following equations (8) and (9).
m _rack · d ² x / dt ² = T _driver · N + F _{p_steer} + F _left + F _right (8)
_{T driver = C s · (δ} -N · x) + K s · (dδ / dt-N · dx / dt) + C d · dδ / dt ... (9)
However, m _rack : Steering rack mass
x: Steering rack position
N: Steering gear ratio
δ: Steering angle
T _driver : Steering torque
F _{p_steer} : Power steering assist force
F _left : Front wheel left axial force
F _right : Front wheel right axial force
C _s : Torsion bar viscosity
K _s : Torsion bar rigidity
C _d : Steering wheel damping

The power steering assist force F _{p_steer} is calculated by applying the power assist gain K _p to the torsion bar twist angle.

For example, this system is applied to a steer-by-wire vehicle in which the initial steering parameter is set to a quick response such as a sports car, and the steering gear ratio among the steering parameters constituting the equation of motion of the steering system model. When N, the power assist gain K _p , the tote bar stiffness K _s , the tote bar viscosity C _s and the steering wheel damping C _d are optimized, the result shown in FIG. 5 is obtained.

  FIG. 5 (a) is a waveform showing the time change of the steering angle when the novice driver drives the vehicle before adapting to this system and performs a double lane change. It can be seen that every minute correction steering is repeated so as not to deviate from the course. On the other hand, FIG. 5B shows the time change of the steering angle of the driver after the vehicle parameter is changed by learning of this system, and there is no correction steering and the waveform is smooth. I can confirm. In this way, it is possible to determine a steering parameter that minimizes the corrected steering amount from the driving characteristics of the driver who is actually driving the vehicle.

  As described above, in this system, it is possible to adapt the parameters in consideration of the compatibility between the steering specifications that are difficult to consider when developing the vehicle and the driver who actually drives the vehicle. Thereby, for example, even when a beginner driver drives a vehicle with a sensitive vehicle response such as a sports car, the correction steering can be minimized. Even if the driving characteristics change in the process of getting used to beginners, it is possible to obtain optimum steering parameters according to the driver characteristics at that time by performing online parameter tuning each time.

  Therefore, regardless of the vehicle characteristics at the time of shipment from the factory, it is possible to achieve a steering seasoning that matches the characteristics of individual drivers who actually drive the vehicle in the market. It becomes possible. Furthermore, even during vehicle development, by applying this system to a large number of subjects and determining the optimal steering parameters, it is possible to significantly reduce the number of man-hours for design and experiments, and to improve costs while improving quality. Reduction can be achieved.

Overall configuration diagram of steering parameter optimization system Functional block diagram of driver-car model Flow chart showing overall system operation Explanatory drawing showing characteristics of steer-by-wire vehicle steering system Explanatory drawing showing the steering angle waveform before and after optimization

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Steering parameter optimization system 10 Vehicle characteristic identification part 20 Driver characteristic identification part 21 Feedforward steering part 22 Feedback steering part 30 Correction steering amount estimation part 40 Parameter learning part 50 Parameter adaptation part

Claims (9)

A steering parameter optimization system for a vehicle capable of changing a steering characteristic,
A driver characteristic identification unit that identifies the driving characteristic of the driver based on the history of the state quantity of the vehicle in progress;
Steering parameters for estimating and correcting a steering parameter that minimizes the corrected steering amount based on the driving characteristics of the driver and the transmission characteristics of the running vehicle A steering parameter optimization system comprising an optimization unit.   2. The steering parameter optimizing system according to claim 1, further comprising a vehicle characteristic identifying unit for identifying a transfer characteristic of the vehicle.   3. The steering parameter optimization system according to claim 1, wherein the target travel locus is determined based on a recognition result of an external environment of the vehicle.   The steering parameter according to any one of claims 1 to 3, wherein the driver characteristic identification unit identifies the driving characteristic of the driver based on a driving operation model of the driver by feedforward steering and feedback steering. Optimization system.   5. The steering parameter optimization system according to claim 4, wherein the driver characteristic identification unit determines a control parameter for the feedforward steering using a least square method and a cross-correlation function.   6. The steering parameter optimization system according to claim 4, wherein the driver characteristic identification unit determines a control parameter for the feedback steering using multiple regression analysis.   5. The steering parameter optimization system according to claim 4, wherein the driver characteristic identification unit collectively identifies both the feedforward steering control parameter and the feedback steering control parameter by a genetic algorithm.   3. The steering parameter optimization system according to claim 1, wherein the target travel locus is determined using road information acquired from a navigation device.   3. The steering parameter optimization system according to claim 1, wherein the target travel locus is determined using road white line information.

Images