JP3246042B2 - Rear wheel steering control device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a rear-wheel steering control device for a front-rear-wheel steering type vehicle in which a rear wheel can be steered in addition to a front wheel.

[0002]

2. Description of the Related Art A front-rear-wheel steering vehicle (4WS vehicle) of this type includes a rear-wheel steering actuator for steering rear wheels, and the rear-wheel steering actuator receives a target rear-wheel steering amount from a 4WS controller. Then, the rear wheels are steered so that the actual rear wheel steering amount matches the target rear wheel steering amount. When the rear wheels are properly steered in addition to the steering of the front wheels, the turning performance and the steering stability performance of the 4WS vehicle are both improved as compared with a normal 2WS vehicle.

[0003] In order to appropriately control the steering of the rear wheels, the 4WS controller sets the target rearward position based on information on the steering wheel angle indicating the steering state of the front wheels, and information on the vehicle speed and the yaw angular speed of the vehicle indicating the motion state of the vehicle. The wheel steering amount is determined. By the way, the 4WS controller determines the target rear wheel steering amount from various kinds of detected information based on the control law of the rear wheel steering considering only the dynamic performance of the vehicle itself. When the rear wheels are actually steered by the actuator, depending on the driving situation,
Thereafter, the wheel steering may give a great sense of discomfort to the user driver of the vehicle.

[0004] Accordingly, by multiplying the target rear wheel steering amount determined by the 4WS controller by a constant gain, the target rear wheel steering amount is uniformly reduced and corrected.
Generally, the correction value is reset to a new target rear wheel steering amount to control the rear wheel steering.

[0005]

However, 4WS
When the target rear wheel steering amount determined by the controller is uniformly reduced and the rear wheel steering is controlled, the rear wheel steering is undesirably suppressed even in an area where the user driver does not feel uncomfortable. As a result, the effect of the original steering control of the 4WS vehicle is not exhibited.

The present invention has been made in view of the above circumstances, and has as its object to provide a rear wheel steering control capable of fully exerting rear wheel steering control without giving a user driver an uncomfortable feeling. It is to provide a device.

[0007]

A rear wheel steering control device according to the present invention comprises: first detecting means for detecting a steering state of a front wheel;
Second detecting means for detecting various motion states of the vehicle, and rear wheel steering amount for calculating the target rear wheel steering amount based on front wheel steering information and motion information detected by the first and second detecting means, respectively. Calculating means, gain calculating means for calculating the control gain of the target rear wheel steering amount calculated by the rear wheel steering amount calculating means, and the target rear wheel steering amount from the rear wheel steering amount calculating means; A value obtained by adjusting the target rear wheel steering amount based on the control gain from the gain calculating means is determined as a new target rear wheel steering amount, and the determined target rear wheel steering amount is determined by a rear wheel steering actuator of a rear wheel steering system. Determining means for providing Then, the gain calculation unit inputs the front wheel steering steering information <br/> Ho及beauty exercise information detected by said first and second detection means, teacher data teachers gain adjusted by sensory evaluation driver When the front wheel steering information and the motion information are input, the learned neural network outputs the control gain corresponding to the teacher gain. So
Then, the gain detecting means performs the first and second detecting steps.
The front wheel steering information and the rear wheel steering information detected at the
Before entering the neural network, these front wheels
Pre-processing means for processing steering information and motion information is further provided.
The pre-processing means includes the front wheel steering information and the front wheel
Calculate the approximate value of the vehicle body weight center slip angle from the exercise information,
This approximation is used as one input to the neural network.
And an approximation value calculating means for calculating the approximate value.

[0008]

According to the above-described rear wheel steering control device, when the front wheel steering information and the vehicle motion information are detected by the first and second detection means during the running of the vehicle, these information are either The gain calculation means, that is, a learned neural network, is provided as an input, and the learned neural network outputs a control gain corresponding to the teacher gain of the sensory evaluation driver. On the other hand, the front wheel steering information and the motion information are supplied to a rear wheel steering amount calculating means, and the rear wheel steering amount calculating means outputs a target rear wheel steering amount. The target rear-wheel steering amount and the control gain are both supplied with determining means, the determining means rear wheel steering amount target rear-wheel steering amount value adjusted by the control gain new target rear-wheel steering from the calculating means The amount is output to the rear wheel steering actuator of the rear wheel steering system, and as a result, the rear wheel steering actuator steers the rear wheels according to the new target rear wheel steering amount. Here, the neural network has an approximation
Calculated from front wheel steering information and rear wheel steering information by value calculation means.
The approximate value of the vehicle body weight center slip angle that was issued is also
This approximation is based on the vehicle motion.
Parameters, which are important for understanding
The learning effect of the neural network is enhanced.

[0009]

FIG. 1 schematically shows a front-rear-wheel steering vehicle (4WS vehicle) according to an embodiment of the present invention. The 4WS vehicle has a power cylinder 2 and a rack
Including a pinion mechanism, the front wheel F
W _L, a front wheel steering system 6 for steering the FW _R, the wheel steering actuator 8 after consisting hydraulic cylinders, the rear wheels RW _L,
The RW _R and a wheel steering system 10 after steering.

The rear wheel steering system 10 will be described in more detail. The rear wheel steering actuator 8 is combined with a rear wheel steering angle sensor 11 and an electromagnetic valve 12. Is connected to the control unit 14 via the feedback controller 13.
Is electrically connected to Wheel RW _L after being determined by the control unit 14 to the feedback controller 13, given the actual rear wheel steering angle and the rear wheel steering amount or target rear-wheel steering angle target, detected by the rear wheel steering angle sensor 11 of the RW _R Then, the feedback controller 13 controls the opening direction and the opening degree of the electromagnetic valve 12, that is, controls the operation of the rear wheel steering actuator 8, and sets the actual rear wheel steering angle to match the target rear wheel steering angle. wheel RW _L, to steer the RW _R.

Various sensors are connected to the control unit 14 for determining the target rear wheel steering angle. These sensors include a steering wheel angle sensor 16, a vehicle speed sensor 18, a longitudinal acceleration sensor 20, There are a lateral acceleration sensor 22 and a yaw angular velocity sensor 24. Steering wheel angle sensor 16 detects the operation angle or the steering wheel angle theta _H of the handle 4, a vehicle speed sensor 18 detects the speed of the vehicle, the vehicle speed V.
Here, to the vehicle speed sensor 18 may be a speedometer of the vehicle, or, in the case where the wheel speed sensor is mounted on each wheel, the front wheels FW _L, the output of the wheel speed sensor FW _R Front wheel speed obtained by averaging the front wheel speed and rear wheel RW _L , R
W may be the vehicle speed V the smaller of the average rear wheel speed obtained by averaging the output of the wheel speed sensors of _R.

Longitudinal acceleration sensor 20, lateral acceleration sensor 2
2 and the yaw angular velocity sensor 24 detect the longitudinal acceleration G _{Z of} the vehicle body, the lateral acceleration G _Y of the vehicle body, and the yaw angular velocity ψ of the vehicle body, respectively. Therefore, the control unit 14 receives the steering wheel angle θ _H , the vehicle speed V, the longitudinal acceleration G _Z , the lateral acceleration G _Y, and the yaw angular velocity ψ from each sensor. Therefore, the control unit 14 calculates the target rear wheel steering angle based on the detection signals from the various sensors described above.

More specifically, as shown in FIG. 2, the control unit 14 includes a 4WS controller 25,
Correction block 27, approximate expression calculation block 26, pre-processing block 28, memory 30, and neural network 3
2 is provided. Detection signals from the steering wheel angle sensor 16, the vehicle speed sensor 18, and the yaw angular velocity sensor 24 are input to the 4WS controller 25, and a target rear wheel steering angle δr is calculated from these detection signals based on, for example, the following equation. δr = K ₁ × θ _H + K ₂ × (ψ _O −ψ) ψ _O : target yaw angular velocity determined from vehicle speed V and steering wheel angle θ _H K ₁ , K ₂ : correction coefficient determined from vehicle speed V 4WS controller Target rear wheel steering angle δ obtained at 25
r is output to the correction block 27, and this correction block 2
In step 7, the target rear wheel steering angle δr is multiplied by a control gain to correct the target rear wheel steering angle δr, and the corrected value is set as a new target rear wheel steering angle δr in the feedback controller 13.
To supply.

On the other hand, detection signals from the steering wheel angle sensor 16, the vehicle speed sensor 18, and the yaw angular speed sensor 24, that is, the steering wheel angle θ _H , the vehicle speed V, and the yaw angular speed ψ are also input to the approximate expression calculation block 26, respectively. , The approximate expression calculation block 26 calculates an approximate value A of the vehicle body weight center slip angle β. The vehicle center-of-gravity slip angle β indicates the angle between the traveling direction of the vehicle and the direction of the vehicle body, as shown in FIG.

The approximation formula calculation block 26 is provided with an approximation formula for calculating the approximation value A.
It is derived from the general linear two-degree-of-freedom vehicle model of FIG. That is, the equation of motion of this vehicle model is described by the following two equations. m × V (dβ / dt) + 2 × (Kf + Kr) × β + {m × V + (2 / V) × (Lf × Kf−Lr × Kr)} × ψ = 2 × Kf × δf (1) 2 × (Lf × Kf−Lr × Kr) × β + I × (dψ / dt) + (2 / V) × (Lf ² × Kf + Lr ² × Kr) × ψ = 2 × Lf × Kf × δf (2) ) M: body mass, Kf, Kr: equivalent cornering power of front / rear wheels I: yaw moment of inertia, Lf: distance between the center of gravity of the front axle Lr: distance between the center of gravity of the rear axle, δf: front wheel steering angle Now, the following equation δf = δf _MAX × (1-e ^−2t ) (3) δf _MAX : maximum steering angle, and the step-like steering input as shown in FIG. 5 is expressed by the equations (1) and (2). By linearly analyzing the response when given, the solutions of the yaw angular velocity ψ and the vehicle body weight center slip angle β in the time domain are obtained, and an approximate expression expressed by the following equation can be obtained from these two solutions. A = β ₀ × ψ + K ₀ × ｛ψ − {(1-e ^−2t )} (4) β ₀ : (β steady value obtained from vehicle speed V and front wheel steering angle δf) /
Yaw angular velocity ψK ₀ : correction coefficient in transient state for the first term Accordingly, when vehicle speed V, front wheel steering angle δf, and yaw angular velocity に are given to the approximate expression of equation (4), approximate value A of vehicle center-of-gravity slip angle β is obtained. Can be calculated. Note that the front wheel steering angle δf can be obtained by performing an arithmetic processing from the steering wheel angle θ _H in the approximation formula calculation block 26. Therefore, the approximate value A is obtained by calculating the vehicle speed V, the steering wheel angle θ _H and the yaw angular speed ψ. When given, it can be calculated from the approximate expression of Expression (4).

The approximate value A calculated by the approximate expression calculation block 26 is given to the neural network 32 as one input information. The neural network 32 is also provided with input information other than the approximate value A, and these other input information are generated by the pre-processing block 28 and the memory 30.

That is, the steering wheel angle θ _H from the steering wheel angle sensor 16 is calculated by the arithmetic processing in the pre-processing block 28.
After being converted into the front wheel steering angle δf, it is stored in the memory 30 in a time series. For example, the memory 30 stores the front wheel steering angle δf ₀ at the current time, the front wheel steering angles δf ₁ , t ₂ before t ₁ (0.1 second).
The front wheel steering angle δf ₂ before the time (0.2 seconds), t ₃ hours (0.3
The time series data of the front wheel steering angle δf ₃ before (seconds) and the front wheel steering angle δf ₄ before t ₄ (0.4 seconds) are stored as parameters of the vehicle. Is entered.

On the other hand, the vehicle speed V, the longitudinal acceleration G _Z , the lateral acceleration G _Y and the yaw angular velocity か ら from the vehicle speed sensor 18, the longitudinal acceleration sensor 20, the lateral acceleration sensor 22 and the yaw angular velocity sensor 24 are given directly to the pre-processing block 28. And
The pre-processing block 28 includes, as other vehicle speed parameters, lateral acceleration-vehicle speed x yaw angular velocity (= G _Y -V x ψ),
Vehicle speed x yaw angular speed (= V x ψ), yaw angular speed / vehicle speed (=
ψ / V), yaw angular acceleration (= ψ _A ), and longitudinal acceleration × yaw angular velocity (= G _Z × ψ) are calculated and input to the neural network 32.

The neural network 32 outputs a control gain C (0 ≦ C ≦ 1) based on the input information described above, and gives the control gain C to the correction block 27. Since the approximate value A of the vehicle body weight center slip angle β given to the neural network 32 is a state quantity defining the yaw motion and the lateral motion of the vehicle, the approximate value A is important input information for calculating the control gain C. The input information other than the value A is determined in consideration of the following equations obtained by respectively modifying the equations (1) and (2). β = {1/2 × (Kf + Kr)} × {2 × Kf × δf− mx × V × (dβ / dt) −mx × V × ψ− 2 (Lf × Kf−Lr × Kr) × ( ψ / V )} ··· (5) β = {1/2 × (Kf × Lf + Kr × Lr)} × {2 × Lf × Kf × δf -I × ψ A -2 × (Lf 2 × Kf + Lr 2 × Kr) × ( Ψ / V )｝ (6) From the above equations (5) and (6), the vehicle body weight center slip angle β
Is described as a linear sum of the understated state quantities.

In the approximate expression, the coefficient of the state quantity is a constant, whereas in an actual vehicle, it can be considered that the coefficient is a function of the non-linearity of the tire, the starting of the load, and the like. Here, if the state quantity is taken as the input information to the neural network 32, the neural network 32 can appropriately correct the error between the vehicle body weight center slip angle β and the approximate value A.

The structure of the neural network 32 includes:
A coordinate system transformation from a vehicle speed V whose sign is reversed in the front-rear direction to a vehicle center-of-gravity slip angle β whose sign is reversed in the left-right direction is included, and a function from the input to the output of the neural network 32 is simplified. Can be. Among the state quantities, the front wheel steering angle δf is an input to the vehicle, and thus is given to the neural network 32 as a time-series pattern as described above.

With respect to the state quantity of V × (dβ / dt) in the equation (5), assuming that the slip angle β of the vehicle center of gravity is sufficiently small, the lateral acceleration G _Y becomes V × (dβ / dt + ψ).
Dβ / dt = G _Y / V−ψ, and the state quantity of V × (dβ / dt) can be expressed by V × (dβ / dt) = G _Y −ψ × V it can.

Further, the neural network 32 includes
The longitudinal acceleration G not considered in the linear two-degree-of-freedom model
state quantity of G _z × [psi about _z are also given as input information. The neural network 32 is of a hierarchical type as schematically shown in FIG. 6, and includes an input layer having an input unit IU for each input information and an intermediate layer having several intermediate units MU. And an output layer having one output unit OU for outputting the correction value C.

The strength of the connection between the units in the neural network 32 is learned in advance by a back propagation method, and this learning is performed using a learning system shown in FIG. This learning system is shown in FIG.
A gain adjustment dial 29 is provided in addition to the functional blocks and the sensor group of the control unit 14 shown in FIG. The gain adjustment dial 29 is operated by a sensory evaluation driver, and the teacher gain selected by the sensory evaluation driver is provided to the neural network 32 as teacher data.

That is, at the time of learning, the teacher gain C _T (0 ≦ C _T ≦ 1) from the gain adjustment dial 29 is input to the correction block 27 and the subtraction unit 31, and the subtraction unit 31 uses the teacher gain C _T and the neural network. An error between the control gain C from the network 32 and the error E is calculated, and the error E is returned to the neural network 32. Also, when learning,
In the correction block 27, as shown in the following equation is multiplied teachers gain C _T to the target rear-wheel steering angle δr from 4WS controller 25, after the new target wheel steering angle δr is calculated.

Δr = C _T × δr During learning, the target rear wheel steering angle δr thus determined
Is output as a control signal to the rear wheel steering system i.e. the rear wheel steering actuator 8, as a result, the rear wheel steering actuator 8, the rear wheel RW _L, actual rear wheel steering angle target rear-wheel steering angle RW _R Feedback control is performed so as to match Δr.

[0027] Thus, sensory evaluation driver various driving conditions, repeated running test at the road surface condition and the front wheel steering pattern, as then In the situation where wheel steering discomfort reducing the teacher gain C _T , by operating the gain adjusting dial 19, in a situation with no uncomfortable feeling to reverse while operating the gain adjusting dial 19 so as to increase the teacher gain C _T, repeated learning of the neural network 32.

As is well known, the backpropagation learning rule using the sigmoid function learns the strength (coupling amount) ω of the connection between the units in the neural network 32 so that the error E is minimized. Is what you do.
More specifically, in the neural network 32, first, the output value X (X _I , X _M , X _O ) of each unit (IU, MU, OU) of each layer is calculated. These output values X are obtained as the value of a sigmoid function using the accumulated value as a parameter by accumulating the product of the input to each unit and the coupling amount in each layer.

As for the error E _O of the output unit OU, the teacher gain C _T and the output value X _{O of the} output unit OU are obtained.
Based on the differential value F _O of the sigmoid function expressing the input and output relationship (= control gain C) and the output unit OU, it is calculated from the following equation. _{E 0 = (C T -X 0} ) × F 0 With respect to the error E _M of each intermediate unit MU of the intermediate layer, the differential value F _M of the sigmoid function expressing the input-output relation of each intermediate unit MU, the output unit OU the amount of coupling between the error E _O, and the respective intermediate units MU and the output unit OU of ω
It is calculated from the following equation based on _OM . E _M = F _M × Σ (E _O × ω _OM ) Further, regarding the error E _I of each input unit IU of the input layer, the differential value F _{I of} the sigmoid function representing the input / output relationship of each input unit IU, and each intermediate value The error E _{M of} the unit MU, and the coupling amount ω between each intermediate unit MU and each input unit IU
It is calculated from the following equation based on the _MI . E _I = F _I × Σ (E _M × ω _MI ) Next, the learning constant η and each unit (IU, MU, O
U), and the output value X (X _I , E _I , E _M , E _O ) of each unit (IU, MU, OU) of each layer.
X _M , X _O ), each unit (IU, MU,
OU) of input data ω (ω _ID , ω _MI , ω _OM )
(Δω _ID , Δω _MI , Δω _OM ) is calculated by the following equation. Note that the coupling amount ω _ID indicates the coupling amount with the input data of the input unit IU. Δω = η × E × X The calculated correction amount Δω is added to the coupling amount ω (ω _ID , ω _MI , ω _OM ) with the input data of each unit, and the updated coupling amount ω (= ω + Δω) is obtained. can get.

Accordingly, since the control gain C output from the learned neural network 32 has a value corresponding to the teacher gain C _T selected by the sensory evaluation driver, as shown in FIG. The control gain C output from the learned neural network 32 and the target rear wheel steering angle δ output from the 4WS controller 25
is supplied to the correction block 27, the correction block 27 outputs a new target rear wheel steering angle δr with a gain adjusted as a control signal to the rear wheel steering system, that is, the rear wheel steering actuator 8. as a result, the rear wheel steering actuator 8, the rear wheel RW _L, is feedback controlled so that the actual rear-wheel steering angle of the RW _R matches the target rear wheel steering angle [delta] r.

The above-described rear wheel steering control can be represented by a steering control routine shown in FIG. Steering Control Routine First, in step S1, when the steering wheel angle θ _{H, that} is, the front wheel steering angle δf is input to the control unit 14, the counter value i is set to 0 (step S2), and the input front wheel steering angle is set. δf is stored or saved in the memory 30 (step S3). The front wheel steering angle δf is calculated by the pre-processing block 28 from the steering wheel angle θ _H as described above.

In the next step S4, after waiting (waiting) for 0.1 second, the value i of the counter is incremented by 1 (step S5), and whether or not the value i of the counter is smaller than 4 Is determined (step S
6). If the determination result of step S6 is true (Yes),
The steps after step S1 are repeated, and when the determination result is false (No), the process proceeds to step S7. Therefore, when the determination result of step S6 becomes false, the memory 30
, Time series data relating to the front wheel steering angle δf, that is, δf ₀ , δf
₁ , δf ₂ , δf ₃ , δf ₄ are saved.

In step S7, the vehicle speed V, the longitudinal acceleration _GZ , the lateral acceleration _GY, and the yaw angular velocity ψ from the sensors 18, 20, 22, and 24 are input to the control unit 14, respectively. Approximate value A
After There is calculated (step S8), and the preprocessing calculation, G _Y -V × ψ described above, V × ψ, ψ / V , ψ A, G Z × ψ is obtained (step S9), and these values And the time series data are given as input information to the learned neural network 32, and the neural network 32
Calculates and outputs the control gain C (step S1).
0). Then, the 4WS controller 25 calculates a target rear wheel steering angle δr (step S11), and multiplies the target rear wheel steering angle δr by the control gain C to calculate a new target rear wheel steering angle δr. (Step S12).

[0034] Thereafter, the target rear wheel steering angle δr is, (step S13) by being output to the feedback controller 13, the rear wheel RW _L, RW _R starts to be actually steered. The feedback controller 13 determines whether or not the actual rear wheel steering angle matches the target rear wheel steering angle δr, and controls the rear wheel steering system 10 so that the determination result becomes true. In the next step S14, the control termination condition is determined.
The termination condition may be, for example, turning off an ignition switch of the vehicle.

According to the rear wheel steering control device described above, 4 W
Since the target rear wheel steering angle δr calculated by the S controller 25 is corrected by the output of the learned neural network 32, that is, the control gain C, the corrected target rear wheel steering angle δr becomes the vehicle itself. It is determined from a non-linear map that takes into account not only the exercise performance but also the driver's own preference. That is, in a situation where the rear wheel control based on the target rear wheel steering angle δr gives the driver an uncomfortable feeling, the control gain C is lowered, and thereafter the wheel steering is suppressed, and conversely, the driver does not feel uncomfortable. By increasing the control gain C, the rear wheel steering is performed without restraint,
The effect of the steering control can be maximized.

Further, since one of the input information of the neural network 32 employs the vehicle weight center slip angle β, which is an important parameter for grasping the motion state of the vehicle, that is, its approximate value A, The learning efficiency of the neural network 32 can be improved. Further, since the input information processed in the pre-processing block 28 is given to the neural network 32, the above-described input information from the pre-processing block 28 is used to calculate the approximate value A of the vehicle body weight center slip angle β. impact and of the longitudinal acceleration G _Z and the lateral acceleration G _Y with respect deviation, the motion state of the motion state and the left and right systems before and after system of a vehicle can be set in consideration of the unified processing on one system, thereby, the neural Not only can the load on the network 32 be reduced, but also the output accuracy can be increased.

Further, not only the current front wheel steering angle δf ₀ but also the past front wheel steering angles δf ₁ , δf ₂ , δf ₃ and δf ₄ are given as input information from the memory 30 to the neural network 32. Another advantage is that the correspondence between the front wheel steering angle Δf and other input information is accurate.

[0038]

I have described above, according to the present invention urchin, wheels according to the steering control apparatus, the target rear wheel calculated from the motion information of the front-wheel steering <br/> information and vehicle at the rear wheel steering amount calculating means after the invention Since the steering amount is adjusted by the output of the learned neural network, that is, the control gain, based on the teacher gain selected by the sensory evaluation driver, the new target rear wheel steering amount after the adjustment is set by the user driver. Will be determined in consideration of As a result, then In the situation where wheel steering gives a sense of discomfort to the user <br/> driver suppresses the rear wheel steering, conversely, in a situation with no discomfort, then effective by wheel steering to maximize This enables steering control of the rear wheels as much as possible. Furthermore, input of neural network
The information includes front wheel steering information and vehicle motion information.
Car weight is an important parameter for grasping exercise information
Since the approximate value of the center slip angle is also added,
Network learning efficiency.

[Brief description of the drawings]

FIG. 1 is a schematic diagram showing a 4WS vehicle.

FIG. 2 is a block diagram showing a part of a control unit of FIG. 1;

FIG. 3 is a diagram for explaining a vehicle body weight center slip angle;

FIG. 4 is a diagram showing a linear two-degree-of-freedom vehicle model.

FIG. 5 is a graph showing a step-like steering input.

FIG. 6 is a schematic diagram showing a configuration of the neural network of FIG. 2;

FIG. 7 is a block diagram showing a learning system of the neural network of FIG. 2;

FIG. 8 is a flowchart showing a rear wheel steering control routine executed in the block diagram of FIG. 2;

[Explanation of symbols]

 6 front wheel steering system 8 rear wheel steering actuator 10 rear wheel steering system 13 feedback controller 14 control unit 16 handle angle sensor 18 vehicle speed sensor 20 longitudinal acceleration sensor 22 lateral acceleration sensor 24 yaw angular velocity sensor 25 4WS controller (rear wheel steering amount calculation means) 26 approximate expression calculation block 27 correction block (decision means) 28 pre-processing block 30 memory 32 neural network

──────────────────────────────────────────────────続 き Continuation of the front page (51) Int.Cl. ⁷ identification code FI B62D 113: 00 B62D 113: 00 137: 00 137: 00 (58) Field surveyed (Int.Cl. ⁷ , DB name) B62D 6 / 00 B62D 7/14

Claims (3)

(57) [Claims] When a front wheel steering system that operates in response to a steering wheel operation and a target rear wheel steering amount are given, the rear wheel steering is controlled via a rear wheel steering actuator according to the target rear wheel steering amount. In a front and rear wheel steering type vehicle provided with a wheel steering system, first detection means for detecting a steering state of a front wheel, second detection means for detecting various motion states of the vehicle, and the first and second detections Means for calculating the target rear wheel steering amount based on the front wheel steering information and the motion information detected by the means, and the target rear wheel steering amount calculated by the rear wheel steering amount calculating means. Gain calculation means for calculating the control gain, based on the target rear wheel steering amount from the rear wheel steering amount calculation means and the control gain from the gain calculation means,
Determining means for determining a value obtained by adjusting the gain of the target rear wheel steering amount as a new target rear wheel steering amount, and providing the determined target rear wheel steering amount to the rear wheel steering system; the front wheel steering rudder information及detected by the first and second detecting means
An input fine exercise information, sensory evaluation are learned teachers gain which is adjusted as teacher data by the driver, when the front-wheel steering information and the motion information is input, the control gain corresponding to the teacher gain while it has a trained neural network outputting front wheel steering information detected by the first and second detecting means
And rear wheel steering information input to the neural network
Before steering, the front wheel steering information and motion information are processed.
Pre-processing means, wherein the pre-processing means includes the front wheel steering information and the motion information.
Calculates the approximate value of the slip angle of the vehicle center of gravity from
As one input to the neural network
Wheel steering control device after that, characterized in that it have a approximate value calculating means. 2. The approximate value calculating means, which regards the vehicle as a linear mathematical model, is derived from the equation of motion of the mathematical model, and calculates the approximate value only from the vehicle speed, the front wheel steering angle, and the yaw angular velocity of the vehicle. The rear wheel steering control device according to claim 1 , further comprising an approximation formula that enables the rear wheel steering control. 3. The vehicle control apparatus according to claim 2, wherein the first detecting means includes a steering wheel angle sensor for detecting a steering angle of the vehicle, and the second detecting means includes a vehicle speed sensor for detecting a vehicle speed, and a yaw angular velocity for detecting the yaw angular velocity. 3. The rear wheel steering control device according to claim 2 , further comprising a sensor, a longitudinal acceleration sensor for detecting a longitudinal acceleration of the vehicle, and a lateral acceleration sensor for detecting a lateral acceleration of the vehicle.