JPH054544A - Automatic traveling device - Google Patents

Abstract

(57) [Summary] [Objective] To provide an automatic traveling device by fuzzy inference incorporating the same driving control method as a skilled driver, and an automatic traveling device capable of learning a driving control method suitable for each user's driving. To do. [Constitution] The fuzzy inference operation unit 3 makes a fuzzy inference based on the input / output data of a skilled driver, and the descent method operation unit 5 repeats the operation until the inference error of the error operation unit 7 becomes smaller than a predetermined value. After that, the parameters of the membership function storage unit 2 are fixed. Further, the user input unit for inputting the human preference for the output of the fuzzy inference calculation unit 3 restarts the calculation by the descent method and changes the parameters of the membership function.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention is to learn an automatic traveling device incorporating the same driving control method as a skilled driver by fuzzy reasoning for driving a vehicle and a driving control method suitable for each user's driving. The present invention relates to an automatic traveling device that can perform.

[0002]

2. Description of the Related Art Fuzzy inference is an attempt to execute knowledge obtained by humans from conventional experience on a computer using inference rules in a complicated controlled object in which a mathematical model cannot be described.

In the conventional fuzzy inference, as shown in FIG. 11, the input information obtained from the control observation value input unit 101, for example, the control deviation e and its change rate Δe, and the control operation amount output unit 103 output. If ~ then the relationship between the manipulated variables u
When describing as a rule, a plurality of the following inference rules are prepared in the fuzzy inference rule storage unit 104.

IF e is Zero (ZO) and Δe is Positive Small (PS) Then u is Negative Small (NS). Here, the if part is called an antecedent part, and the part of then is called a consequent part. Zero, PositiveSmall, Negative Small, etc. are labels representing membership functions of inputs and outputs used to describe inference rules. The membership function is stored in the membership function storage unit 105.

FIG. 12 shows an example of the membership function. The membership function is a symmetric triangle.

Commonly used membership functions include NB (negatively large), NS (negatively small), ZO (approximately zero), PS (positively small), PB (positively large), and the like.

Next, the fuzzy inference process performed by the fuzzy inference operation unit 102 will be described. Now, it is assumed that the following n inference rules are stored in the fuzzy inference rule storage unit 104. However, R ⁱ (i = 1,2, ... n) is an inference rule.

Here, the inference rule for the input information e, Δe
The first rule R is used to calculate the suitability μ _i of the antecedent part of R ^i.
Description will be given by taking ¹ as an example. Where μ _zo (e) and μ _ps (△ e) are the input information for the membership functions ZO and PM of the antecedent proposition.
Indicates the membership value of e, △ e. Now eo from the control observation value input portion 101, △ when eo have been inputted, rules fit mu ₁ for R _{^1, μ 1 = μ zo (eo} ) Λμ ps (△ eo) (1) However, lambda is min It will be an operation.

The membership function ω ¹ of the conclusion of the consequent part of the inference rule R ¹ is the membership function NS of the consequent proposition.
Using the membership value μ _ns (u) of, it is calculated as follows. ω1 = μ ₁ Λμ _ns (u) (2) Since there are multiple inference rules R ⁱ , the membership function that combines the membership functions of all conclusions is u _T = ω1Vω2Vω3V ・ ・ ・ Vωn (3) where V indicates max operation. Becomes The membership function u _T is a conclusion membership function indicating the control operation amount, but the actual control operation amount u o is a real number, so it is necessary to convert the membership function u _T to a real value. The weighted center of gravity shown below is adopted as the conversion method. The control operation amount uo is

[0010]

[Equation 1]

And is output to the control operation amount output unit 103.

[0012]

However, in the above configuration, it is difficult to optimally construct the inference rules and membership functions for the following reasons. In particular, like driving a car, the distance to the following vehicle and its degree of change, the distance to the preceding vehicle and its degree of change, the traveling speed of the vehicle,
And its degree of change, all or any of these, and complex control with many input and output parameters, such as controlling accelerator opening, braking force, and clutch operation amount, can no longer be constructed. It is impossible. This is because the skilled driver cannot express what the inference rules and membership functions are, and therefore the rules that the skilled driver is controlling well cannot be expressed. There was a problem that the device could not be realized.

Further, in the conventional structure, since the inference rule and the membership function are fixed, there is a problem that the driving cannot be changed to the user's preference.

In view of the problems of the conventional automatic traveling apparatus, the present invention automatically tunes (adjusts) fuzzy inference using the descent method from input / output data obtained from a skilled driver or a user. is there. Thus, it is an object of the present invention to provide an automatic traveling device which can automatically create a desired fuzzy inference rule without trial and error and can incorporate the driving method of a skilled driver or each user.

[0015]

A first aspect of the present invention is directed to a distance to a following vehicle and a degree of change thereof, a distance to a preceding vehicle and a degree of change thereof, a running speed of an own vehicle, and a degree of change thereof. Fuzzy inference with all or any of the above as input, and fuzzy inference calculation unit that outputs accelerator opening, braking force and clutch operation amount, and inference rule memory that stores inference rules used for fuzzy inference Part, a membership function storage part that stores the shape data of the membership function of the antecedent part used in the inference rule and the function of the consequent part, and the input / output data and the fuzzy data obtained by an experiment on the manufacturer side. The descent method operation part that performs the descent method from the inference result obtained from the inference operation part and the output of the descent method operation part reduces the membership function of the antecedent part and the function of the consequent part A membership function adjusting unit that changes one of the two, an inference error is calculated from the inference data obtained from the input / output data and the fuzzy inference operation unit, and when the inference error is smaller than a predetermined value, the descent method operation unit and the membership function The automatic traveling device includes an error calculation unit that stops the operation of the adjustment unit.

The second aspect of the present invention is characterized in that the consequent part used for the inference rule in the inference rule storage part is a real value.

According to the third aspect of the present invention, the fuzzy inference operation section has a user input section for inputting a human preference for the output of the fuzzy inference operation section, and when the user input section receives an input, it is obtained from a preset fuzzy inference operation section. The membership function is changed by performing a calculation by the descent method that minimizes the error between the inference result and the output of the user input unit.

In the fourth aspect of the present invention, when the user drives the vehicle, the inference result obtained from the preset fuzzy inference operation section and the output data controlled by the user are taken in by the output detection sensor and the inference result is obtained. The calculation is performed by the descent method that minimizes the error.

According to a fifth aspect of the present invention, in the third and fourth aspects of the present invention, the inference error calculation unit calculates an inference error from the inference result obtained from the fuzzy inference operation unit, and when the inference error is larger than a predetermined value. The number of occurrences is counted, and only when the count number becomes larger than a predetermined number, the descent method calculation unit and the membership function adjustment unit are operated until the inference error becomes smaller than a predetermined value. .

In the sixth aspect of the present invention, the inference rule storage unit and the membership function storage unit are searched, and the inference rule display unit displays the inference rules.

[0021]

With the above configuration, the distance to the following vehicle and the degree of change thereof, the distance to the preceding vehicle and the degree of change thereof, the traveling speed of the own vehicle, and the degree of change thereof, and all or any of these The fuzzy reasoning inference rules and membership functions are automatically generated based on the output data when a skilled driver successfully controls the accelerator opening, braking force, and clutch operation for input data in various situations. . After that, fuzzy inference can be performed by the automatically generated inference rule and the membership function, and a driving method similar to that of a skilled driver can be applied to the automatic traveling device.

Further, by changing the inference rules and the membership function again from the driving method of the skilled driver at the time of shipment from the manufacturer to further match the input / output data from the user, the automatic driving for learning the driving method of the user. The device becomes possible.

[0023]

Embodiments of the present invention will be described below with reference to the drawings.

An embodiment of the first present invention will be described. Collect the following input / output data obtained by experiments of skilled drivers on the manufacturer side. Distance to the following vehicle and its degree of change, distance to the vehicle in front and its degree of change, the traveling speed of the own vehicle and its degree of change, all of these, or any of various input data. It is output data of an accelerator opening, a braking force, and a clutch operation when a skilled driver successfully controls the vehicle. Based on these input / output data, the driving rules are automatically generated using the descent method, which is a kind of nonlinear search method. After that, based on the generated rule, fuzzy inference such as the braking force is performed on the input data from the sensor and the like, and the automatic driving is performed.

As a sensor, the signals of the front vehicle and the following vehicle input from the CCD camera are subjected to digital signal processing and automatically tracked to obtain their distance and the degree of change thereof in real time. This image tracking technique uses the well-known pattern matching method or motion detection vector method. Alternatively, an ultrasonic sensor may be used. The running speed of the vehicle and the degree of change can be easily obtained from the speedometer.

The operation of the automatic traveling apparatus of the first invention is shown in FIG.
This will be described with reference to the block diagram of FIG.

In FIG. 1, reference numeral 1 denotes an inference rule storage unit that stores inference rules for fuzzy inference, and 2 stores shape data of a membership function of an antecedent part used for fuzzy inference and a functional expression of a consequent part. Membership function storage,
3 is a fuzzy inference operation unit that performs inference operations for fuzzy inference, 4 is a control target, and 5 is a descent method that performs an operation by a descent method from input / output data obtained from a skilled driver and the inference result to obtain a tuning direction. A calculation unit 6 is a membership function storage unit 2 based on the calculation result of the descent method calculation unit 5.
A membership function adjusting unit for updating the parameters stored in the, and an error calculating unit 7 for calculating an inference error from the inference result of fuzzy inference and input / output data.

To simplify the description, a two-input one-output control system will be described as an example. This algorithm can be similarly applied to N inputs and M outputs.

The inference rule storage unit 1 stores the following inference rules. R ¹ : IF x ₁ = A ₁₁ & x ₂ = A ₁₂ THEN y = f ₁ (x ₁ , x ₂ ) R ² : IF x ₁ = A ₂₁ & x ₂ = A ₂₂ THEN y = f ₂ (x ₁ , x ₂ ) .............................. R ⁿ : IF x ₁ = A _n1 & x ₂ = A _n2 THEN y = f _n (x ₁ , x ₂ ) R ⁱ is the inference rule number, n is the inference rule number, A _ij (i = 1,. ,, nj = 1,
2) is the membership function of the antecedent part, and f _i (x ₁ , x ₂ ) is the linear function of the antecedent part.

The membership function A _ij is an isosceles triangle type as shown in FIG. 3, and its center value is a _ij and its width is b _ij .
The linear function of the consequent part is f _i (x ₁ , x ₂ ) = p _i · x ₁ + q _i · x ₂ + r _i (5) (i = 1, ..., n).

The membership function memory 2 stores the parameters to be tuned, a _ij , b _ij , p _i ,
q _i , r _i (these are called tuning parameters)
Store in the order of inference rules. [Step a1] First, the membership function adjusting unit 6 initializes the membership function used for fuzzy inference and the input / output data number p. For example, the initial value of the central value a _ij of the membership function A _{ij in} the antecedent part is
Set to divide the entire set of input variables into equal parts. Width b
_ij is set to be larger than the interval between the center values of the membership functions, and the membership functions are set to overlap. The coefficients p _i , q _i , r _i of the linear function of the consequent part are initialized to 0. The input / output data number is initialized to 1. "Step a2" Input / output data (x
_1p , x _2p , y _p ^r ). The input data (x _1p , x _2p ) is input to the fuzzy inference operation unit 3. The output data y _p ^r is input to the descent method calculation unit 5 and the error calculation unit 7. [Step a3] The fuzzy inference operation unit 3 performs fuzzy inference with (x _1p , x _2p ) as an input. In this inference, the operation represented by the following equation is performed to obtain the manipulated variable y _p ^* for the controlled object 4. μ _i = A _i1 (x _1p ) ・ A _i2 (x _2p ) (6)

[0032]

[Equation 2]

[Step a4] In the descent method computing unit 5, the tuning parameters a _ij , b _ij , p _i , are calculated from the inference result y _p ^* obtained in step a3 and y _p ^r input in step a2.
Calculate the tuning direction of q _i and r _i . Here, as a goal of tuning the membership function, the evaluation function of the following equation is minimized.

[0034]

[Equation 3]

This equation represents the difference between the inference result y _p ^* and the data y _p ^r obtained from the skilled driver, that is, the inference error. The descent method calculation unit 5 uses the steepest descent method which is one of the descent methods. In the steepest descent method, the differential value of the evaluation function is used to find the direction of updating the tuning parameter for minimizing the evaluation function.

Here, a method for obtaining a direction for minimizing the evaluation function E regarding the tuning parameter r _i will be described.

In FIG. 4, the horizontal axis is r._iShows the evaluation function E as
It was done. r_i= r_iThe differential value for'∂E (r_i') / ∂r_i
Is r as shown in FIG._i'The slope of the evaluation function at
means. Figure 4 (a) shows ∂E (r_i') / ∂r_iIs positive, Fig. 4 (b)
Is ∂E (r_i') / ∂r_iIs a negative time. Here, in FIG.
Tuning parameter r_i∂E (r_i') /
∂r_iWhen a small amount is moved in the direction opposite to the sign of, the evaluation function E
Decreases. Similarly, ∂E (r_i') / ∂r_iIs negative
Even when the tuning parameter is ∂E (r_i') / ∂r _iof
When a small amount is adjusted in the direction opposite to the sign, the evaluation function E decreases
To do. That is, the tuning parameter r_iAnd the differential amount ∂
E / ∂r_iAdjusting in the direction opposite to the sign of causes the evaluation function E to decrease.
By repeating this a little, the evaluation function E becomes minimal.
It will converge to the value.

Here, ∂E / ∂r _ij is calculated from the equations (7) and (8).

[0039]

[Equation 4]

[Mathematical formula-see original document] Then, a numerical operation is performed by the equation (9) to obtain the value of ∂E / ∂r _ij .

Similarly, in the descent method calculation unit 5, ∂E / ∂
a _ij , ∂E / ∂b _ij , ∂E / ∂p _ij , ∂E / ∂q _ij , ∂E / ∂r _ij are calculated, and the adjustment direction to reduce the evaluation function is calculated. “Step a5” In the membership function adjusting unit 6, ∂E / ∂a _ij , ∂E / ∂b _ij , ∂E / ∂ calculated by the descent method calculating unit 5
Tuning parameters stored in the membership function storage unit 2 using p _i , ∂E / ∂q _i , and ∂E / ∂r _i
Update a _ij , b _ij , p _i , q _i , r _i . Update is performed by the following formula. Where K _a , K _b , K _p , k _q , and K _r are constants.

[0042]

[Equation 5]

"Step a6" The membership function adjusting unit 6 continues the updating of step a5 as follows. Compare the input / output data number p with the total number N of input / output data. If the input / output data number p is smaller than the total number N of input / output data, proceed to step a7 and increase the value of p by 1 and return to step a2. , Steps a2 to a7 are repeated until the data number p becomes equal to the total number N of input / output data. If the input / output data number p is larger than the total number N of input / output data, the update is stopped and the process proceeds to step a8. “Step a8” The inference error D and its variation ΔD are obtained by the error calculation unit 7 from the equations (15) and (16).

[0044]

[Equation 6]

Here, t represents the number of times of tuning, and ΔD is the difference between the inference error at the time of tuning one time before and the current inference error. [Step a9] The error calculator 7 further compares the variation amount ΔD of the inference error with a predetermined threshold value T. If the amount of change ΔD is larger than the predetermined threshold value T, the process proceeds to step a10 to initialize the input / output data number p to 0, and step a2
To step a8 are repeated. If the amount of change ΔD is smaller than the predetermined threshold value T, it is determined that the tuning has converged, the error calculation unit 7 stops the operations of the descent method calculation unit 5 and the membership function adjustment unit 6, and ends the tuning.

In this way, at the end of tuning, the tuning parameters a _ij , b _ij , p _i corresponding to the inference rules that have acquired the knowledge of the skilled driver are stored in the inference rule storage unit 1 and the membership function storage unit 2. , Q _i , r _i have been set.

At this point, if shipped, when the user instructs automatic driving, the fuzzy inference operation section 3 makes a result inferred by a fuzzy inference calculation section 3 in the same manner as a skilled driver. Can drive

As described above, according to this embodiment, it is possible to obtain the optimum inference rule by the descent method from the input / output data obtained from the skilled driver. Therefore,
By using the acquired inference rules, it is possible to realize an automatic traveling device in which the knowledge and know-how of a skilled driver are easily installed in a device.

In this embodiment, the membership function is an isosceles triangle type, but the same effect can be obtained even if the membership function is another type. The consequent part may be a non-linear function or a membership function instead of a linear function.

Further, in the present embodiment, the steepest descent method is used as the descent method of the descent method computing unit 5, but the Newton method, the conjugate gradient method, the Porell method or the like may be used. Although the error calculation unit 7 determines the end of tuning based on the value of the inference error, a method of giving the number of times of tuning in advance before starting the tuning may be used.

Next, an embodiment of the second invention will be described. The present invention is to perform tuning with the consequent part of the membership function storage unit 2 described in the first present invention as a real value.

Since the consequent part of the inference rule is a real value, the number of tuning parameters is small, and faster automatic tuning and fuzzy inference are possible as compared with the first embodiment.

FIG. 5 is a block diagram of the fuzzy inference apparatus according to the second aspect of the present invention. In FIG. 5, reference numeral 21 denotes an inference rule storage unit that stores inference rules for fuzzy inference, 2
2 is a membership function storage unit that stores the shape data of the membership function of the antecedent part used for fuzzy inference and the functional expression of the consequent part. 3 is a fuzzy inference operation unit that performs inference operations for fuzzy inference. 4 is a controlled object, 23 is a descent method calculation unit that calculates the direction of tuning by performing a descent method calculation from input / output data obtained from an expert and inference results, and 24 is a membership function based on the calculation result of the descent method calculation unit 23. A membership function adjustment unit for updating the parameters stored in the storage unit 22, and an error calculation unit 7 for calculating an inference error from the inference result of fuzzy inference and input / output data. The same components as those in FIG. 1 are designated by the same reference numerals, and detailed description thereof will be omitted.

Differences from the first aspect of the present invention will be described.

In the inference rule storage unit 21, the consequent part is a real number. In the inference rule storage unit 1 of FIG. 1, the consequent part is used in the linear form represented by the equation (5), but in the present invention, the real value w _i (i = 1 ,.
n). n is the number of inference rules. Therefore, in the fuzzy inference operation unit 3, f _{i in the} equation (7) becomes a real value w _i , which simplifies the operation and speeds up the inference time.
The tuning parameters stored in the membership function storage unit 22 are a _ij , b _ij , and w _i , which are compared with a _ij , b _ij , p _i , q _i , and r _{i in the} membership function storage unit 2 in FIG. ,Decrease. In the descent method computing unit 23, in order to obtain the updating direction of the parameters, ∂E / ∂a _ij , ∂E / ∂b _ij , ∂E / ∂p _i , ∂E /
Find ∂w _i . ∂E / ∂a _ij , ∂E of descent method calculation unit 5
Compared with / ∂b _ij , ∂E / ∂p _i , ∂E / ∂q _i , and ∂E / ∂r _i , the number of calculations is reduced. In the membership function adjusting unit 24, the parameter updating work is performed by the membership function adjusting unit 6
Compared with, the equations (10) and (11) are the same, but (12),
Equations (13) and (14) are simplified by the following one equation. Where K _w is a constant.

[0056]

[Equation 7]

The algorithm for minimizing the inference error by the descent method computing unit 23 is performed by the same method as described in FIG. 2, and the tuning is completed. At the end of tuning, the inference rule storage unit 21 and the membership function storage unit 2
In No. 2, an inference rule incorporating the knowledge of the skilled driver will be constructed.

As described above, according to this embodiment, it is possible to obtain the optimum inference rule by the descent method from the input / output data obtained from the skilled driver. In particular, since the consequent part of the inference rule is a real value, the number of tuning parameters is small, and higher-speed automatic tuning is possible as compared with the first embodiment of the present invention. Further, by using the present invention, since the consequent part is a real number, even after tuning, fuzzy inference can be realized by a simple program or by hardware, and a driving method similar to that of a skilled driver can be easily mounted on the device. it can.

An embodiment of the third invention will be described.

In the automatic traveling device based on the inference rules that are tuned based on the input / output data obtained by the experiment on the manufacturer side, the rules are for the skilled driver or the driver with the greatest common divisor.

On the other hand, the driving method differs slightly depending on the driving skill of each user, age, and gender.
For example, on a highway, when the vehicle travels at the maximum speed limit, some people feel uncomfortable, or on the contrary, travel at the maximum speed limit, and some people are satisfied with the sense of speed. In the third aspect of the present invention, the user's preference is input from the driving method set like a skilled driver, and the fuzzy inference rules are sequentially tuned to realize control suitable for the user's preference and sensitivity. As a result, the automatic traveling device is equipped with the adaptive fuzzy control that learns the user's preference in real time.

FIG. 6 is a block diagram of the fuzzy inference apparatus according to the third aspect of the present invention. In FIG. 6, 1 is an inference rule storage unit that stores inference rules for fuzzy inference, and 2 is a member that stores the shape data of the membership function of the antecedent part and the functional expression of the consequent part used for fuzzy inference. A ship function storage unit, 3 is a fuzzy inference operation unit that performs fuzzy inference operations, 5 is a descent method operation unit that obtains a tuning direction by descent method operation from input / output data obtained from an expert and inference results, and 6 is a descent method. Membership function adjusting unit for updating the membership function based on the operation result of the operation unit 5, 7
Is an error calculator that calculates inference errors from fuzzy inference results and input / output data. The same components as those in FIG. 1 are designated by the same reference numerals, and detailed description thereof will be omitted.

Differences from the first aspect of the present invention will be described. The difference from the configuration of FIG. 1 is that the inference result display section 31 that displays the inference result calculated by the fuzzy inference operation 3 is displayed.
And a user input unit 32 for inputting the user's preference for the inference result.

The operation of this embodiment will be described with reference to the flowchart of FIG. "Step b1" As the fuzzy inference rules stored in the inference rule storage unit 1 and the membership function storage unit 2, the inference rules of the skilled driver are set before shipment. The construction of this initial inference rule may use the first aspect of the present invention, or an experiment by trial and error or a method by an interview with a skilled driver. [Step b2] Next, when the user operates in the automatic driving mode, the input data such as the distance to the preceding vehicle and the following vehicle is fetched from the sensor including the CCD described above. [Step b3] Based on these input data, the fuzzy inference operation unit 3 performs fuzzy inference, and the inference result y ^*
To get The operation procedure of fuzzy inference is the same as step a3 of the first embodiment. "Step b4" The inference result y ^* is displayed to the user on the inference result display unit 31. The display method may be a display, but for safety, voice output by voice synthesis is preferable. Inference result
Based on y ^* , "Brake force is 3, accelerator opening is 8"
The man-machine interface can be further improved by using a quantitative expression such as, or converting the inference result y ^* into a fuzzy number and outputting it as "medium braking force, a little step on the accelerator". "Step b5" Next, the user inputs the correction amount from the user input unit 32 according to his / her preference. For example, when the user is in that input situation, it is better to drive a little slower to suit his or her driving feeling, and if that is desirable, a quantitative expression such as "brake force is 5 and accelerator opening is 5" Or, enter an instruction with a correction using a fuzzy number such as "the braking force is small and the accelerator is quite large". This may also be input with a touch panel or keyboard, but for safety it is preferable to input with voice recognition. If the user makes no changes to the displayed inference results,
The correction amount is 0. "Step b6" The error calculation unit 7 checks whether the correction amount y ^', which is the output from the user input unit 32, is zero. In the membership function adjusting unit 6, if the correction amount y ′ is 0,
I think that the fuzzy inference rules that led to the inference result represent the user's preference or that it is better to use the method of the expert driver, and proceed to step b2 without tuning the inference rules, and the next different situation. Wait for sensor input at. If the correction amount calculated by the error calculation unit 7 is not 0, the membership function adjustment unit 6 performs step b7.
Go ahead.

“Steps b7, b8” Inference result y ^* is modified by y ^′
And make y ^r . This y ^r is the output data and the input data
The membership function is tuned by the descent method with x ₁ and x ₂ as one input / output data. The calculation of this tuning is performed in steps a4 and a of the first embodiment of the present invention.
Same as 5. After updating the membership function, the process returns to step b2 and waits for the next sensor input.

In this way, when the user is driving by automatic driving, in various situations, when the driving method of the manufacturer's experienced driver is sensuously different, tuning is started and parameter adjustment is carried out to drive the driver in his or her own way. The method can be changed later. The user inputs whether or not to adjust the parameters.

As described above, according to the present embodiment, the result of the fuzzy inference is displayed to the user by the inference result display section 31 and the user's preference for the fuzzy reasoning is input from the user input section 32. Use to change the membership function of fuzzy reasoning. As a result, the more the device is used, the more it is possible to realize an automatic traveling device that can be controlled according to the user's preference.

In this embodiment, the structure of the portion for automatic tuning is the same as that of the first embodiment of the present invention, but the structure of the second embodiment of the present invention may be used for this portion. In this case, the consequent part of the inference rule becomes a real value, and the merits of the second aspect of the present invention such as high-speed tuning are also realized.

Next, a fourth embodiment of the present invention will be described.

In the third aspect of the present invention, the user has instructed the preference, but in the present invention, while the user is using,
Automatically learns user preferences without giving instructions,
The fuzzy inference rules are tuned sequentially to realize the control that suits the user's taste and sensibility. As a result, the automatic traveling device is equipped with the adaptive fuzzy control that learns the user's preference in real time.

FIG. 8 shows a block diagram of an apparatus according to the fourth aspect of the present invention. 8, the same components as those in FIG. 6 are designated by the same reference numerals, and detailed description thereof will be omitted.

Differences from the third invention will be described. The difference from the configuration of FIG. 6 is that the inference result display unit 31 that displays the inference result calculated by the fuzzy inference operation 3 is displayed.
Instead of the user input unit 32 for inputting the user's preference for the inference result, an output detection sensor 41 for detecting the user's braking force, accelerator opening, and clutch operation amount.
That is the point.

The operation of the fuzzy inference apparatus of the embodiment constructed as above will be described below.

The operation of this embodiment is the step of the third invention.
Steps b4 and b5 are changed to step c5, and step b6 is changed to step C6. Others are the same, so detailed description will be omitted. [Step c5] Under the situation at the time when the inference result y ^* is obtained, the user may change the braking force, the accelerator opening, or the clutch operation according to his / her preference. The operation amount is taken in from the output detection sensor 41. Then calculates the correction amount y ^'is the difference between the operation amount and the respective inference results y ^*. If the user does not make any changes, the modification amount is zero. "Step c6" correction amount y ^'is ascertain whether 0. If the correction amount y'is 0, the process proceeds to step b2 without tuning the inference rule, and waits for the input of the output detection sensor 41 in the next different situation. If the correction amount is not 0, proceed to step b7.

As described above, when the user corrects and operates the automatic driving method in various situations, the output detection sensor 41 takes in the driving data of the user, starts tuning, and adjusts the parameter to allow the user to operate. After that, the driving method can be changed.

As described above, according to this embodiment, the user's driving method is incorporated by the output detection sensor 41,
The membership function of fuzzy reasoning is changed by using the descent method so as to minimize the difference from the result of fuzzy reasoning similar to that of a skilled driver. As a result, the more the device is used, the more the user can realize the automatic traveling device which can be controlled according to the user's preference without knowing it.

In the embodiment, the construction of the portion for automatic tuning is the same as that of the first embodiment, but the construction of the second embodiment may be used for this portion. In this case, the consequent part of the inference rule becomes a real value, and the merits of the second aspect of the present invention such as high-speed tuning are also realized.

Next, a fifth embodiment of the present invention will be described.

According to the third and fourth aspects of the present invention, the user can automatically instruct the user's preference without instructing the preference while the user is using it. The fuzzy inference rules are tuned to realize the control that suits the user's preference and sensitivity. In this case, if the user has a different driving method with respect to the inference rule like a skilled driver set by the manufacturer at the time of shipping, tuning is started regardless of the difference. Therefore, if there is a big difference in some situation, start tuning to match it. In this case, it can be said that it is dangerous if the difference is large, but in the present invention, the tuning of the membership function is not started immediately when the difference is larger than a predetermined value.

FIG. 9 shows a block diagram of an apparatus of the fifth invention. 9, the same components as those in FIG. 6 are designated by the same reference numerals, and detailed description thereof will be omitted.

Differences from the third invention will be described. The difference from the configuration of FIG. 6 is that the error level determination unit 51 determines the magnitude of the error from the output of the error calculation unit 7, and when it is larger than a predetermined value, the occurrence frequency is counted by the counter 52. In the count number comparison unit 53, when the count number exceeds a certain number of times, the descent method calculation unit 5 is driven,
The parameters of the membership function adjusting unit 6 are modified.
Since this tuning operation is similar to that of the third aspect of the present invention, detailed description will be omitted. If the output of the error level determination unit 51 shows a large error, the counter may be reset and the correction may not be performed.

Needless to say, the present invention can be applied to the fourth invention.

As described above, according to the present invention, when the driving instruction of the user is largely irrelevant, the same instruction is repeatedly received without correcting the parameters of the membership function or immediately. Since it is sometimes corrected for the first time, the automatic driving device can be safely equipped with a driving method according to the user's preference.

A sixth embodiment of the present invention will be described. Figure 10
FIG. 9 is a block diagram of a fuzzy reasoning device of a sixth aspect of the present invention. In the figure, 1 is an inference rule storage unit that stores inference rules for fuzzy inference, and 2 is a membership that stores shape data of membership functions in the antecedent part and functional expressions in the consequent part used for fuzzy inference. Function storage unit, 3 is a fuzzy inference operation unit that performs fuzzy inference operations, 4 is a control target, and 5 is a descent method operation unit that obtains a tuning direction by an operation by a descent method from input / output data and inference results obtained from an expert , 6 is a membership function adjustment unit that updates the membership function based on the calculation result of the descent method calculation unit 5, and 7 is an error calculation unit that calculates an inference error from the fuzzy inference result and input / output data. The above is the same as the configuration of FIG. The difference from the configuration of FIG. 1 is that the inference rule storage unit 1 and the membership function storage unit 2 are searched to obtain an inference rule with a large adaptation range, and an inference rule obtained by the inference rule search unit 61. The reason is that an inference rule display unit 62 for displaying is provided.

The operation of the fuzzy inference apparatus of this embodiment constructed as above will be described below.

In the first and second aspects of the present invention, the construction of the fuzzy reasoning membership function similar to that of a skilled driver is automatically performed. In the third to fifth aspects of the present invention, the membership function is set according to the user's preference. It could be tuned to the membership function. However, if the input / output data obtained from a skilled driver or user contains a large amount of noise, tuning may proceed in an unexpected direction. In addition, due to excessive tuning, inference with a small inference error can be performed on given input / output data, but the inference error may be extremely large for data not given during tuning. In order to avoid such a situation, it is necessary for designers and users to understand and check the automatically constructed inference rules and membership functions.

In the present invention, in consideration of these points, the inference rule display section 62 displays the inference rules obtained by the automatic tuning. Further, the inference rule searching unit 61 displays the inference rules having a wide matching range first in order to efficiently check the inference rules.

The operation procedure of the automatic tuning is the same as the flowchart of FIG. 2 of the first embodiment. The difference is that the inference rules are displayed after the tuning is completed. After the tuning is completed, the inference rule search unit 61 searches the membership function storage unit 2 and calculates the following equation.

[0089]

[Equation 8]

S _i indicates the width of the applicable range of the inference rule R _i .
The inference rule R _i , the shape of the membership function thereof, and the parameters of the functional expression of the consequent part are sent to the inference rule display unit 62 in the order of increasing S _i .

The inference rule display section 62 is composed of, for example, a CRT or a liquid crystal display, and displays information such as the inference rules sent from the inference rule retrieval section 61 and the membership function used therein.

As described above, according to the present embodiment, the fuzzy inference rules obtained by the descent method can be displayed in the descending order of matching range. Therefore, the user can know the inference rule obtained by the automatic tuning, and the progress of tuning and the inference rule can be checked.

Although the present invention has been described based on the first aspect of the present invention, it goes without saying that the second to fifth aspects of the present invention can also have the same display function.

[0094]

As described above, according to the first aspect of the present invention,
It is possible to obtain the optimum inference rule by the descent method from the input / output data obtained from the skilled driver. Therefore, the knowledge and know-how of the skilled driver can be easily installed in the automatic traveling device as an inference rule.

According to the second aspect of the present invention, it is possible to obtain the optimum inference rule by the descent method from the input / output data obtained from the skilled driver. In particular, since the consequent part of the inference rule is a real value, the number of tuning parameters is small, and higher-speed automatic tuning and inference are possible as compared with the first embodiment.

According to the third aspect of the present invention, the result of fuzzy inference is displayed to the user by the inference result display unit, and the user's preference for the result is input from the user input unit. Change the membership function. With this, the more you use the device,
It is possible to realize an automatic traveling device that can be controlled according to the user's preference.

According to the fourth aspect of the present invention, the driving method of the user is taken in by the output detection sensor, and the descent method is used to minimize the difference from the result of the fuzzy inference similar to that of a skilled driver. By changing the ship function, it is possible to realize an automatic traveling device that can be controlled according to the user's preference without knowing it as the device is used.

According to the fifth aspect of the present invention, when the driving instruction of the user is largely irrelevant, the same instruction is given many times without correcting the parameters of the membership function or immediately. Since the correction is performed for the first time when coming, it is possible to realize an automatic traveling device safely equipped with a driving method according to a user's preference.

According to the sixth aspect of the present invention, by displaying the fuzzy inference rules obtained by the descent method in descending order of matching range, the user can know the inference rules obtained by automatic tuning, and the tuning The progress status and inference rules can be checked.

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration of an apparatus according to an embodiment of the first present invention.

FIG. 2 is a flowchart showing the operation of the apparatus of the same embodiment.

FIG. 3 is a configuration diagram of a membership function.

FIG. 4 is an operation explanatory diagram of a descent method.

FIG. 5 is a block diagram showing a configuration of an apparatus according to an exemplary embodiment of the second present invention.

FIG. 6 is a block diagram showing the configuration of an apparatus according to an embodiment of the third invention.

FIG. 7 is a flowchart showing an operation of the apparatus of the same embodiment.

FIG. 8 is a block diagram showing a configuration of an apparatus of an embodiment in the fourth invention.

FIG. 9 is a diagram showing a configuration of an apparatus according to an embodiment of the present invention.

FIG. 10 is a diagram showing the configuration of an apparatus according to an embodiment of the sixth present invention.

FIG. 11 is a block diagram showing a configuration of a conventional fuzzy inference apparatus.

FIG. 12 is a configuration diagram of a membership function.

[Explanation of symbols]

1, 21 Inference rule storage 2.22 Membership function storage 3 Fuzzy inference operation section 4 controlled objects 5, 23 Descent method calculation unit 6, 24 Membership function adjustment department 7 Error calculator 31 Inference result display section 32 User input section 41 Output detection sensor 51 Error level determination unit 52 counter 53 Count Count Comparison Section 61 Inference rule search unit 62 Inference rule display section

Claims (6)

[Claims] 1. A distance to a following vehicle and a degree of change thereof, a distance to a vehicle ahead, and a degree of change thereof, a traveling speed of the own vehicle,
And the degree of change thereof, all or any of these, are used as inputs to perform fuzzy inference, and fuzzy inference calculation means that outputs accelerator opening, braking force, and clutch operation amount, and inference rules used for the fuzzy inference. The stored inference rule storage means, the membership function storage means that stores the shape data of the membership function of the antecedent part and the function of the consequent part used in the inference rule, and the input obtained by the experiment. A descent method computing means for computing a descent method from output data and an inference result obtained from the fuzzy inference computing means, and at least one of a membership function of the antecedent part and a function of the consequent part by the output of the descent method computing means. The membership function adjusting means for changing the value, and the inference result obtained from the input / output data and the fuzzy inference calculating means. Automatic traveling apparatus characterized by inference error by calculating the error is and a error calculating means for stopping the operation of the said descent calculation means membership function adjusting means when less than a predetermined value. 2. A distance to a following vehicle and a degree of change thereof, a distance to a vehicle ahead, and a degree of change thereof, a traveling speed of the own vehicle,
And the degree of change thereof, all or any of these, are used as inputs to perform fuzzy inference, and fuzzy inference calculation means that outputs accelerator opening, braking force, and clutch operation amount are output, and inference rules used for the fuzzy inference. The stored inference rule storage means, the antecedent part parameter storage means that stores the parameter representing the shape of the membership function of the antecedent part used for the inference rule, and the actual part of the consequent part used for the inference rule. The real part numerical value storage means for storing numerical values, the input / output data obtained by an experiment, and the inference result obtained from the fuzzy inference operation means, to the parameter representing the shape of the membership function of the antecedent part. Stored in the antecedent part parameter storage means by the output of the antecedent part descent method calculation means for performing a modal operation and the antecedent part descent method calculation means. The antecedent part parameter adjusting means for changing the parameter representing the shape of the membership function, and the descent method operation for the real value of the consequent part from the input / output data given in advance and the inference result obtained from the fuzzy inference operation means. Consequent part descent method computing means for performing and consequent part real value adjusting means for changing the real value of the consequent part stored in the consequent part real value storage means by the output of the consequent part descent method computing means An inference error is calculated from the input / output data and the inference result obtained from the fuzzy inference operation means, and when the inference error is smaller than a predetermined value, the antecedent part descent method operation means and the antecedent part parameter adjustment means An automatic traveling device comprising: a consequent part descent method computing means and an error computing means for stopping the operation of the consequent part real value adjusting means. 3. A distance to a following vehicle and a degree of change thereof, a distance to a vehicle ahead, and a degree of change thereof, a traveling speed of the own vehicle,
And the degree of change thereof, all or any of these, are used as inputs to perform fuzzy inference, and fuzzy inference calculation means that outputs accelerator opening, braking force, and clutch operation amount are output, and inference rules used for the fuzzy inference. Inference rule storage means that is stored, membership function storage means that stores the shape data of the membership function of the antecedent part and the function of the consequent part used in the inference rule, and inference by the fuzzy inference operation means. An inference result display section for displaying a result, a user input means for inputting human preference, an inference result obtained from the fuzzy inference operation means and an output of the user input means when an input is made to the user input means A descent method calculation means for performing a calculation by the descent method, and an output of the descent method calculation means stored in the membership function storage means. The membership function adjusting means for changing either one of the membership function of the antecedent part and the function of the consequent part, and the inference error is calculated from the inference result obtained from the input / output data and the fuzzy inference operation means. An automatic traveling device comprising: a descent method computing means and an error computing means for stopping the operations of the membership function adjusting means when an inference error is smaller than a predetermined value. 4. A distance to a following vehicle and a degree of change thereof, a distance to a vehicle ahead, and a degree of change thereof, a traveling speed of the own vehicle,
And the degree of change thereof, all or any of these, are used as inputs to perform fuzzy inference, and fuzzy inference calculation means that outputs accelerator opening, braking force, and clutch operation amount are output, and inference rules used for the fuzzy inference. Inference rule storage means that is stored, membership function storage means that stores the shape data of the membership function of the antecedent part used in the inference rule and the function of the consequent part, and when the user drives a car Descent method computing means for performing a descent method from the inference result obtained from the fuzzy inference computing means and the output data controlled by the user, and the membership function by the output of the descent method computing means. A member that changes either the membership function of the antecedent part or the function of the consequent part stored in the storage means. A ship function adjusting means, an inference error is calculated from the input / output data when the user drives and the inference result obtained from the fuzzy inference operation means, and when the inference error is smaller than a predetermined value, the descent method operation means and the An automatic running device comprising: an error calculating means for stopping the operation of the membership function adjusting means. 5. The inference error calculation means calculates an inference error from the inference result obtained from the fuzzy inference operation means, and when the inference error is larger than a predetermined value, the number of occurrences thereof is counted, and the count number is a predetermined number. The automatic running according to claim 3 or 4, wherein the descent method computing means and the membership function adjusting means are operated until the inference error becomes smaller than a predetermined value only when the number becomes larger than the number. apparatus. 6. A distance to a following vehicle and a degree of change thereof, a distance to a vehicle ahead, and a degree of change thereof, a traveling speed of the own vehicle,
And the degree of change thereof, all or any of these, are used as inputs to perform fuzzy inference, and fuzzy inference calculation means that outputs accelerator opening, braking force, and clutch operation amount are output, and inference rules used for the fuzzy inference. Inference rule storage means that is stored, membership function storage means that stores the shape data of the membership function of the antecedent part used in the inference rule and the function of the consequent part, the input / output data and the fuzzy A descent method computing means that performs a descent method from the inference result obtained from the inference computing means, and a membership that changes at least one of the membership function of the antecedent part and the function of the consequent part by the output of the descent method computing means. Inference is performed by calculating an inference error from the inference result obtained from the function adjusting means, the input / output data, and the fuzzy inference operation means. When the difference is smaller than a predetermined value, error calculation means for stopping the operations of the descent method calculation means and the membership function adjustment means, the inference rule storage means and the membership function storage means are searched to find the inference rule. And an inference rule display means for displaying the inference rule display means.