DE4242218A1 - Road vehicle dynamics control method - using actuator position and rate of error used in fuzzy logic control stage to regulate output - Google Patents

Abstract

The control of an actuator in a road vehicle, such as a pulse width modulated hydraulic device (18) used in rear axis control, is provided with the aid of a fuzzy logic unit (20). The position of the actuator is measured (12) and is fed back (y) for comparison (11) with a reference (w). The error signal (e) and its derivative (e') are provided as input to the fuzzy logic controller that provides a setting input to the control action device (16). USE/ADVANTAGE - Provides optimal control over wide range of conditions.

Description

State of the art

The invention relates to a method for controlling a driving Dynamic influencing or the driving condition of a vehicle according to the preamble of claim 1 and a device for Execution of the procedure.

DE 37 34 477 A1 describes a device for controlling the steering angle of the wheels of a vehicle known. There is steering wheel turning angle and vehicle speed using Senso be scanned on the basis of a predetermined mathematics model represents the movement of the vehicle train variable determined, which serves as the setpoint. For the actual A reference variable is also determined when the vehicle is moving. A controller compares the two reference variables with each other and controls an adjusting means for adjusting the steering angle of the wheels such that the deviation between the two reference variables is reduced.

From US 5 005 133 is a system for regulating the speed of a vehicle using fuzzy logic. The actual value the speed is determined and a setpoint is taken into account  specified driver request. Depending on the driving condition (e.g. Ge Suitable gear functions are selected. On the basis of the membership functions found in this way and the Setpoints and actual values of the vehicle speed are used as a manipulated variable Control of an actuator determined using fuzzy rules.

The invention is based on the object of driving dynamics or variable influencing the driving condition of a vehicle if possible optimally regulated. The difference between the target and actual should value as low as possible under as many operating conditions are maintained and the control behavior stable even in extreme cases stay.

This object is characterized by the in claim 1 characterized features and the advantageous described below Refinements and further developments solved.

Advantages of the invention

The invention has the advantage that it provides optimal control of a influencing the driving dynamics or the driving state of a vehicle Size allows. Since none for the control loop according to the invention mathematical model of the controlled system is required, the Modeling effort. Furthermore, the simple one affects Structure of the fuzzy controller advantageous because the adaptation to ei a specific area of application is made easier.

Another advantage is the robustness of the fuzzy controller also in Extreme situations. The use of the inventin is particularly advantageous according to the control loop if the actuator is arranged as a switching valve voltage is executed, for example to control a rear axle stellers. The switching valve arrangement has similar advantages as that Fuzzy controller on: It has a simple structure, is robust  and can be realized inexpensively. Overall, this leads to a lot good control characteristics.

drawing

The invention is based on the Darge in the drawing presented embodiments explained.

Show it

Fig. 1 is a schematic diagram of a known Re gelkreises, Fig. 2 is a schematic diagram of the Re gelkreises invention with fuzzy controller, Fig. 3 shows the schematic structure of the fuzzy controller, Fig. 4 the membership functions, Fig. 5 the Pha senebene the normalized control deviation e and the normalized time derivative e 'of the control deviation, Fig. 6 is a table for the assignment of areas of the phase level to categories of the standardized manipulated variable u, Fig. 7 is a listing of the fuzzy rules and Fig. 8 ei ne graphical Presentation of the application of the fuzzy rules.

The structure and operation of the invention are exemplary for use in connection with rear wheel steering ben. In this embodiment, the invention provides signals to control electromagnetic switching valves that the Hydraulic flow to at least one rear axle actuator for control of the steering angle. The following is an embodiment described with a rear axle actuator. In principle, however several rear axle actuators can also be controlled. Furthermore can the invention, for example, in connection with a Chassis control or used in connection with the braking system become.

Internally, the fuzzy controller works with standardized sizes. It is assumed that all sizes are normalized when reading into the fuzzy controller and this normalization is canceled when the fuzzy controller is output. In order to be able to differentiate between the normalized and the non-standard sizes more easily, the symbols of the non-standard sizes are underlined, e.g. B. e (non-standardized control deviation) or e (standardized control deviation).

Fig. 1 shows a basic circuit diagram of a per se conventional control loop. A block with the reference number 10 provides a setpoint value w for a connection point 11 . An actual value y is detected by a sensor 12 and serves as the 2nd input signal of the node 11 . At node 11 , a control deviation e is determined by forming the difference between the desired value w and the actual value y . The control deviation e is fed to a controller 14 , which determines a manipulated variable u therefrom and forwards it to an actuator signal generator 16 . The actuator signal generator 16 he witnesses a corresponding drive signal u ₁ and acted upon an actuator 18, whose position with the sensor 12 as an actual value y he construed, with the signal u ₁.

Fig. 2 shows a schematic diagram of the Regelkrei ses invention. The rough structure corresponds to the known control loop from FIG. 1. The internal structure of the individual blocks, in particular block 14, is essential to the invention. Block 14 consists of a fuzzy controller 20 and a means 22 for forming the time derivative e 'of the control deviation. The fuzzy controller receives the control deviation e as the input signals and the derivative e 'of the control deviation according to time and supplies the control variable u as the output signal, which is passed on to the control signal generator 16 . The Stellersignalge generator 16 generates a corresponding pulse width modulated signal u _PWM to control the actuator 18th

The actuator 18 consists of a rear axle actuator and a switching valve arrangement through which the rear axle actuator is acted upon by hydraulic fluid. The switching valves are opened and closed in time with the pulse-width-modulated signal, which leads to a correspondingly modulated hydraulic signal. Due to the controlled hydraulic flow, the piston of the rear axle adjuster can be moved, thereby influencing the wheel lock.

The actual position y of the rear axle adjuster is detected by sensor 12 and enters node 11 with a negative sign. With a positive sign, the target position w of the rear axle actuator enters the link point 11 . In node 11 , the two input signals y and w are additively superimposed. The result of the overlay is the control deviation e .

To determine the desired position w of the rear axle actuator by block 10 , methods known from the prior art can be used. In one of these methods, a suitable sensor senses the steering angle of the steering wheel. This turning angle is converted via a characteristic curve into a corresponding target position w of the rear axle actuator. In the implementation, other variables that characterize the driving state, such as. B. the greed speed, are taken into account.

FIG. 3 shows a schematic illustration of the fuzzy controller 20 , which internally consists of 3 functional units 30 , 32 and 34 connected in series . These 3 functional units are usually referred to as fuzzyfication ( 30 ), fuzzy reasoning ( 32 ) and defuzzyfication ( 34 ).

The 1st functional unit 30 receives the control deviation e and its time derivative e 'as input signals. The input signals are normalized, ie the normalized control deviation e and the normalized time derivative e 'of the control deviation are determined. The quantities e and e 'are assigned to two groups of membership functions µ and µ _e' , which are shown graphically in FIG. 4.

In each membership function, the current values dro e or e ′ are noted and then the membership functions are passed on to the second functional unit 32 .

In the second functional unit 32 , the fuzzy rules are applied to the membership functions provided by the first functional unit 30 and, as a result, a membership function µ _{u is obtained} for the manipulated variable. A graphical representation of this operation, which is usually referred to as fuzzy reasoning, is shown in FIG. 8 described below . The membership function μ _u is forwarded to the third function unit 34 , which generates a standardized manipulated variable from it by averaging. U is from the normalized control value A to the system, in which the fuzzy controller is used, adapted manipulated variable u ermit telt and providing at the output of the fuzzy controller 20th

In Fig. 4 membership functions are shown. To define the membership functions, the value ranges of the standardized control deviation e, the standardized time derivative e ′, the control deviation and the standardized manipulated variable u are each assigned several categories, with each category qualitatively describing the position of a sub-range within the value range. The strength of the assignment of each individual value of a value range to a category is determined by a membership function. For example, for the normalized control deviation e, the strength of the assignment to the "negative small" category is described by the membership function μ _e of FIG. 4, designated "NS".

Fig. 4 shows membership functions for the categories "negative big" (NB), "negative small" (NS), "positive small" (PS) and "positive big" (PB). The normalized quantities e, e ′ and u are plotted on the abscissa and the membership functions µ _e , µ _{e ′} and µ _u for the different categories on the ordinate. The membership functions each take values between 0 and 1. The value 0 indicates that there is no membership in the category under consideration, and the value 1 indicates complete membership. In the exemplary embodiment described here, the membership functions of the standardized control deviation e, the standardized time derivative e 'of the control deviation and the standardized manipulated variable u are identical for the same category. In principle, however, different membership functions can be selected for each of these three sizes. Depending on the application, other functional courses than those shown in FIG. 4 can also be selected. For the control of a rear axle actuator via a switching valve arrangement, however, the membership functions shown here have proven to be particularly expedient.

In Fig. 5 is a phase plane for the normalized control deviation e and the normalized time derivative E 'of the control deviation is Darge is that u is a particularly efficient and intuitive derivation of the fuzzy rules for determining the normalized control value of the normalized control deviation e and the normalized time Derivation e 'of the control deviation allows. The normalized control deviation e is plotted on the abscissa and the normalized time derivative e ′ of the control deviation is plotted on the ordinate. The phase level is divided into several areas, which are designated with the letters A to F. The straight line "e + e ′ = 0" runs diagonally through the phase plane from top left to bottom right. The fuzzy rules are generated with the help of the phase level by assigning a category of the standardized manipulated variable u to an area or a combination of several areas of the phase level.

In Fig. 6, the assignments between the areas of the phase level (left column) and the categories of the standardized manipulated variable u (middle column) are shown in a table. Fuzzy rules (right column) are obtained from these assignments by describing the areas of the phase level by logical combinations of the categories for the normalized control deviation e and the normalized time derivative e ′ of the control deviation.

Fig. 7 shows a collection of fuzzy rules (R1 to R4). A set of rules is defined for each of the two areas of the phase level "e '<- e" (R1 and R2) and "e'-e" (R3 and R4). Every fuzzy rule is introduced by the word "IF", which is followed by a premise. The premise is followed by the word "THEN" and a conclusion. The premise consists of a category specification or several category information, linked by logical operators, for the standardized rule deviation e and / or the standardized time derivative e ′ of the rule deviation. The size to which the respective category specification refers is between the two abbreviations for the category, e.g. B. NEB means: The normalized control deviation e is "negative big". The conclusion contains a category specification for the standardized manipulated variable u, e.g. B. "NUB".

In Fig. 8 the application of fuzzy rules for the example values e _{0 of} the normalized control deviation e and e ' _{0 of} the normalized time derivative e' of the control deviation is shown. It is first determined which area of the phase level ("e '<- e" or "e'-e") the example values e ₀ and e' _{0 are} to be assigned and accordingly the rules R1 and R2 or R3 and R4 selected. The following applies to the example values e ₀ and e ′ ₀ shown in FIG. 8: e ′ ₀ <-e ₀ . Consequently, the fuzzy rules that are defined for the range "e '<- e" of the phase plane are to be applied, ie R1 and R2.

In the upper part of FIG. 8, the membership functions µ _e , µ _{a '} and µ _u from Fig. 4 are shown. Below, the application of the fuzzy rules R1 and R2 is shown, the application of each individual rule being read from left to right, in other words, the results of the rule applications are each shown on the right-hand side of FIG. 8. The results of the control applications are superimposed from top to bottom and lead to the final result, which is shown at the bottom right in FIG. 8.

A fuzzy rule is used as follows:

First, the category information of the premise (e.g. R1: (PES AND (PE′S OR NE′S)) OR (NES OR NEB)) is evaluated by adding e ₀ or e ′ to each category specification at the location of the example value ₀ (vertical, dashed lines) the value of the membership function µ _e or µ _{e ′ is determined} for the size e or e ′ indicated by the category specification. For rule R1, the values determined in this way are shown in FIG. 8 as points and designated I, II, III, IV and V, in accordance with the individual category specifications of the premise of rule R1: PES, PE′S, NE′S, NES and NEB. The premise of rule R1 can therefore also be represented as: (I AND (II OR III)) OR (IV OR V). If the premise consists of several categories - in rule R1 there are 5 categories - the largest function value of the category information linked by "OR" or the smallest function value of the category information linked by "AND" is successively selected. In the case of bracketed expressions, the innermost bracket is used. In rule R1, the expression (II OR III) is evaluated first. For this purpose, one of the points labeled II and III is selected which represents the larger value of the membership function µ _{e '} . This is point II. The premise of rule R1 after this first step is: (I AND II) OR (IV OR V). The expression (I AND II) is then evaluated, ie points I and II are compared with one another and the smaller of the two is selected for further evaluation, namely point I. This simplifies the premise of rule R1 to: I OR (IV OR V). Thus, the expression IV OR V is evaluated next, with IV being obtained as a result. The premise of rule R1 is now: I OR IV. The result of this expression is IV.

The function value determined according to the procedure described above (e.g. point IV when evaluating the premise for rule R1) is transferred to the category specification of the conclusion (e.g. NUS for rule R1) (dashed, horizontal lines) by the Corresponding membership function µ _{u of} the normalized manipulated variable u is cut off horizontally at the level of the function value, so that the area shown in hatching in FIG. 8 is obtained.

Rule R2 uses the same procedure. There the function value determined in this way is zero in this case, there is no space left.

The results of the control applications are overlaid by forming the union of the areas determined by the control applications. The area determined in this way is shown in FIG. 8 below. It is identical to the area that was determined when rule R1 was applied, since rule R2 does not contribute to the example values e ₀ and e ′ ₀ .

A numerical value for the normalized manipulated variable u is obtained by averaging over the area obtained by the union operation. This averaging can be done, for example, by determining the geometric center of the area and projecting it onto the abscissa. From the standardized manipulated variable u determined in this way, a manipulated variable u adapted to the controlled system is finally determined, for example by multiplication with an empirically determined parameter.

Claims (10)

1. Method for controlling a variable influencing the driving dynamics or the driving state of a vehicle depending on the deviation ( e ) between a setpoint ( w ) and an actual value ( y ) of a signal related to this variable, characterized in that that by means of a fuzzy controller, a manipulated variable (20) (u) from the deviation (e) between desired and actual values and the time derivative (s') of this deviation is determined and the manipulated variable (u) to an actuator signal generator (16) is forwarded, which _forwards a signal ( u _PWM ) to an actuator ( 18 ). 2. The method according to claim 1, characterized in that for the normalized control deviation (e), the normalized time derivative (e ') of the control deviation and the normalized manipulated variable (u) each

• - several categories ("negative big", "negative small", "positive small" and "positive big") are defined as qualitative sizes and
• - Membership functions (µ _e , µ _{e ′} and µ _u ) are defined, which determine the degree of belonging to the categories as a function of the values derived from the standardized control deviation (e) or the normalized time derivative (e ′) of the control deviation or the standardized manipulated variable (u).

3. The method according to claim 2, characterized in that the train hearing functions (µ _e ) of the normalized control deviation (e), (µ _{e '} ) of the standardized time derivative (e') of the control deviation and (µ _u ) of the normalized manipulated variable (u ) are the same for the same category. 4. The method according to claim 3, characterized in that

• - the membership functions in a category "negative big" decrease from 1 to 0 if the values used in the membership functions pass through a range from -1 to 0,
• - the membership functions for a category "negative small" increase from 0 to 1 if the values used in the membership functions pass through a range from -1 to 0,
• - the membership functions in a category "positive small" decrease from 1 to 0 if the values used in the membership functions pass through a range from 0 to 1 and
• - The membership functions for a category "positive big" increase from 0 to 1 if the values used in the membership functions pass through a range from 0 to 1.

5. The method according to any one of the preceding claims, characterized in that the fuzzy controller ( 20 ) determines the normalized manipulated variable (u) by applying fuzzy rules to the normalized control deviation (e) and the normalized time derivative (e ') of the control deviation and then the results of these control applications are superimposed by averaging. 6. The method according to claim 5, characterized in that a 1st set of 2 fuzzy rules is applied if the normalized temporal derivative (e ') of the control deviation is greater than the negative of the normalized control deviation (e), wherein

• - A 1st fuzzy rule (R1) states that the normalized manipulated variable (u) is "negative small" if a 1st or a 2nd condition is fulfilled, whereby
  • - The 1st condition is met if the normalized control deviation (e) is "positive small" and the normalized time derivative (e ′) of the control deviation is "positive small" or "negative small" and
  • the second condition is met if the normalized control deviation (s) is "negative small" or "negative big",
• - A second fuzzy rule (R2) states that the normalized manipulated variable (u) is "negative big" if a 1st or a 2nd condition is fulfilled, whereby
  • - The 1st condition is met if the normalized control deviation (e) is "positive big" and
  • the second condition is met if the normalized control deviation (e) is "positive small" and the standardized time derivative (e ′) of the control deviation is "positive big",

and a second set of 2 fuzzy rules is applied if the normalized time derivative (e ′) of the control deviation is less than or equal to the negative of the standardized control deviation (e), where

• - A 3rd fuzzy rule (R3) states that the standardized manipulated variable (u) is "positive small" if a 1st or a 2nd condition is fulfilled, whereby
  • - The 1st condition is met if the normalized control deviation (e) is "negative small" and the normalized time derivative (e ′) of the control deviation is "positive small" or "negative small" and
  • the second condition is met if the standardized control deviation (s) is "positive small" or "positive big",
• - A 4th fuzzy rule (R4) states that the normalized manipulated variable (u) is "positive big" if a 1st or a 2nd condition is fulfilled, whereby
  • - The 1st condition is met if the normalized control deviation (e) is "negative big" and
  • - The 2nd condition is met if the normalized control deviation (e) is "negative small" and the normalized time derivative (e ′) of the control deviation is "negative big".

7. The method according to claim 1, characterized in that the set lersignalgenerator ( 16 ) is a pulse width modulator, which delivers a control variable ( u ) corresponding pulse width modulated signal ( u _PWM ). 8. The method according to any one of the preceding claims, characterized ge indicates that at least one of the systems steering, braking system and chassis are controlled. 9. Device for controlling a driving dynamics or the driving state of a vehicle influencing variable depending on the deviation ( e ) between a target ( w ) and an actual value ( y ) of a signal related to this variable, characterized in net that means are provided in the form of a fuzzy controller ( 20 ) which, based on this deviation ( e ′) between the setpoint ( w ) and actual value ( y ) and the time derivative ( e ′) of this deviation, are a manipulated variable ( u ) identify and further means (16) a of the manipulated variable (u) abhängendes signal (u _PWM) generate and transmit it to an actuator (18).