DE19932492B4 - Method for controlling an automatic transmission for motor vehicles - Google Patents

Abstract

The invention relates to a method for controlling an automatic transmission, in which a separation between the situation detection and the resulting effects takes place. In addition, harmonization of effects is not made until immediately prior to the conversion of the effects into signals to the actuators.

Description

The The invention relates to a method for controlling an automatic transmission for motor vehicles.

In Many areas of technology use controls in which a complex situation assessment is made. Under one In this context, situation becomes one affecting the system History of an input vector E (t) understood, which are assigned to a typical input pattern can. In this case, there is a specific reaction of the controller required and the further course of the input variables can be predicted within limits become.

• – den Eingangsgrößenvektors E(t) kontinuierlich auf Ähnlichkeiten mit vorgegebenen Mustern zu untersuchen,
• – ein Maß für die Übereinstimmung mit diesen Mustern zu bestimmen,
• – die für diese Situation abgespeicherten Regeln anzuwenden, um die Reaktion der Steuerung auf diese Situation zu bestimmen,
• – im Falle des gleichzeitigen Auftretens mehrerer Situationen die sich aus den Regeln ergebenden Reaktionen zu harmonisieren und
• – eine eindeutige Reaktion der Steuerung zu ermitteln.

The task of situation-aware control is to

• - to continuously examine the input variable vector E (t) for similarities with given patterns,
• - to determine a measure of compliance with these patterns,
• - apply the rules stored for this situation to determine the controller's response to this situation,
• - in the case of the simultaneous occurrence of several situations, to harmonize the reactions resulting from the rules and
• - to determine a clear reaction of the controller.

In The rule is not only the currently incoming input variables for situation detection decisive, but also their temporal development in the past. In addition, the gradients generally not by concrete values, but only by the characteristic Overall appearance of one or more input variables detected can be.

One example for A situation-recognizing control is the control of the automatic transmission of Motor vehicles with adaptive switching programs. From these switching programs is expected in their reactions to drivers, driving style and adjust vehicle condition. The driver and the driving situation, however, can only indirectly via a limited number of input variables are detected. In addition, there are hardly two identical drivers or driving situations. It are therefore characteristic input pattern recognizable and driving situation, To classify driver and vehicle condition in a rough grid. For all combinations must the control adequate Find behaviors that take into account some peculiarities also considers the manual shift driver in the gear selection into consideration.

In Previous controls of automatic transmissions is a module of Control not only for the recognition of a situation in charge, but calculated at the same time the time course of a signal or a characteristic value, the the desired effect on the manipulated variables of the Control corresponds; d. H. the "detection a situation "and the resulting "reaction" is immediate linked together. This link is called "impact relationship" in the following. The of The manipulated variables issued to the control essentially become from a total characteristic value formed from the result of all modules calculated. The recognition of a situation and its effect on the manipulated variable of the controller are thus summarized in a control module.

Overlapping relationships are taken into account by a series connection of detection blocks. The consequence of the series connection is that, when an effect arranged further behind in the action sequence occurs, the further effects above are modified. Overlapping effects are represented by a parallel connection; ie the effects are taken into account pro rata. This approach results in several disadvantages, which are explained below:
The relationships in the structure quickly become intransparent; This means that it is only comprehensible with considerable effort how the individual control modules contributed to the overall result. This is increasingly true when an attempt was made to model different effects (multidimensional structure of effect) on one parameter.

at the introduction a new feature for situation detection needs this to be "appropriate" location of the structure be introduced to not the function of the entire controller to endanger. In general, the function continues with increasing priority Rear arranged in the action strand. With increasing complexity it will increasingly difficult to determine this position. With many impact dimensions However, there is usually not only one, but for each dimension one of the best suitable place.

at multidimensional relationships can happen that the priorities the modules for every dimension is different. In this case, the structure looks in each effect relationship differently, which further complicates the clarity. In practice it becomes common one for all impact dimensions uniform structure chosen, however the output of a another module in addition introduced as an input, around for single effect dimensions to carry out a compensation.

at a known from DE-OS 43 12 717 A1 control of an automatic transmission was trying to select a circuit program using a Rating table in which the driving conditions city traffic, mountain driving and trailer operation to enter, to systematize. The evaluation of the evaluation table allows the definition of a second parameter for the selection of the circuit program. simultaneously becomes a priority the driving conditions established with each other, so that when two driving states occur clearly It is clear which driving state when selecting the shift program the rash.

adversely At this control is that always only exactly one recognition block determines the effect, namely the one who after the entries in a Evaluation table and a procedural instruction at a time allocated. This also leads at gradual changes the situation to a sudden behavioral change of control. Furthermore does not find a situation adequate proportionate consideration several influences at one time instead, because only the dominant situation is evaluated.

Of the Invention has for its object to provide a control which is constantly changing adapted to situations that take several influences into account simultaneously and their structure in the detection of complex patterns and situations in the input variable vector and in multi-dimensional relationships, as in recognition and processing complex situations, remains transparent, easy to maintain and expand.

These The object is achieved by a method for controlling an automatic transmission solved by motor vehicles, having the features of claim 1.

This The method has the advantage that several effects are considered simultaneously and the behavioral changes the control is steady. Furthermore are in this process in every situation all impact relationships, which influence the behavior of the controller, transparent. Overlapping and overbearing ones Wirkungsäste are at a glance. This can be a fault diagnosis perform in a simple manner and parameterize the control more easily.

In addition, new ones Detection blocks in a clear hierarchical recognition scheme to get managed. Your position in the structure will not work anymore by the priority the effect of this function. The effects can be in managed in a similar scheme. The multiple use The often extensive effects are thereby structurally supported.

Farther is for the relationship of effects control-specific a catalog of possible Created properties from which the user is to create a new one configured. This also allows the relationships between Detection and effect in the form of standardized rules memory space and make optimal computing time.

Finally results for all impact dimensions even with individual prioritization for every Dimension a unique arrangement of recognition and action elements. In the structure you can thus easily find your way and the extension the structure with additional Recognition and effect functions are easy.

• – die Fahrsituationen des Kraftfahrzeugs erkannt werden,
• – der oder die Kennwerte der Wirkungen der Fahrsituation nach einer oder mehreren vorgegebenen Regeln bestimmt werden,
• – die Wirkungen harmonisiert werden, sowie
• – die Wirkungen umgesetzt und eine Schaltstufe gewählt wird.

The object mentioned in the introduction is also achieved by a method for controlling an automatic transmission of motor vehicles, in which

• The driving situations of the motor vehicle are recognized,
• The characteristic (s) of the effects of the driving situation are determined according to one or more predetermined rules,
• - the effects are harmonized, as well
• - The effects implemented and a switching stage is selected.

The Advantages resulting from this process correspond to the o. G. Advantages of the method according to the invention.

In an embodiment of the invention, the recognition of driving situations is carried out by continuously comparing an input quantity vector E (t) with patterns of different driving situations, and depending on the degree of correspondence between the relevant quantities e _{u of} the input quantity vector E and the patterns of driving situations, Characteristics of the driving situations f _l , where l = 1 ... number of patterns, formed so that a quantification takes place.

A variant provides that from the driving situation characteristic values f _l one or more control output values r _i , with i = 1 ... number of possible effects, are formed according to predetermined rules R _k , with k = 1 ... number of rules, so that the control output values are available in quantified form.

In addition to the invention, it is provided that one or more rules R map the driving situation characteristic values f _l to a control output value r _i by means of a linear or non-linear characteristic, and that each rule R contains a parameter q which determines whether the rule supersedes mend or superimposing and a parameter p, which acts as a weighting factor a in the case of overlapping rules and as a priority factor s in the case of overriding rules, so that the conversion of the situation characteristic values into action characteristics is uniform and the further processing, in particular the harmonization of the effects, is facilitated.

In another embodiment of the invention, it is provided that all control output values r _s (R) acting on an effect, with s ε i, by matching or superimposing on an action characteristic w _i , with i = 1 ... number of effects, mapped , so that an effect index is determined for each effect.

at A variant of the invention is provided that the figure has the parameters q and p, where q is overconcentration or superposition signals and thereby the value of the parameter p as a weighting factor a or priority factor s is interpreted so that the harmonization of the effects facilitates becomes.

Another variant of the method provides that all effect parameters w _{i are} transformed such that they can be expressed as a linear combination of linearly independent effects w _{j, h} , so that the effects are described completely and without contradiction. Hereinafter, these effects are referred to as "harmonized".

Another embodiment of the invention provides that the implementation of the harmonized efficiency parameter w _{j, h is carried out} by interpolation between an upper and a lower limit G _o and G _{u of} the effect s according to the action characteristics w _s , so that the control continuously to changed driving situations is adapted.

Further Advantages and advantageous embodiments of the invention are the following description, the drawing and the claims. In the drawing are exemplary embodiments of the subject invention and described in more detail below. Show it:

1 : the schematic structure of a controller according to the invention;

2 a schematic representation of another embodiment of a controller according to the invention;

3 a detail of a control according to the prior art and

4 : an example of the structure of a control according to the prior art

• 1. (Situations- und Muster-) Erkennung
• 2. Wirkung
• 3. Regeln zur Verbindung von "Erkennung" und "Wirkung"
• 4. Umsetzung

In 1 the schematic structure of a controller according to the invention is shown. In this control, detection and effect are arranged separately and connected with rules. These also contain the rules for prioritization, so that recognition and impact modules can be arranged thematically. The controller consists of the following components:

• 1. (situation and pattern) detection
• 2. Effect
• 3. Rules for Linking "Detection" and "Effect"
• 4. Implementation

Ad 1 .: Each function in the "Detection" block is responsible for recognizing a particular situation or pattern. It continuously searches the input variable vector E (t) for the patterns and situations characteristic of it. In principle, it has access to all input variables. It makes sense to restrict access to the relevant input variables for this situation. When a pattern or situation is detected, the output of that function changes. Depending on the character and the desired effect, a recognition can output its output in binary quantities (eg detected / not recognized), in quantized quantities (eg not recognized, weakly recognized, recognized) or continuously (eg on a percentage scale).

The Detection functions can be hierarchically nested in different levels. Lying close The order criterion is the content of a situation. The Functions can within a level and across levels arbitrarily linked be. For example, a special function can only be activated if the parent Level has recognized a more general situation. A link is For example, it makes sense to exclude two situations. In addition, the detections can be provided with a prioritization: the detection of safety-relevant conditions can be more common then be activated as the rest Recognition functions.

To 2nd: In the block "Effect" are from the rules and detection resulting effects on control behavior collected and summarized, that she can be implemented without contradiction.

This block is comparable in its structure (output variables, levels, links) to the "detection" block, except that possible effects on the control behavior are hierarchically arranged here. Input variables here are the results of the rules applied to the block "detection": usually continuous sizes, the depend on the output size of the detection and the type of rule.

The listed Effects affect the overall control behavior and can do this roughly or describe with keywords (for example, behavior should become "sluggish"), so far it for each effect can be clearly formulated mathematically / physically can. The impact catalog is best designed to suit the user the rule formulation is facilitated. It should be him - similar to the strategy pursued in fuzzy logic - taken from his way of thinking Terms are offered. The offered effects are allowed to each other supplement, overlap, contradict.

task of the action block it is, the effects without contradiction and under Attention to the boundary conditions. You are then from Implement the "Implementation" block directly into control signals. This can be understood mathematically that the n arranged hierarchically Impact dimensions basically just a m-dimensional sphere of effect describe where m <n is. The action block therefore transforms the n-dimensional space of action into an m-dimensional: Any point in this space can be unique denoted by a coordinate system of m linearly independent vectors become. The transformation does not have to be linear and not ambiguous.

clear expressed have to the terms of action adapted to the mindset of the user in clear Control instructions are implemented and interact with the others activated effects. For example a little slower Control behavior required, so may have a response time elevated become. At the same time, however, a critical operating state is detected been fast on the at certain development of input variables must be reacted. The reaction time is thus generally prolonged, but shortened for effects that concern the critical operating condition.

effects can also generally with priorities provided, for example, to show that an effect always takes precedence over another.

To 3 .: Rules for linking of detection and effect. The patterns and situations are in the Presentation of the user with conclusions for the control behavior connected. For every Situation are all the necessary conclusions in terms of rules and into the system. The formulation is control-specific to choose: It could For example, weighting factors may be selectable to main and side effects to quantize. Furthermore, it is conceivable to introduce an option "overlaying" / "overruling" to if one rule is effective, the others take a back seat allow. For Safety-related functions make sense of priority factors. For this should analyze the character of the impact relationships and then the impact relationships are categorized and cataloged. On this basis, each rule can then reference to this Catalog with parameters to be configured.

each Recognition function can with any impact function - even cross-level - with a rule connected become. For A detection function may have several outgoing rules give. For An effective block can have several rules that affect it.

To 4 .: In the implementation, the contradiction-free and exactly specified Information from the block "Effect" converted into manipulated variable curves and spent. In the example of a reaction time, this means: Depending on the course of the input variables responded to these differently fast.

at the conversion of a control according to the prior art to a Control according to the invention Become modules that detect a situation and its impact in the control, in recognizing and acting components divided up. These are then separated into functions hierarchically or object-oriented arranged. For the connection of the recognition and The effect block determines the relationship of effects that affect the relationship of effects in the previous structure. It is recommended for the sake of Space and computing time needed to catalog these and to configure each rule with parameters.

Based two embodiments should the above be illustrated.

Embodiment 1

A Adaptive transmission control fits their behavior to the driver, driving situation and vehicle condition. It However, it is based on specific driving situations and drivers, for example in the same or similar Way reacts. Safety functions in the transmission come first and foremost others. Input variables in the exemplary embodiment are the parameters "accelerator pedal position", "speed" and "lateral acceleration".

In the "Detection" block, all the situations occurring at the moment are first recognized from the input data. From a moving average of the accelerator pedal speed and an average duration of the accelerator pedal in one range Strong activity calculates the driver's sportiness recognition a sportiness of 90%. As a result - since nothing else has been detected - the shift points are shifted towards greater speeds: the vehicle only shifts at higher engine speeds.

Now the vehicle is driven rapidly in a curve (high lateral acceleration, high speed): Curve detection has a critical situation recognized. This affects on the one hand in the confirmation or increase in the Sportiness, on the other hand, a shift should now be avoided as possible to avoid torque fluctuations. There is therefore the Recognition link "When cornering strongly active, then high sportiness is present "and the rule leading to the effect" If curve recognition very active, then activate effect, switching as possible avoid '. "The latter is overriding, because it is security relevant. Slows down the vehicle, should be due to the high level of sportiness even at high engine speeds be switched. The reconciliation the other rule leads however, that only the effect "as possible no circuit "in the action block activated and the circuit is not executed.

The Vehicle now leaves the Turn and drive a gradient. Detection recognizes "slope". It should only the switching points for the switchback to higher Speeds which remain upshift points unaffected. By the overlay with the recognized high sportiness now the block "effect" must have both effects harmonize. This is achieved by all switching points in the direction of higher speed be moved for the switchback but a little bit stronger. It will be checked at the same time whether the upshift points are not below the switchback points. hereby it would come to a malfunction.

Embodiment 2:

to 1. Detection: In this exemplary embodiment, the input data vector E (t) consists of the time profiles of the measured values "accelerator pedal position", "speed" and "lateral acceleration" and there are the recognition functions:
Driver Sportlichkeits detection,
Curve detection and
Slope detection.

by virtue of the structural freedom afforded by those proposed in the patent Arrangement results, one finds the driver-sportiness recognition in the group "driver recognition", the curve recognition is located in the driving situation detection, the "slope detection" in the "mountain detection" and in turn in the "driving situation recognition".

All detection blocks are always active and search the input data vector for the corresponding patterns that lead to detection.

output For example, a characteristic value between 0 and 100%, depending on the intensity of the occurred Situation, condition, driving style.

to 2. Rules: In the present example, the rule consists of a combination of an output value of the recognition e with a nonlinear characteristic f. Another parameter of the rule is the effect "overriding s" / "overriding a". The rule r _i is therefore an element (f _r , p _r ) with f _r ε {set of all possible characteristics and p _r ε {s, a}. The evaluation of the rules takes place continuously at any time to a value r.

Thus, for the example for every rule k: r _i = f _k (e _l )
With i ε {1, ..., number of effects}
k ε {1, ..., number of rules}
l ε {1, ..., number of recognition functions}

regulate can be formulated between each recognition and effect blocks, and although on every level of both blocks. In an effect can be arbitrary many rules flow From a recognition as many rules go out.

to 3. Effect: Input variables in this case are the evaluated control values. In this example are relevant: "circuit possible avoid "," shift the Upshift points "and" shift all Switching points " Block "detection" is the effect "shifting the Upshift points "the subsequent subordinate. All blocks are always active. The output is in a default position, generally 0, and will only change if a situation has been identified that with a rule on the present Effect linked is.

Also the effects can depending on the extent of Effectiveness take different characteristics, e.g. on a Percentage scale.

• Wenn "Sportlichkeits-Erkennung", dann (f₁, a) "Verschiebung aller Schaltpunkte".
• Wenn "sehr scharfe Kurvenfahrt-Erkennung", dann (f₂, s) "Möglichst keine Schaltung"
• Wenn "Bergab-Erkennung", dann (f₃, a) "Verschiebung aller Schaltpunkte",
• Wenn "Bergab-Erkennung", dann (f₄, a) "Möglichst keine Schaltung", so errechnet sich die die Wirkung wie folgt: Haben nur "Sportlichkeits-Erkennung" (Ausgabe es) und "Bergab- Erkennung" (Ausgabe e_B) die entsprechende Situation erkannt, so ist die Wirkung "Verschiebung aller Schaltpunkte" w_V: WV = r1 + r3 = f1(eS) + f2(eB) (Überlagerung der Regeln)Hat zusätzlich die "scharfe Kurvenfahrt-Erkennung" (Ausgabe e_K) angesprochen, so gilt für die Wirkung "Möglichst keine Schaltung": WK = f2(eK)

Now, according to the embodiment, the rules apply:

• If "sportiness detection", then (f ₁ , a) "shift all switching points".
• If "very sharp cornering detection", then (f ₂ , s) "possible no circuit"
• If "downhill detection", then (f ₃ , a) "shift all switching points",
• If "downhill detection", then (f ₄ , a) "possible no circuit", then the effect is calculated as follows: If only "sportiness recognition" (output it) and "downhill detection" (output e _B ) have recognized the corresponding situation, then the effect "shift of all switching points" w _V : W V = r 1 + r 3 = f1 (e S ) + f2 (e B ) (Superposition of rules) In addition, if the "sharp cornering detection" (output e _K ) has been addressed, the effect "possible no switching" applies: W K = f2 (e K )

rule 4 does not play because it was overruled by rule 2.

The output quantities of the effective block are linearly independent in the action space, ie they can be implemented without contradiction and an output variable of the effect y can not be expressed by a combination of others. This results in the output variables for the switching point control:
Shift of the upshift points y _VH
Shift of the switch-back points y _VR
If possible no circuit y _K

In In this case, therefore, the whole impact catalog on three effects be imaged, anyway already as an effect in the block of action are included. This must, however not be like that.

In the exemplary embodiment, there is one of the 1st and 2nd effect superordinate effect "shift of all switching points" w _V.

The contradiction-free conversion to the above quantities 1 and 2 is done in this case over the arithmetic mean: y VH = ½ (w V + w VH ) and y VR = ½ (w V + w VR ) with w _VH ... value of the effect "shift of the upshift points" and w _VR ... value of the effect "shift of the switchback points"

to 4. Implementation: The block implementation refers in the following to the state of the art, ie the implementation with shift maps S. Here there are two shift maps for the downshift S _RU and S _RO , which form the lower and upper limits of the allowed variation. The same applies to the upshift with the switching maps S _HU and S _HO . The controller performs interpolation between these fields. The currently valid fields are to be determined.

• Für die Hochschaltung: Ein zwischen S_HU und S_HO in der Gewichtung 2:3 interpoliertes Feld S_Hres
• Für die Rückschaltung: Ein zwischen S_RU und S_RO in der Gewichtung 1:1 interpoliertes Feld S_Rres

If, for example, due to the previous processing, y _VH = 60%, y _VR = 50% and y _K = 70%, the following fields would be used:

• For the upshift: A field S _Hres interpolated between S _HU and S _HO in the weighting 2: 3
• For the downshift: A field S _Rres interpolated between S _RU and S _RO in the weighting 1: 1

Should a circuit take place on the basis of the fields thus determined, then, for example, the ascertained _upshift field S _Hres snichmal _{would be} interpolated with S _HO with the weighting 3: 7 for an _{upshift to be} performed. This would avoid a circuit under certain circumstances.

To These three steps can thus have a clear switching map be determined. Applied to the current accelerator pedal position, the speed and the current gear, can be read from these, if a circuit is performed. The selected gear is the output of the block "implementation".

DKl:

Stellung der Drosselklappe;

A_IST:

Ist-Beschleunigung;

DK_HS:

Dynamikkennzahl für die Hochschaltpunkte;

DK_RS:

Dynamikkennzahl für die Rückschaltpunkte;

A_QUER:

Querbeschleunigung und

N_MOT:

Motordrehzahl.

In 2 is a further example of a controller according to the invention shown. In this case, the time-variable quantities DK1, A_IST,... Form the input variable vector, by means of which the situation is recognized. Mean

DKL:

Position of the throttle valve;

A_IST:

Actual acceleration;

DK_HS:

Dynamic index for the upshift points;

DK_RS:

Dynamic index for the switch-back points;

a_Quer:

Lateral acceleration and

N_MOT:

Engine speed.

The Effect relationships are formed in this case by fuzzy rules. The output variables DK_HS (Relocation of Upshifts) and DK_RS (Relocation of Downshifts) can be done in each Combination can be carried out without contradiction, even in Combination with the not yet named circuit prevention.

In the controller after 2 There are, among others, the characteristic "DK_HS" for the switching points of the upshifts and "DK_RS" for the downshifts. High sportiness leads to an increase in both, the gradient only to increase the switch-back points. The previously common common characteristic DK for upshifts and downshifts has been eliminated. Its maintenance could lead to contradictory and ambiguous information - high DK, lower DK_RS and DK_HS - which should not appear at this point after the description.

In 3 a detail of a control according to the prior art is shown. The module has n higher priority over the others and is therefore downstream. "Module n + 1" does not work parallel to module 2 in at least one effect dimension. Therefore, partial compensation takes place with its output variable.

In 4 For example, a prior art controller combining detection and action is shown.

The Concept underlying the control according to the invention It is also possible to transmit other situation-aware control systems. examples for Such controls are the control of the brewing process during beer brewing and anti-sound systems that are sensitive to certain noise phenomena in of the kind that they react generate opposite sound pattern caused by interference the noise pollution the actual source of noise minimized.

Claims (5)

Method for controlling an automatic transmission for motor vehicles, characterized by the following method steps: detecting the driving situations of the motor vehicle by the continuous comparison of an input variable vector E (t) with patterns of different driving situations, forming driving situation characteristics f _l , with l = 1 ... number the pattern, depending on the degree of correspondence between the relevant quantities e _{u of} the input quantity vector E and the patterns of driving situations, - determining one or more control output values r _i , where i = 1 ... number of possible effects, according to given rules R _k , with k = 1 ... number of rules, from the driving situation parameters f _l , - mapping of all control output values r _s (R) acting on an effect, with s ε i, by overconciliation or superimposition on an effect characteristic value w _i , with i = 1 ... Number of effects, - Selection of a switching stage as a function of the effect characteristics w _i . Method according to claim 1, characterized in that one or more rules R map the driving situation characteristic f _l to a control output value r _i by means of a linear or non-linear characteristic, and that each rule R contains a parameter q which determines whether the rule is overriding or superimposing has the effect that each rule R contains a parameter p, which acts as a weighting factor a in the case of overlying rules and as a priority factor s in the case of overriding rules. Method according to Claim 2, characterized in that the mapping of all control output values r _s acting on an effect has the parameters q and p, where q indicates overshoot or superimposition and as a result the value of the parameter p is interpreted as weighting factor a or priority factor s. Method according to one of the preceding claims, characterized in that all effect parameters w _{i are} transformed in such a way that they can be expressed as a linear combination of linearly independent effects w _{j, h} . A method according to claim 4, characterized in that the implementation of the harmonized efficiency index w _{j, h is carried out} by interpolation between an upper and a lower limit G _o and G _{u of} the effect s.