JPH11212763A - Electronic controller - Google Patents

Abstract

PROBLEM TO BE SOLVED: To appropriately handle data within the accuracy of a desired physical quantity in an electronic controller for handling floating point type data. SOLUTION: An engine control ECU(electronic controller) 10 executes a fuel purge control 1 by a purge valve 8 or the like based on detection signals of a sensor 9. An FPU(floating point arithmetic processor) 15 inside the ECU 10 executes floating point type arithmetic operations. Floating point type data are composed of a code part of 1 bit, an exponent part of 8 bits and a mantissa part of 23 bits. The ECU 10 approximates the mantissa part corresponding to the valid but number of the mantissa part so as to provided a desired resolution. At this time, whether the next bit of the valid bit number of the mantissa part is '0' or '1' is discriminated, and when the next bit of the valid bit number of the mantissa part is '1', the bit is carried by '1' and the next and succeeding bits of the valid bit number of the mantissa part are cleared to '0', Also, the ECU 10 compares respective values after approximation for a control value and a prescribed judgment value and ON/OFF controls the purge valve 8 corresponding to the compared result.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an electronic control unit for calculating floating-point data including a sign part, an exponent part and a mantissa part.

[0002]

2. Description of the Related Art For example, in an electronic control unit (ECU) that controls an engine, various controls are performed using integer data. The integer type data in the ECU is LSB
The physical quantity per bit is used, and the comparison instruction for the integer type data is always performed within the precision of the LSB, that is, within a certain physical quantity precision range. This was very effective in that subtle changes in physical quantities were not reflected in engine control.

On the other hand, in recent years, it has been studied to introduce a floating-point arithmetic processor (FPU: Floating-Point Unit) and perform engine control using floating-point data capable of directly handling physical quantities. Floating-point data is composed of a sign part, an exponent part, and a mantissa part. For example, in the case of floating-point data consisting of 4 bytes, having a mantissa of 23 bits makes it always 7
The operation is performed with a precision of a digit (a precision of 1/2 ^ 23 ≒ 0.0000001), and a very fine precision can be obtained as compared with the conventional integer type data.

[0004]

However, when the above-mentioned floating-point operation is applied to engine control, the control value represented as floating-point data fluctuates inadvertently, causing a problem that control stability is lacking. . For example, in general-purpose engine control in which a valve (actuator) is turned on / off when a ≧ b, hysteresis substantially corresponding to LSB exists as shown in FIG. On the other hand, in the case of floating-point data, as shown in FIG. 8B, the accuracy is improved, so that the hysteresis width (the hysteresis width) is reduced, which causes a problem that the valve hunts.

SUMMARY OF THE INVENTION The present invention has been made in view of the above problems, and an object of the present invention is to provide an electronic control device that handles floating-point data, in which data can be appropriately converted within a desired range of physical quantity accuracy. It is an object of the present invention to provide an electronic control device that can be handled in a computer.

[0006]

According to a first aspect of the present invention, there is provided an electronic control apparatus for calculating floating-point data comprising a sign part, an exponent part, and a mantissa part. There is provided an approximating means for designating the number of effective bits from the upper part of the mantissa to give a desired resolution, and approximating the mantissa in accordance with the designated number of effective bits.

In short, floating-point data is approximated by arbitrary specification of the number of effective bits so that the floating-point data can be compared within a certain physical quantity precision range such as integer-type data. The precision was reduced to the minimum required. That is, the precision is arbitrarily adjusted by manipulating the number of effective bits of the mantissa part of the floating-point data. As a result, even if the data is floating-point type data, it is possible to have the His width with the same accuracy as that of the integer type data. As a result, in an electronic control device that handles floating-point data, data can be appropriately handled within a desired range of physical quantity accuracy.

Actually, as described in claim 3, the next bit of the number of effective bits of the mantissa is "0" or "1".
If the next bit of the number of effective bits of the mantissa is “1”, the bit is rounded up by “1” (round-up means), and the number of effective bits of the mantissa is calculated. The next bit and subsequent bits are cleared to "0" (clearing means). According to the above configuration, the approximation processing of the floating-point type data can be easily and accurately performed.

According to a second aspect of the present invention, in order to provide a desired resolution, the number of effective bits is designated from the upper part of the mantissa, and the mantissa is designated according to the designated number of effective bits. An approximation unit for approximation and a comparison unit for comparing a control value and a predetermined determination value with values approximated by the approximation unit are provided.

In this case, similarly to the first aspect, in the electronic control device that handles floating-point data, the data can be appropriately handled within a desired range of physical quantity accuracy. Further, even when the control value fluctuates, the comparison determination can be performed with a desired hiss width (width of physical quantity accuracy). Therefore, in general-purpose engine control in which a valve (actuator) is turned ON / OFF according to a result of a magnitude comparison between a control value and a determination value, the problem of the related art such as hunting of a valve is solved.

Incidentally, it has a 23-bit mantissa,
When the number of effective bits is 15 bits, FIG.
As shown in (b), when the determination value b is “128”, the floating-point type data (control value a) has a hysteresis width (LSB: 1/256 = 0.00) equivalent to the existing integer type data.
390625).

[0012] In the second aspect of the present invention, the comparing means includes, as described in the fourth aspect, first comparing means for comparing the sign parts of two data to be compared, A second comparing means for comparing the exponent parts of the two data, and a third comparing means for comparing the mantissa parts approximated by the approximating means for the two data similarly to be compared.
And comparing means. In this case, by comparing each of the sign part, the exponent part, and the mantissa part of the floating-point data (however, the comparison of the mantissa part is an approximate value), the data can be reliably compared and determined.

According to a fifth aspect of the present invention, there is provided an electronic control unit for performing various controls of a vehicle engine, wherein the number of effective bits is set according to an engine control type to which an approximation value by the approximation means is applied. It is supposed to be. For example, when the required control accuracy is relatively high, such as fuel injection control or electronic throttle control, increase the number of effective bits. Conversely, when the required control accuracy, such as fuel gas purge control, is relatively low, , The number of effective bits is reduced. As a result, it is possible to set an appropriate physical quantity accuracy for the control value while ensuring the control accuracy in any engine control.

According to a sixth aspect of the present invention, there is provided an electronic control apparatus having a floating-point arithmetic processor for calculating floating-point data comprising a sign part, an exponent part, and a mantissa part, wherein the floating-point arithmetic processor is In order to have a desired resolution, the number of effective bits is specified from the upper part of the mantissa, and the mantissa is approximated according to the specified number of effective bits, and the approximated value is determined by a predetermined determination value. Compare with

[0015] In this configuration, similarly to the first aspect, in the electronic control device that handles floating-point type data, the data can be appropriately handled within the range of the desired physical quantity accuracy. Further, by performing approximation and comparison operations using a floating-point arithmetic processor, it is possible to realize high-speed operations.

[0016]

DESCRIPTION OF THE PREFERRED EMBODIMENTS One embodiment of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram schematically showing an engine control system. In the present system, the on-vehicle engine 1 is configured as, for example, a gasoline injection type multi-cylinder internal combustion engine, and its intake pipe 2 is provided with a throttle valve 3.

A canister 6 is connected to a fuel tank 4 containing liquid fuel (gasoline) through a purge pipe 5. An adsorbent 7 made of, for example, activated carbon for adsorbing the fuel gas generated in the fuel tank 4 is accommodated in the canister 6. With such a configuration, the fuel gas generated in the fuel tank 4 is taken into the canister 6 via the purge pipe 5 and is adsorbed by the adsorbent 7 in the canister 6.
Is adsorbed. In the purge pipe 5, an ON / O connection is provided between the canister 6 and the throttle downstream of the intake pipe 2.
An FF-driven purge valve 8 is provided.

In such a case, the purge valve 8 is opened (O
N), the intake pipe 2 and the canister 6 are in communication with each other, and conversely, the purge valve 8 is closed (OFF), so that the intake pipe 2 and the canister 6 are closed. The fuel gas adsorbed by the adsorbent 7 in the canister 6 is connected to the intake pipe 2 based on the negative pressure generated in the intake pipe 2 in the communication state based on the opening of the purge valve 8.
Introduced within.

The engine control ECU 10 includes a microcomputer (hereinafter referred to as a microcomputer) 11, which includes a central processing unit (CPU) 12, a readable and writable storage device (RAM) 13, and a read-only storage device (R).
OM) 14 and a floating point processor (FPU) 1
5 and an input / output device (I / O) 16 are provided.
The ECU 10 receives various kinds of information indicating the engine operating state from a sensor 9 provided in the engine 1. The sensor 9 includes, for example, an air flow meter for detecting an intake air amount, a water temperature sensor for detecting a cooling water temperature, a rotation speed sensor for detecting an engine rotation speed, and the like. The ECU 10 performs control of fuel injection by an injector (not shown), control of fuel purge by the purge valve 8, and the like based on the input sensor signal.

Here, the FPU 15 performs an operation in a floating-point format. The floating-point data calculated by the FPU 15 is configured in accordance with, for example, the IEEE 754 standard. As shown in FIG. 2, the breakdown includes a 1-bit sign part, an 8-bit exponent part, and a 23-bit mantissa part. Have. In the case of 4-byte floating point data in which the mantissa part is composed of 23 bits, a resolution of 7 digits (0.0000001)
Resolution), but in the present embodiment,
By designating the upper n bits of the mantissa as the “number of effective bits”, the resolution of the mantissa is reduced to the minimum necessary.

Next, the operation of the present embodiment will be described by taking ON / OFF control of the purge valve 8 as an example. As an outline of the purge control, the control amount a and the determination value b are compared in magnitude. If a ≧ b, the purge valve 8 is turned on, and if a <b, the purge valve 8 is turned off. In the course of such basic control, the FPU 15 performs a floating-point operation on the control amount a and the determination value b, and specifies the upper n valid bits for the mantissa of the floating-point data (see FIG. 2). Then, the control amount a and the determination value b are approximated within the range of the effective bit number n.

FIG. 3 is a flowchart showing a purge control routine, which is executed by the engine control ECU 10.
The processing is performed by, for example, a cycle of 64 ms. Referring to FIG. 3, the CPU 12 first
Then, a basic control amount a1 in the purge control is obtained, and in a subsequent step 120, a water temperature correction value a2 is obtained. At this time,
In step 110, a basic control amount a1 is calculated in accordance with, for example, the engine speed and the intake air amount detected by the sensor 9, and in step 120, a water temperature correction value a2 is similarly calculated in accordance with the water temperature detected by the sensor 9. Is calculated.

Next, the CPU 12 multiplies the basic control amount a1 obtained in step 130 by the water temperature correction value a2 to obtain a final control amount a (a = a1 · a2). On this occasion,
The values of a1, a2, and a are calculated by the FPU 15 in a floating-point type at high speed.

Thereafter, the CPU 12 calls the comparison function cpm (a, b, n) in step 200. The details of the comparison function will be described later with reference to FIGS.
The outline is as follows.
4 and a predetermined determination value b set in advance is compared with each other, and a determination of “true” or “false” is made according to the comparison result.

Thereafter, in step 140, the CPU 12 determines whether the result of the determination by the comparison function is true or false. If the determination result is true (if a ≧ b), the CPU 12 turns on the purge valve 8 in step 150, and then ends this routine once. If the determination result is false (if a <b), the CPU 12 turns off the purge valve 8 in step 160.
F, and thereafter this routine is once ended.

FIG. 4 and FIG.
6 is a flowchart illustrating a comparison function at 0. FIG.
In step 201, the CPU 12 first divides the final control amount a into a sign part as, an exponent part ax, and a mantissa part ay in step 201, and then, in step 202, determines the determination value b in the sign part bs
And an exponent part bx and a mantissa part by.

After that, the CPU 12 determines in step 203 the sign portions as and b of the final control amount a and the determination value b.
Determine if s matches. If as ≠ bs (NO in step 203), the CPU 12 proceeds to step 2
Proceeding to 04, it is determined whether or not as = positive and bs = negative. If step 204 is YES, the CPU 12
In FIG. 3, the determination result is “true” in step 205.
Return to the routine. If step 204 is NO, C
The PU 12 returns the determination result to “false” in step 206 and returns to the original routine of FIG.

On the other hand, when as = bs (step 203)
Is YES), the CPU 12 proceeds to step 207, in which the mantissa part ay of the final control amount a (23-n
-1) It is determined whether or not the bit is "1". Here, "n" is the number of effective bits in the mantissa part ay, and in step 207, the (23-n-1) th bit counted from the right end in the floating point data of FIG. 2 is "0". Or “1”.

If step 207 is NO, the CPU 1
2 proceeds directly to step 209, and if step 207 is YES, step 20
Go to 9. In step 208, the CPU 12 adds "2 ^ (23-n-1)" to the mantissa part ay. As a result, the (23-n) th bit of the mantissa part ay is rounded up. Further, the CPU 12 determines in step 209 that the mantissa part ay is 0 to 0.
Clear all (23-n-1) bits to "0". According to steps 207 to 209, the mantissa part ay is approximated by the effective bit number n.

The approximation processing of the mantissa part ay will be described with an actual example. As shown in FIG. 6, when the number n of effective bits is "10", for example, floating-point type (float type) control values "0x3F800FFF (1.0004882 in decimal)" to "0x3F7FF000"
(0.99975586 in decimal) "are all 0x3F
800000 (1.0000000 in decimal) ". When (23-n-1) = the 12th bit of the mantissa part ay is "1" (for example, in the case of 0x3F7FF000 in FIG. 6), "2 ^ 12" is added to the mantissa part ay. (Step 2
08). Also, the 0th to 12th bits of the mantissa ay are forcibly cleared to "0" (step 209). Thereby, “0x3F7FF000” is approximated as “0x3F800000”. When the (23−n−1) = 12th bit of the mantissa part ay is “0” (for example, in the case of 0x3F800FFF in FIG. 6), the 0th to 12th bits of the mantissa part ay are forcibly set to “0”. (Step 209). As a result, "0x3F
"800FFF" is approximated as "0x3F800000".

Thereafter, the CPU 12 determines in step 210 whether the 23rd bit of the mantissa part ay is "1". Only when the 23rd bit of the mantissa ay is "1",
The CPU 12 adds “1” to the exponent part ax in step 211, and in the subsequent step 212, adds 2 to the mantissa part ay.
The third bit is set to “0”.

On the other hand, in steps 213 to 218 in FIG. 5, the mantissa part by of the decision value b is approximated by the number of effective bits n in the same procedure as in steps 207 to 212 in FIG. Each step is simply described: In step 213, it is determined whether or not the (23-n-1) th bit is "1" in the mantissa part by of the determination value b. In step 214, "2 ^ (23-n-1)" is added to the mantissa part by.
Is added. In step 215, the mantissa part by of 0 to (23-n-
1) Clear all bits to "0". At step 216, it is determined whether or not the 23rd bit of the mantissa part by is "1". Step 217 only when step 216 is YES
At step 218, "1" is added to the exponent part bx, and at step 218, the 23rd bit of the mantissa part ay is set to "0".

After approximating the final control amount a and the determination value b with the number of effective bits n as described above, the CPU 12 determines in step 219 whether or not the sign as of the final control amount a is positive. If as = positive, the CPU 12 proceeds to step 220
If as = negative, go to step 225.

In step 220, the CPU 12 compares the exponent part ax of the final control amount a with the exponent part bx of the determination value b to determine whether or not ax> bx. ax> bx
Holds (if step 220 is YES), C
The PU 12 proceeds to step 221, sets the determination result to “true”, and returns to the original routine of FIG.

If ax ≦ bx (step 220)
Is NO), the CPU 12 determines at step 222 that ax =
bx is determined. Then, on condition that ax = bx, the CPU 12 proceeds to step 223,
The mantissa part ay of the final control amount a is compared with the mantissa part by of the determination value b to determine whether or not ay ≧ by. ay ≧
If by is satisfied (YES in step 223), the CPU 12 proceeds to step 221, sets the determination result to "true", and returns to the original routine of FIG.

If ax ≠ bx in step 222,
Alternatively, if ay <by in step 223, the CPU 12
Goes to step 224 and sets the judgment result to “false”.
It returns to the original routine of FIG.

On the other hand, in step 225, the CPU 12
Determines whether ax <bx. If ax <bx holds (if step 225 is YES), CP
U12 proceeds to step 221, sets the determination result to "true", and returns to the original routine of FIG.

If ax ≧ bx (step 225)
Is NO), the CPU 12 determines at step 226 that ax =
bx is determined. Then, on condition that ax = bx, the CPU 12 proceeds to step 227,
It is determined whether or not ay ≦ by. If ay ≦ by is satisfied (YES in step 227), the CPU 1
2 proceeds to step 221, sets the judgment result to "true", and returns to the original routine of FIG.

If ax ≠ bx in the step 226,
Alternatively, if ay> by in step 227, the CPU 12
Goes to step 228 and sets the judgment result to “false”.
It returns to the original routine of FIG.

In this embodiment, steps 207 to 209 in FIG. 4 and steps 213 to 215 in FIG.
Correspond to the approximation means described in the claims, of which steps 207 and 213 are used as the determination means, and steps 208 and 214 are used.
Corresponds to the raising means, and steps 209 and 215 correspond to the clearing means. Step 203 of FIG.
204 corresponds to the first comparing means,
Are the second comparing means, and steps 223 and 227 are the third comparing means.
, Respectively.

According to the embodiment described above, the following effects can be obtained. (A) In the present embodiment, in order to provide a desired resolution, the final control amount a and the determination value
Approximate the mantissa of (floating point type data)
Are compared with each other after approximation. In such a case, even in the case of floating-point data, it is possible to have a His width with the same accuracy as that of the integer data. As a result, in the ECU 10 that handles floating-point data, the data can be appropriately handled within the range of the desired physical quantity accuracy. Further, even when the control amount fluctuates, it is possible to perform the comparison determination with a desired hiss width. For example, a ≧ b
In general-purpose engine control such as turning on / off the purge valve 8 at the time of the above, the problem of the prior art such as the hunting of the purge valve 8 is solved.

(B) Upon approximation of each value a, b, it is determined whether the bit next to the number n of effective bits of the mantissa is “0” or “1”, and the bit next to the number n of effective bits of the mantissa is determined. Is "1", the bit is rounded up by "1", and the bit following the effective bit number n of the mantissa is cleared to "0". According to this configuration, the approximation process of the floating-point data can be easily and accurately performed.

(C) When comparing the values a and b, the sign, the exponent and the mantissa of the floating-point data are compared with each other (however, the comparison of the mantissa is an approximate value). In this case, the comparison of floating-point data can be reliably performed.

The embodiment of the present invention can be realized in the following modes other than the above. In the above-described embodiment, in the processing of the comparison function in FIGS. 4 and 5, the approximation processing is performed for each of the final control amount a and the determination value b, but if the determination value b is a fixed value, the same determination is performed. An approximate value of the value b is calculated in advance. Thereby, the process of approximating the determination value b (FIG. 5)
Steps 213 to 218) can be omitted.

The number n of effective bits is variably set each time according to the type of engine control. For example, when the required control accuracy such as fuel injection control or electronic throttle control is relatively high, the effective bit number n is increased, and conversely, the required control accuracy such as fuel gas purge control is relatively low. In this case, the number n of effective bits is reduced. As a result, it is possible to set an appropriate physical quantity accuracy for the control value while ensuring the control accuracy in any engine control.

The present invention can be applied to other than the above-described control of turning on / off the purge valve (actuator) according to the comparison result of a ≧ b. The point is that if an electronic control unit that introduces floating-point data to improve controllability and increase the operation speed can adjust the accuracy by arbitrarily specifying the effective number of bits, The invention can be applied at any time.

In the above embodiment, the configuration for approximating the floating-point data and the configuration for comparing and judging the approximated data are embodied by the CPU 12, but this is changed. For example, the FPU 15 may be configured to approximate floating-point data and compare and determine the approximate value. In the case of such a configuration, similarly to the above-described embodiment, it is possible to appropriately handle data within a desired range of physical quantity accuracy, and by performing approximation and comparison operations using the FPU 15, High-speed operation can be realized.

In the above embodiment, as shown in FIG. 2, the number of effective bits is designated by counting from the most significant bit (23rd bit) of the mantissa, but the number of effective bits is counted by counting from the least significant bit (1st bit). May be specified. In short, any configuration may be used as long as the upper n bits of the mantissa can be designated as the number of effective bits.

[Brief description of the drawings]

FIG. 1 is a block diagram schematically showing an engine control system according to an embodiment of the present invention.

FIG. 2 is a diagram showing a configuration of floating-point data.

FIG. 3 is a flowchart illustrating a purge control routine.

FIG. 4 is a flowchart illustrating processing of a comparison function.

FIG. 5 is a flowchart showing processing of a comparison function, following FIG. 4;

FIG. 6 is a diagram specifically showing an approximation process of a mantissa part.

FIG. 7 is a diagram for comparing the precision of integer type data and the precision of floating point type data.

FIG. 8 is a diagram for comparing the precision of integer type data and floating point type data in the related art.

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 1 ... Engine, 10 ... Engine control ECU (electronic control device), 11 ... Microcomputer, 12 ... CPU which comprises approximation means and comparison means, 15 ... FPU (floating point arithmetic processor).

Claims (6)

[Claims] 1. An electronic control unit for calculating floating-point data comprising a sign part, an exponent part and a mantissa part, wherein an effective number of bits is designated from the upper part of the mantissa in order to have a desired resolution. And an approximating means for approximating the mantissa in accordance with the designated number of effective bits. 2. An electronic control device for calculating floating-point data comprising a sign part, an exponent part and a mantissa part, wherein an effective number of bits is designated from the upper part of the mantissa in order to have a desired resolution. And approximating means for approximating the mantissa in accordance with the designated number of effective bits; and comparing means for comparing a control value and a predetermined determination value after approximation by the approximating means. Electronic control device. 3. An approximation means comprising: a discrimination means for discriminating whether the next bit of the number of effective bits of the mantissa is “0” or “1”; 3. If the bit is "1", the bit is rounded up by "1", and a clearing unit that clears to "0" bits subsequent to the number of valid bits of the mantissa is set.
An electronic control unit according to claim 1. 4. The comparison means comprises: first comparison means for comparing the sign parts of two data to be compared; and second comparison means for comparing the exponent parts of two data to be compared. 3. The electronic control device according to claim 2, further comprising third comparison means for comparing the mantissa part approximated by the approximation means for the two data to be compared. 5. An electronic control unit for performing various controls of a vehicle engine, wherein the number of effective bits is set according to an engine control type to which an approximation value by the approximation unit is applied. The electronic control device according to claim 1 or 2. 6. An electronic control device having a floating-point arithmetic processor for calculating floating-point data comprising a sign part, an exponent part and a mantissa part, wherein the floating-point arithmetic processor has a desired resolution. Therefore, the number of valid bits is specified from the upper part of the mantissa, the mantissa is approximated according to the specified number of effective bits, and the approximated value is compared with a predetermined determination value. Electronic control device.