JPH06249007A - Driving force control device for vehicle - Google Patents

Abstract

PURPOSE:To control driving force which is suited to a driver's characteristic at all times without influencing on driver's conscious condition and driving environment, and which is reflected disposition of running speedily, and continue its control extending in the all operating range of an acceleration pedal and the like. CONSTITUTION:In a neuro-computer 22, acceleration G of a vehicle 1 is learnt as a teacher data for comparison, and the relation of acceleration G between acceleration stroke S and car speed V is learnt as a demand acceleration model for a driver DR. Throttle sensibility is calculated from the deviation between the demand acceleration model and a second reference acceleration model which is reflected disposition of running of the driver DR and the deviation between the second reference acceleration model and a standard reference acceleration model. In a throttle computer 21, a D.C. motor 8 is driven according to throttle sensibility and acceleration stroke S so as to open/close a throttle valve 7. It is thus possible to control output of an engine 2 so as to control driving force of the vehicle.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a vehicle driving force control device for controlling a driving force so that the vehicle acceleration becomes an acceleration required by a driver.

[0002]

2. Description of the Related Art Generally, a vehicle is required to run under various environmental conditions, and various driving operations are performed by individual drivers. As for the behavior of the vehicle, the responsiveness and smoothness intended by the driver are required. Conventionally, regarding the behavior related to the driving force of the vehicle,
For example, in a vehicle equipped with an internal combustion engine, it is known that control is performed according to the amount of depression of an accelerator pedal by a driver.

For example, in the technique disclosed in Japanese Patent Laid-Open No. 1-294925, the target acceleration intended by the driver is estimated from the amount of depression of the accelerator pedal (accelerator opening) by the driver and the vehicle speed at that time. . Then, the deviation between the estimated target acceleration and the actual acceleration is found, and the opening of the throttle valve is corrected based on the size of the found deviation so that the actual acceleration matches the target acceleration. To be controlled.

However, in the above-mentioned conventional technique, the target acceleration to be estimated is only uniformly set in advance by a map from the relationship between the accelerator opening and the vehicle speed. Therefore, the setting of the target acceleration in the map does not always match the characteristics of the driver, and it is difficult to appropriately control the driving force of the vehicle by matching the characteristics of the individual driver. Moreover, even if the driver is the same, the map for estimating the target acceleration is always the same, so if the driver's consciousness or driving environment changes, the target that the driver can satisfy is changed. Acceleration is not obtained,
The driving performance of the vehicle tended to deteriorate.

Therefore, in addition to addressing the above problems,
A technique aiming at efficient use of a memory in a control device and prevention of delay in calculation of target acceleration has been proposed by the applicant of the present application in Japanese Patent Application No. 3-80103. In this proposed technique, in a vehicle equipped with an internal combustion engine, the opening of a linkless type throttle valve is controlled according to the depression amount (accelerator stroke) of the accelerator pedal by the driver. Here, data for determining the target acceleration corresponding to the accelerator stroke is stored in advance in the backup RAM as a map. The opening of the throttle valve is controlled so that the actual acceleration becomes the target acceleration determined from the map, and thus the driving force of the vehicle is controlled. Further, in this proposed technique, the change in accelerator stroke and the actual acceleration are detected as the change in the degree of acceleration demand of the driver. Then, the map data is modified and stored again in the backup RAM so that the deviation between the detected degree of acceleration demand and the target acceleration determined from the above map is minimized. Mathematically, the map is rewritten by correcting (correcting) the target acceleration according to the above deviation. That is, the data of the target acceleration corresponding to the accelerator stroke is learned.

Therefore, since the target acceleration data is learned so as to match the degree of acceleration required by the driver as described above, the target acceleration that always matches the characteristics of the driver is determined. As a result, regardless of the driver's consciousness and driving environment, the driving force that always matches the driver's characteristics can be obtained.

[0007]

In the latter proposed technique, however, the learning of the target acceleration data is performed only by correcting (correcting) the data and rewriting the map. When rewriting the map, the target acceleration data is only learned for a certain point or a certain range of the accelerator stroke. For example, in the driving region of steady running, the target acceleration data is only learned for the range of the accelerator stroke corresponding to it. Alternatively, in the operating region of the sudden acceleration running, the data of the target acceleration is only learned for the range of the accelerator stroke corresponding to it. Therefore, the target acceleration is only corrected (corrected) only in a specific driving region, and the rewritten map is locally biased. That is, even if the correction (correction) regarding the target acceleration is performed on the specific range of the accelerator stroke, the correction (correction) is not reflected in other ranges of the accelerator stroke. As a result, in the rewritten map, the relationship between the target acceleration and the accelerator stroke may be partially discontinuous, and the control of the driving force of the vehicle may be partially discontinuous with respect to changes in the accelerator stroke. there were.

Further, in the latter proposed technique, it takes some time to learn the acceleration required by the driver. That is, when the driver is replaced with respect to a specific vehicle, if the driver's driving orientation is different before and after the replacement, the driver's orientation after the replacement can be immediately reflected in the vehicle driving. There was a tendency that I could not. For example, when a driver who always tends to drive faster is replaced by a driver who always tends to drive slowly,
There was a tendency for some delay before the driver's orientation after the replacement was reflected in the running of the vehicle.

The present invention has been made in view of the above-mentioned circumstances, and an object thereof is to always control a driving force that matches the characteristics of a driver regardless of the driver's consciousness and driving environment. It is possible to control the driving force that more quickly reflects the driving direction (preference) of each driver, and the driving force can be controlled by the driver's operation amount of the accelerator pedal or the like. It is an object of the present invention to provide a driving force control device for a vehicle which can be continuous over the entire range.

[0010]

In order to achieve the above object, in the present invention, as shown in FIG.
Control amount changing means M3 for changing the control amount of the drive source M2 mounted on the drive unit 1, output operation means M4 operated by the driver to arbitrarily control the output of the drive source M2,
An operation amount detecting means M5 for detecting the operation amount of the output operation means M4 is provided, and the output of the drive source M2 is controlled by driving the control amount changing means M3 according to the detection result of the operation amount detecting means M5. In the vehicle driving force control device configured to control the driving force of the vehicle M1, the acceleration detecting means M6 for detecting the acceleration of the vehicle M1, and the vehicle M
The speed detection means M7 for detecting the speed of No. 1 and the vehicle M with respect to the operation amount of the output operation means M4 and the speed of the vehicle M1.
1 standard pre-learned about the relationship of acceleration of 1
The first reference acceleration model storage unit M8 for storing the reference acceleration model of No. 1, the relationship between the operation amount of the output operation unit M4 and the acceleration of the vehicle M1 with respect to the speed of the vehicle M1 is learned, and the driver's running orientation is directed. The second reference acceleration model storage unit M9 for storing the second reference acceleration model learned about (preference) and the teacher data as the teacher data to be compared with the acceleration obtained by the detection of the acceleration detection unit M6. And the output of the means M10 is used as an error signal, and the operation amount of the output operation means M4 and the speed detection means M7 are set so that the error amount becomes small.
The required acceleration model learning means M10 for learning the relationship of the acceleration of the vehicle M1 with respect to the speed obtained by the detection as the acceleration model required by the driver, and the second reference acceleration model storage means M9. The deviation between the second reference acceleration model and the first reference acceleration model stored in the first reference acceleration model storage means M8 is calculated, and the output of the acceleration model learned by the required acceleration model learning means M10 and A deviation from the first reference acceleration model stored in the first reference acceleration model storage means M8 or the second reference acceleration model stored in the second reference acceleration model storage means M9 is calculated, and each deviation is calculated. The drive source M2 for the speed obtained by the operation amount of the output operation means M4 and the detection of the speed detection means M7 based on
Control amount sensitivity calculation means M11 for calculating the relationship of the control amount as the control amount sensitivity, and the control amount sensitivity calculation means M1.
It is intended to include drive control means M12 for controlling the drive of the control amount changing means M3 according to the operation amount detected by the operation amount detecting means M5 based on the control amount sensitivity calculated by 1.

[0011]

According to the above construction, as shown in FIG.
The reference acceleration model storage means M8 of the output operation means M
The standard first reference acceleration model learned in advance regarding the relationship between the operation amount of No. 4 and the acceleration of the vehicle M1 with respect to the speed of the vehicle M1 is stored. Further, in the second reference acceleration model storage means M9, the relationship between the operation amount of the output operation means M4 and the acceleration of the vehicle M1 with respect to the speed of the vehicle M1 is learned, and the orientation (preference) for the driver's running is learned. The second reference acceleration model is stored. Further, the acceleration obtained by the detection of the acceleration detecting means M6 is calculated by the required acceleration model learning means M10.
Used as teacher data for comparison. That is, the actual acceleration of the vehicle M1 is used for comparison as the acceleration requested by the driver. Then, in the required acceleration model learning means M10, the deviation between the teacher data and the output of the means M10 is reduced, that is, the deviation between the acceleration required by the driver and the output of the required acceleration model is reduced. , Operation amount of output operation means M4 and speed detection means M7
The relationship of the acceleration of the vehicle M1 with respect to the speed obtained by the detection of is learned as an acceleration model required by the driver. Further, the control amount sensitivity calculating means M11 calculates the deviation between the second reference acceleration model and the first reference acceleration model, and outputs the learned acceleration model and the first reference acceleration model or the second reference acceleration model. The deviation from the reference acceleration model is calculated. Similarly, in the control amount sensitivity calculating means M11, the relationship between the operation amount of the output operating means M4 and the control amount of the drive source M2 with respect to the speed obtained by the detection of the speed detecting means M7 is calculated as the control amount sensitivity on the basis of the respective deviations. . Then, in the drive control means M12, based on the control amount sensitivity calculated as described above, the drive of the control amount changing means M3 is controlled according to the operation amount of the output operation means M4 by the driver's operation. As a result, the output of the drive source M2 is controlled, and thus the driving force of the vehicle M1 is controlled.

Therefore, according to the present invention, the acceleration model that always meets the driver's request is estimated. The deviation between the output of the acceleration model and the standard first reference acceleration model or the second reference acceleration model reflecting the driver's driving direction (preference), and the driver's driving direction (preference) 2nd reference acceleration model reflecting the standard 1st
The control amount sensitivity is obtained according to the deviation from the reference acceleration model of. Therefore, from the former deviation, the strength of the driver's demand for acceleration at that time is reflected in the control amount sensitivity, and from the latter deviation, the driver's normal driving direction (preference) is strong. Is reflected in the control amount sensitivity. Based on the control amount sensitivity, the control amount of the drive source M2 is controlled with an acceleration that always meets the driver's request and with an acceleration that reflects the driver's driving direction (preference).

Further, according to the present invention, the output operating means M
4, the entire relation of the acceleration of the vehicle M1 to the operation amount of the vehicle M1 and the speed of the vehicle M1 is learned as a continuous model, and the entire relation of the operation amount of the output operation means M4 and the control amount of the drive source M2 to the speed of the vehicle M1 Is learned as a continuous model. Therefore, in the entire range of the speed of the vehicle M1, the relationship between the acceleration and the control amount with respect to the entire range of the operation amount of the output operation means M4 does not become partially discontinuous.

[0014]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment in which a vehicle driving force control device according to the present invention is embodied will be described in detail below with reference to FIGS.

FIG. 2 is a schematic diagram showing a driving force control device for a vehicle in this embodiment. A vehicle 1 is equipped with a gasoline engine (hereinafter simply referred to as “engine”) 2 as a drive source. The engine 2 is of a multi-cylinder in-line type. Outside air is taken into the intake passage 3 of the engine 2 and fuel injected from an injector (not shown) is supplied. Then, a mixture of outside air and fuel is taken into each combustion chamber (not shown) of the engine 2 through the intake passage 3. Further, the air-fuel mixture taken into each combustion chamber is exploded and burned by the operation of each spark plug (not shown), whereby the piston (not shown), the crankshaft, etc. are operated and the output of the engine 2 is obtained. Further, the burnt gas after being burned in each combustion chamber of the engine 2 is discharged to the outside through the exhaust passage 4.

In this embodiment, the vehicle 1 is of a front engine / rear drive system (FR system), and the crankshaft of the engine 2 is provided with a transmission, a propeller shaft, a differential gear, a drive shaft, etc., which are not shown. It is drivingly connected to a pair of left and right rear wheels 5, which are driving wheels. The pair of left and right front wheels 6, which are driven wheels, are steering wheels that interlock with the operation of a steering wheel (not shown) provided in the driver's seat.

In this embodiment, a linkless type throttle valve 7 constituting a control amount changing means is provided in the middle of the intake passage 3. That is, the throttle valve 7 is connected to the DC motor 8 provided in the vicinity thereof. Then, the operation of the DC motor 8 controls the opening of the throttle valve 7 as the control amount of the engine 2, that is, the throttle opening Th. As a result, the amount of air taken into each combustion chamber of the engine 2 through the intake passage 3 is adjusted, and the output of the engine 2 is controlled by adjusting this amount of air.

A throttle sensor 9 is provided near the throttle valve 7. This throttle sensor 9
Then, the throttle opening Th is detected and a signal corresponding thereto is output. The driver's seat of the vehicle 1 is provided with an accelerator pedal 10 as an output operating means. The accelerator pedal 10 is operated by the driver DR in order to arbitrarily control the output of the engine 2.
Further, an accelerator sensor 11 as an operation amount detecting means is provided near the accelerator pedal 10. The accelerator sensor 11 detects an operation amount of the accelerator pedal 10, that is, an accelerator stroke S, and outputs a signal corresponding to the detected amount. Further, a well-known acceleration sensor 12 which constitutes an acceleration detecting means is provided at substantially the center of the vehicle 1. The acceleration sensor 12 detects the acceleration G in the front-rear direction of the vehicle 1 and outputs a signal corresponding thereto.
In addition, the front wheel 6 is provided with a known vehicle speed sensor 13 which constitutes a speed detecting means. In the vehicle speed sensor 13, the speed of the vehicle 1, that is, the vehicle speed V is determined according to the rotation speed of the front wheels 6.
Is detected and the corresponding signal is output.

In addition, in this embodiment, the driver's seat of the vehicle 1 is provided with a mode select switch 14 which is operated to select a preferred mode for the driving of the driver DR, that is, a directivity mode DM. . In this embodiment, as the directional mode DM for running, "high mode (H)" that directs high-acceleration running, "low mode (L)" that directs low-acceleration running, and standard acceleration running are directed. The three normal modes DM of “normal mode (N)” are set. Then, in order to select each of the pointing modes DM, the mode select switch 1
When the switch 4 is operated, the switch 14 outputs a signal instructing the directivity mode DM according to the selection.

Further, in this embodiment, a throttle computer 21 and a neuro computer 22 are provided in order to suitably control the opening / closing of the throttle valve 7 in accordance with the demand and direction (preference) of the driver DR. The throttle computer 21 constitutes drive control means, and the DC motor 8 and the throttle sensor 9 are electrically connected to the throttle computer 21, respectively.
Further, the neuro computer 22 constitutes a first reference acceleration model storage means, a second reference acceleration model storage means, a required acceleration model learning means and a control amount sensitivity calculation means, and is constructed by applying a neural network technique. Has been done. An accelerator sensor 11, an acceleration sensor 12, a vehicle speed sensor 13, and a mode select switch 14 are electrically connected to the neuro computer 22. Further, the neuro computer 22 and the throttle computer 21 are electrically connected to each other.

FIG. 3 is a block diagram showing the electrical configuration of the throttle computer 21 and the neuro computer 22. The neuro computer 22 includes a central processing unit (CPU) 23, a read-only memory (ROM) 24 that stores a predetermined learning control program and the like in advance, and a CPU 23.
Random access memory (RAM) 25 for temporarily storing the calculation results and the like, backup RAM 26 for storing previously stored data, and the like. The neuro computer 22 is configured as a logical operation circuit in which the respective units 23 to 26, the external input / output circuit 27 and the like are connected by a bus 28. The accelerator sensor 11, the vehicle speed sensor 13, and the mode select switch 14 described above are connected to the external input / output circuit 27, respectively. or,
The acceleration sensor 12 is connected to the external input / output circuit 27 via a low-pass filter 29. The low-pass filter 29 freely passes a signal having a frequency lower than a predetermined cutoff frequency serving as a reference among the detection signals of the acceleration sensor 12, and gives a large attenuation to a high frequency. In addition, the above-mentioned throttle computer 21 is connected to the external input / output circuit 27. Also, RO
In M24, a learning control program using the neural network technology is stored in advance.

The CPU 23 uses the external input / output circuit 2
Various signals from each of the sensors 11 to 13 and the mode select switch 14 input via 7 or the like are read as input values. CPU23, based on those input values, ROM
According to the learning control program stored in 24, the learning control of the “requested acceleration model” requested by the driver DR is executed. Further, the CPU 23 calculates a deviation between the learning result and the “second reference acceleration model” selected according to the above-described directivity mode DM, and similarly the “second reference acceleration model” and the standard “standard reference acceleration model”. The deviation from the “first reference acceleration model” is calculated. The CPU 23 calculates the throttle sensitivity as the control amount sensitivity from the respective deviations. And CP
U23 outputs the learning result and the like to the throttle computer 21 via the external input / output circuit 27. Here, the “first reference acceleration model” and the “second reference acceleration model” are learned in advance and stored in the ROM 24.

On the other hand, the throttle computer 21 basically has the same structure as the neuro computer 22, and is composed of a CPU 30, a ROM 31, a RAM 32, a backup RAM 33, an external input / output circuit 34, a bus 35 and the like. The aforementioned DC motor 8, throttle sensor 9 and neuro computer 22 are connected to the external input / output circuit 34, respectively. Also ROM3
In FIG. 1, a throttle opening control program for controlling opening / closing of the throttle valve 7 based on a learning result of the neuro computer 22 is stored in advance.

Then, the CPU 30 reads data such as learning results input from the neuro computer 22 via the external input / output circuit 34 as an input value. Also, CPU
30 reads the signal from the throttle sensor 9 as an input value. Also, the CPU 30 uses the input values to
The DC motor 8 is preferably controlled according to the throttle opening control program stored in the ROM 31.

Here, the conceptual configuration of the neural network technology applied to the neurocomputer 22 will be described with reference to FIGS. The neural network in this embodiment includes two multilayer neural networks, as shown in FIGS. Each multi-layered neural network basically has the same configuration, and is an "input layer" composed of two neurons n1.
And an "intermediate layer" consisting of 2 to 10 neurons n2
And an "output layer" composed of one neuron n3. In addition, each neuron n1, n2, n3 between each layer
Are bound by synapses sp. In each multilayer neural network, signals flow in one direction from the “input layer” to the “intermediate layer” and from the “intermediate layer” to the “output layer”. In the neurons n1, n2, n3 of each layer, the state is determined based on the signal received from the previous layer, and the signal is transmitted to the next layer. Then, the output result of each multilayer neural network is the neuron n in the “output layer”.
It is obtained as a state value of 3.

Here, the multilayer neural network shown in FIG. 4 is learned and updated from time to time, and the accelerator sensor 1 is connected to each neuron n1 of the "input layer".
The accelerator stroke S detected by 1 and the vehicle speed V detected by the vehicle speed sensor 13 are input.
Further, the output obtained from the neuron n3 in the "output layer", that is, the required acceleration model output Gx, is compared with the "teacher data" by using the acceleration G of the vehicle 1 obtained by the detection of the acceleration sensor 12 as "teacher data". It Then, the acceleration deviation ΔG (= G−Gx) obtained by the comparison is used as an “error signal”, and the “weighting coefficient” of the synapse sp of all the neurons n1, n2, n3 is corrected in the direction in which the error becomes smaller. That is, the acceleration G of the vehicle 1 is set as the acceleration requested by the driver DR, and the acceleration G is set as "teacher data" to be compared. Then, the relationship between the accelerator stroke S and the acceleration G with respect to the vehicle speed V is learned as a “requested acceleration model” required by the driver DR so that the deviation from the “teacher data” becomes small. The output result of this multilayer neural network is obtained as the required acceleration model output Gx. That is, the “requested acceleration model” is learned as a characteristic as shown in FIG. 9 in the direction in which the requested acceleration model output Gx approaches the acceleration G.

On the other hand, in the multilayer neural network shown in FIG. 5, the standard (or “average”) relationship between the accelerator stroke S, the vehicle speed V, and the acceleration G is “first” based on the traveling data of a plurality of drivers DR. It is uncorrectable data learned in advance as “1 reference acceleration model”. In this multilayer neural network, the accelerator stroke S and the vehicle speed V are input to each neuron n1 in the "input layer" as described above. And "output layer"
From the neuron n3, the first reference acceleration model output Gs1 corresponding to the accelerator stroke S and the vehicle speed V is obtained as an output result. That is, the "first reference acceleration model" is learned as the characteristic shown in FIG.

Each of the multi-layered neural networks shown in FIGS. 6 to 8 is based on the traveling data of a plurality of drivers DR, and various relations between the accelerator stroke S and the vehicle speed V and the acceleration G with respect to the traveling of the driver DR. This is uncorrectable data that has been learned in advance as a “second reference acceleration model” reflecting the orientation. The multi-layer neural network of FIG. 6 is one in which “high mode (H)” is set among the various directional modes DM described above, and the “input layer” thereof is set.
Similarly to the above, the accelerator stroke S and the vehicle speed V are input to each neuron n1. Then, from the neuron n3 of the "output layer", as a result of output, a second reference acceleration model output Gs2 (H) reflecting a high acceleration running according to the accelerator stroke S and the vehicle speed V is obtained.

The multilayer neural network shown in FIG.
The "normal mode (N)" is set among the various directional modes DM described above, and the accelerator stroke S, the same as the above, is applied to each neuron n1 of the "input layer".
The vehicle speed V is input respectively. Then, from the neuron n3 of the "output layer", as the output result, the second reference acceleration model output Gs2 (N) reflecting the standard acceleration running according to the accelerator stroke S and the vehicle speed V is obtained.

The multilayer neural network shown in FIG.
The "low mode (L)" is set among the above-mentioned pointing modes DM, and each neuron n in the "input layer" is set.
The accelerator stroke S and the vehicle speed V are input to 1 as described above. Then, the neuron n in the “output layer”
From FIG. 3, a second reference acceleration model output Gs2 (L) reflecting a low acceleration running according to the accelerator stroke S and the vehicle speed V is obtained as an output result.

That is, the "second reference acceleration model" reflecting these various directional modes DM is learned in advance as various characteristics as shown in FIG. In the following description, each second reference acceleration model output Gs2 (H),
When Gs2 (N) and Gs2 (L) are collectively described, the second reference acceleration model output Gs2 is used for convenience.

The conceptual structure of the neural network as described above has been described for convenience only, and the substance of the neural network is the learning control program stored in the ROM 24 of the neurocomputer 22 in advance. The neural network is based on a mathematical operation in the learning control program, and a generally known "error back propagation learning algorithm" is applied as a learning method.
In this embodiment, a learning control program is created in order to finally obtain the characteristic of the relationship of the throttle sensitivity Thg with respect to the accelerator stroke S as shown in FIG.

Next, the processing operation for learning the "requested acceleration model" and the like executed by using the neural network technology as described above in the neuro computer 22 will be described. FIG. 13 is a flowchart showing a "learning control routine" of the learning control program executed by the neurocomputer 22. After the processing of this routine is started, a fixed period, for example, "0.1
It is executed periodically with a time interval of "second".

When the processing of this routine is started, at step 101, the accelerator sensor 11, the acceleration sensor 12, the vehicle speed sensor 13 and the mode select switch 1 are selected.
The accelerator stroke S, the acceleration G, the vehicle speed V, and the pointing mode DM are read based on various signals from 4.

Next, at step 102, the required acceleration model output Gx is determined from the accelerator stroke S and the vehicle speed V read this time. That is, using the accelerator stroke S and the vehicle speed V as input values, the required acceleration model output Gx is obtained from the characteristics of the already learned "requested acceleration model" (see FIG. 9).

Next, at step 103, the first reference acceleration model output Gs1 is determined from the accelerator stroke S and the vehicle speed V read this time. That is, the characteristics of the “first reference acceleration model” that has been learned in advance using the accelerator stroke S and the vehicle speed V as input values (see FIG. 10).
From this, the first reference acceleration model output Gs1 is obtained.

Further, in step 104, the accelerator stroke S, the vehicle speed V and the pointing mode D which have been read this time are read.
The second reference acceleration model output Gs2 is determined from M.
That is, the second reference acceleration model for “high mode (H)” or the second reference acceleration model for “normal mode (N)” that has been learned in advance corresponding to the designated pointing mode DM, or Select the second reference acceleration model for "low mode (L)". Then, using the accelerator stroke S and the vehicle speed V as input values for the selected second reference acceleration model, the second reference acceleration model output Gs2 is obtained from the characteristics of the second reference acceleration model (see FIG. 11).
(H), or the second reference acceleration model output Gs2
(N), or the second reference acceleration model output Gs2
(L) is obtained.

Further, in step 105, the throttle sensitivity Thg is determined from the required acceleration model output Gx obtained this time, the first reference acceleration model output Gs1 and the second reference acceleration model Gs2. That is, the throttle sensitivity Thg is determined according to the following calculation formula (1).

Thg = α + (Gs2-Gs1) * K0 + (Gx−Gs2) * K (1) Here, “α” is a reference value, and in this embodiment, “α =
It is 1.0 ". Further, “K” and “K0” are positive constants, respectively, and have a relationship of “K> K0”.

For example, as shown by the solid line in FIG. 14, with respect to the characteristic of the standard first reference acceleration model output Gs1,
It is assumed that the characteristic of the second reference acceleration model output Gs2 determined by the directional mode DM (in this case, “high mode (H)”) selected at that time is shown by a broken line. At this time, the second stroke for a certain accelerator stroke S
The deviation (Gs2-Gs1) between the reference acceleration model output Gs2 and the first reference acceleration model output Gs1 is reflected in the throttle sensitivity Thg. That is, driver D
The strength of the normal running direction of R, that is, the high-acceleration running direction is reflected in the throttle sensitivity Thg.

Further, it is assumed that the characteristic of the current required acceleration model output Gx determined by learning is shown by a two-dot chain line with respect to the characteristic of the second reference acceleration model output Gs2 shown by the broken line in FIG. . At this time, the deviation (Gx-Gs2) between the required acceleration model output Gx and the second reference acceleration model output Gs2 for a certain accelerator stroke S.
Will be reflected in the throttle sensitivity Thg. That is, the strength of the driver DR's request for the acceleration G at that time (the strength of the request for higher acceleration than in the “high mode (H)” in this case) is reflected in the throttle sensitivity Thg. .

Then, in step 106, the throttle sensitivity Thg and the accelerator stroke S which are determined this time are set.
And are output to the throttle computer 21. Alternatively,
The product of throttle sensitivity Thg and accelerator stroke S,
That is, the target throttle opening Thg · S is obtained, and the target throttle opening Thg · S is set to the throttle computer 21.
Output to.

Then, in step 107, the vehicle 1
The learning of the “requested acceleration model” requested by the driver DR is executed using the acceleration G of “1” as “teacher data”. That is,
The “teacher data” to be compared with the acceleration G of the vehicle 1 detected by the acceleration sensor 12
The relationship between the accelerator stroke S and the acceleration G with respect to the vehicle speed V is learned as a “requested acceleration model” requested by the driver DR so that the deviation between the acceleration and the vehicle speed V becomes smaller.

For example, it is assumed that the curve shown by the solid line in FIG. 9 is the characteristic of the current "requested acceleration model". Then, in order for the driver DR to drive faster than at present, the accelerator pedal 10
It is assumed that the acceleration G of the vehicle 1 has become larger than the current required acceleration model output Gx by operating. The acceleration G at this time is the new required acceleration,
Is updated to a characteristic like the curve shown by the broken line in FIG. That is, the entire relationship of the required acceleration model output Gx with respect to the accelerator stroke S and the vehicle speed V is learned as a continuous model. And this characteristic is not partially discontinuous.

Incidentally, FIG. 9 shows the characteristic when the vehicle speed V is "0", but in the "requested acceleration model", the entire range of the accelerator stroke S and the entire range of the vehicle speed V and the acceleration of the vehicle 1 are shown. The relationship with G has been learned.

Then, after the processing of step 107 is executed, the subsequent processing is once terminated and "0.1
When “second” has elapsed, the process starts again from step 101.

In this way, the learning control process using the neural network technique is executed, and the driver DR
The characteristics of the "requested acceleration model" required by are learned.
Here, the "requested acceleration model" that is learned from time to time
The “weight coefficient” of the synapse sp as the characteristic of is rewritten and stored in the backup RAM 26.

The initial value of the "weighting coefficient" of the "requested acceleration model" when the vehicle 1 is shipped from the factory is that of the "first reference acceleration model". Next, the processing operation of the throttle opening control executed by the throttle computer 21 based on the throttle sensitivity Thg determined by the above processing operation and the accelerator stroke S at that time will be described. FIG. 15 is a flowchart showing a “throttle opening control routine” of the throttle opening control program executed by the throttle computer 21. After the processing of this routine is started, it is periodically executed at predetermined time intervals.

When the processing of this routine is started, first, at step 201, the throttle opening Th is read based on the signal from the throttle sensor 9, and the latest throttle sensitivity Thg and accelerator stroke S output from the neuro computer 22 are read. Alternatively, the target throttle opening Thg · S is read. Here, if it is premised that the throttle sensitivity Thg and the accelerator stroke S are read, in Step 201, both Th
The product of g and S is obtained as the target throttle opening Thg · S.

Next, at step 202, it is judged if the current throttle opening Th is smaller than the target throttle opening Thg · S. Here, when the throttle opening Th is smaller than the target throttle opening Thg · S, in step 203, the DC motor 8 is normally rotated so as to drive the throttle valve 7 in the opening direction.
Further, in step 204, the throttle opening Th is read based on the signal from the throttle sensor 9.

Then, at step 205, it is judged if the throttle opening Th is smaller than the target throttle opening Thg · S. Here, when the throttle opening Th is smaller than the target throttle opening Thg · S,
To jump to step 203 and drive the throttle valve 7 further in the opening direction, steps 203, 20
The process of No. 4,205 is repeated. On the other hand, when the throttle opening Th is equal to or larger than the target throttle opening Thg · S, it is determined that the throttle valve 7 is not driven in the opening direction any more, and the subsequent processing is temporarily terminated.

On the other hand, in step 202, when the current throttle opening Th is not smaller than the target throttle opening Thg · S, the routine proceeds to step 206, where the throttle opening Th is the target throttle opening Thg · S. Is greater than or equal to. Here, the throttle opening T
When h is not larger than the target throttle opening Thg · S, that is, when “Th ≦ Thg · S”, the subsequent processing is temporarily terminated.

If the throttle opening Th is larger than the target throttle opening Thg · S at step 206, the DC motor 8 is rotated in reverse so as to drive the throttle valve 7 in the closing direction at step 207. In step 208, the throttle sensor 9
The throttle opening Th is read based on the signal from.

Then, in step 209, it is determined whether or not the throttle opening Th is larger than the target throttle opening Thg · S. Here, when the throttle opening Th is larger than the target throttle opening Thg · S,
To jump to step 207 and drive the throttle valve 7 in the closing direction further, steps 207, 20
The process of 8,209 is repeated. On the other hand, when the throttle opening Th is equal to or smaller than the throttle sensitivity Thg, the throttle valve 7 is not driven further in the closing direction, and the subsequent processing is temporarily ended.

In this way, the DC motor 8 is adjusted so that the throttle opening Th matches the target throttle opening Thg · S.
The rotation of the throttle valve 7
Is controlled to open and close. As a result, the output of the engine 2 is controlled, and as a result, the driving force of the vehicle 1 is controlled.

As described above, in this embodiment, the request of the driver DR for the running of the vehicle 1 is estimated as the "requested acceleration model" from the acceleration G at that time.
Further, the deviation between the required acceleration model output Gx obtained from the “requested acceleration model” and the second reference acceleration model output Gs2 reflecting the driving direction of the driver DR obtained from the “second reference acceleration model”. (Gx-Gs2) is calculated. In addition, the second reference acceleration model output Gs2
And a deviation (Gs2-Gs) from the standard first reference acceleration model Gs1 obtained from the "first reference acceleration model".
1) is required. Further, according to the deviation (Gx-Gs2) and the deviation (Gs2-Gs1), the throttle sensitivity Thg
Is determined. Then, the opening / closing control of the throttle valve 7 is performed so that the target throttle opening Thg · S obtained from the product of the determined throttle sensitivity Thg and the accelerator stroke S matches the throttle opening Th. That is, according to this embodiment, the throttle opening Th of the engine 2 is controlled with the acceleration G that always meets the request of the driver DR and with the acceleration G that reflects the driving direction of the driver DR. Therefore, from the former deviation (Gx-Gs2), the strength of the driver DR's request for the acceleration G at that time is reflected in the control amount sensitivity. Therefore, the acceleration G requested by the driver DR at that time, that is, the requested acceleration model output Gx is larger than the characteristic of the second reference acceleration model output Gs2 corresponding to the directional mode DM selected at that time. At times, the throttle sensitivity Thg becomes relatively large. Therefore, the change range of the accelerator stroke S for obtaining the same acceleration G becomes relatively narrow, and the accelerator pedal 10
A large acceleration G can be obtained by a small number of operations, and the driver DR can feel that the acceleration performance of the vehicle 1 is improved. For example, driver DR
When the user wants to drive the vehicle 1 quickly due to a consciousness of urgency or the driving environment of the vehicle 1 is an expressway with no traffic congestion, a large acceleration G can be obtained by operating the accelerator pedal 10 less. Therefore, the feeling of acceleration can be improved.

On the other hand, when the required acceleration model output Gx at that time is smaller than the characteristic of the second reference acceleration model output Gs2 corresponding to the directional mode DM selected at that time, the throttle sensitivity Thg is It becomes relatively small. Therefore, the change range of the accelerator stroke S for obtaining the same acceleration G becomes relatively wide, and the acceleration G can be delicately changed by many operations of the accelerator pedal 10, and the accelerator pedal for the driver DR. The operating performance of 10 can be improved. For example, when the driver DR is in a relaxed state of consciousness, or the driving environment of the vehicle 1 is a congested road, a snowy road, or the like, and when it is desired to drive the vehicle 1 slowly, it is necessary to operate the accelerator pedal 10 a lot. The acceleration G can be changed subtly, and the operation feeling of the vehicle 1 can be improved.

That is, in this embodiment, since the learning is performed so as to match the degree of demand for the acceleration G of the driver DR, the throttle sensitivity Thg that always matches the characteristics of the driver DR is determined. As a result, the driver's DR consciousness (hurrying, relaxing, etc.) and driving environment (road condition, day / night, tunnel, rainy road, snowy road,
Regardless of mountain road, traffic jam, etc.)
It is possible to always control the driving force that matches the characteristics of the driver DR.

In addition, the latter deviation (Gs2-Gs)
From 1), the directional strength of the normal running of the driver DR is reflected in the throttle sensitivity Thg. That is,
The directional strength corresponding to the directional mode DM of the “high mode (H)”, the “normal mode (N)”, or the “low mode (L)” is selectively reflected on the throttle sensitivity Thg. Therefore, when the characteristic of the second reference acceleration model output Gs2 corresponding to the directional mode DM selected at that time is larger than the characteristic of the standard first reference acceleration model output Gs1, the throttle sensitivity T
hg becomes relatively large. As a result of the increase, it becomes possible to obtain a large acceleration G by operating the accelerator pedal 10 less, and the acceleration performance of the vehicle 1 is improved by reflecting the driving direction of the driver DR. It can make the driver DR feel.

On the other hand, compared to the characteristic of the standard first reference acceleration model output Gs1, the second reference acceleration model output Gs corresponding to the directional mode DM selected at that time.
When the characteristic of 2 is smaller, the throttle sensitivity Thg
Becomes relatively small. The acceleration G can be delicately changed by a large number of operations of the accelerator pedal 10 by the reduced amount, and the driver DR's driving direction is reflected to reflect the accelerator pedal 10 of the driver DR. The operation performance can be improved.

Therefore, the driver DR, who always tends to drive the vehicle 1 at high speed, selects the "high mode (H)" by the mode select switch 14 so that the vehicle 1
It is possible to immediately set the running of the car to the running that matches the direction of high acceleration. Similarly, the driver DR, who always tends to run the vehicle 1 slowly, selects the "low mode (L)" by the mode select switch 14 to make the running of the vehicle 1 match the direction of low acceleration. It can be set immediately. Alternatively, the driver DR, who always aims to run the vehicle 1 as a standard, selects the “normal mode (L)” by the mode select switch 14 so that the running of the vehicle 1 matches the orientation of medium acceleration. It becomes possible to immediately set the running.

In other words, in this embodiment, the driving force of the engine 2 can be controlled in a manner that always matches the characteristics of the driver DR regardless of the driver's DR consciousness and the driving environment. The driving force of the engine 2 can be controlled in such a manner as to more rapidly reflect the driving tendency of the driver DR.

Further, in this embodiment, since the neural network technique is used for the learning control in the neuro computer 22, the entire relation of the required acceleration model output Gx with respect to the accelerator stroke S and the vehicle speed V is learned as a continuous model. However, the characteristics are not partially discontinuous. This is because the "requested acceleration model" learned between the discontinuity points of the accelerator stroke S and the vehicle speed V is interpolated by using the neural network technology. That is, the correction of the required acceleration model output Gx performed in a specific range of the accelerator stroke S and the vehicle speed V is reflected in the correction of the required acceleration model output Gx corresponding to other ranges of the accelerator stroke S and the vehicle speed V. Of.

As a result, the control of the driving force of the vehicle 1 is controlled by the driver DR over the entire range of the vehicle speed V.
The operation amount of 0, that is, the operation amount of the accelerator stroke S can be made continuous over the entire operation range. Therefore, when the driver DR continuously depresses the accelerator pedal 10, the acceleration G of the vehicle 1 does not suddenly change, and the increase in the vehicle speed V can be made smooth at all times.

Further, in this embodiment, the throttle sensitivity Thg is determined from the deviation between the required acceleration model output Gx estimated from the actual acceleration G and the reference acceleration model output Gs. Therefore, as compared with the case where the map is rewritten by the correction (correction) as in the conventional technique, the map interpolation calculation is unnecessary and the calculation time can be further shortened.

In addition, in this embodiment, in the neuro computer 22, the acceleration sensor 12 is connected to the external input / output circuit 27 via the low pass filter 29. Therefore, when noise is generated in the detection signal of the acceleration sensor 12 due to the harshness when the vehicle 1 is traveling on an uneven road, the low-pass filter 29 attenuates the high-frequency component correlated with the noise. It That is,
The signal of the acceleration G from the acceleration sensor 12 is shown in FIG.
Even if a large noise due to the harshness is included as shown in (a), the signal passes through the low-pass filter 29, so that the signal of the acceleration G with less noise as shown in FIG. 16 (b) is obtained. Will be adjusted.

As a result, in the neurocomputer 22, the signal of the acceleration G used for learning can be a signal from which noise due to harshness has been removed. Therefore, erroneous learning about the “requested acceleration model” can be prevented in advance, and in turn, the throttle sensitivity Thg can be prevented from being adjusted in the wrong direction.

The present invention is not limited to the above-described embodiment, but may be implemented as follows with a part of the structure appropriately changed without departing from the spirit of the invention. (1) In the above embodiment, the gasoline engine 2 is used as the drive source and the linkless type throttle valve 7 is used as the control amount changing means, but it can be embodied as other drive source and control amount changing means. . For example, in an electric vehicle, an electric motor such as a DC motor may be used as a drive source, and a current control circuit that controls the current to the electric motor may be used as the control amount changing means.

(2) In the above embodiment, the accelerator pedal 10 is used as the output operating means operated by the driver DR, but an accelerator lever or other operating member may be used as the output operating means.

(3) In the above embodiment, the accelerator sensor 1
Although the accelerator stroke S is detected by using 1 as the operation amount detection means, the following may be performed. That is, instead of using the accelerator stroke S, a sensor for detecting the accelerator pedal force is used, or an accelerator sensor for detecting the accelerator stroke S and a sensor for detecting the accelerator pedal force are used together.

(4) Although the acceleration sensor 12 is used as the acceleration detecting means in the above embodiment, the acceleration G is obtained by differentiating the vehicle speed V detected by the vehicle speed sensor 13 with respect to the vehicle speed sensor 13 as the acceleration detecting means. You may get it.

(5) In the above embodiment, as the neural network technology in the neuro computer 22,
Although a multi-layered neural network is adopted, an interconnected neural network can also be adopted.

(6) In the above embodiment, "K>K0" is used in the calculation formula (1), but "K <K0" or "K =
It may be "K0". When “K <K0” is set, it is possible to set the specification in which various directions (preferences) of the driver DR are reflected more strongly (largely) on the throttle sensitivity Thg.

(7) In the above embodiment, the throttle sensitivity Thg is calculated according to the calculation formula (1), but the throttle sensitivity Thg can be calculated according to the following calculation formula (2).

Thg = α + {(Gs2 / Gs1) -1.0} * K0 + {(Gx / Gs2) -1.0} * K (2) (8) In the above embodiment, the required acceleration model output Gx. Between the second reference acceleration model output Gs2 and (Gx-Gs
2) was obtained and used for calculation of the throttle sensitivity Thg. On the other hand, the deviation (Gx-Gs1) between the required acceleration model output Gx and the first reference acceleration model output Gs1
May be used to calculate the throttle sensitivity Thg. Then, the throttle sensitivity Thg can be obtained according to the following calculation formulas (3) and (4).

Thg = α + (Gs2-Gs1) * K0 + (Gx-Gs1) * K (3) Thg = α + {(Gs2 / Gs1) -1.0} * K0 + {(Gx / Gs1) -1. 0} * K (4) Even in this case, the same operation and effect as those of the above-described embodiment can be obtained.

(9) In the above embodiment, the standard "first reference acceleration model" learned in advance is stored in the ROM of the neuro computer 22 in the form of a multilayer neural network.
However, the “first reference acceleration model” learned in advance may be stored in the ROM 24 of the neurocomputer 22 in the form of a map.

(10) In the above embodiment, the "second reference acceleration model" reflecting various directions of the driver DR for running is stored in the ROM 24 of the neurocomputer 22 in the form of a plurality of multilayer neural networks. However, the "second reference acceleration model" corresponding to the various directions may be stored in the ROM 24 of the neurocomputer 22 in the form of a plurality of maps.

(11) In the above-mentioned embodiment, the "second reference acceleration model" reflecting the driver's driving orientation is run.
In the ROM2 of the neurocomputer 22 in the form of three types of pre-learned multilayer neural networks.
I memorized it in 4. On the other hand, the "second reference acceleration model" may be always learned by using one multi-layered neural network, like the "requested acceleration model". That is, as the “second reference acceleration model”, one learned in advance may be selected and used, or one obtained by learning at each time may be used.

In this case, the learning rate of the "second reference acceleration model", that is, the number of times of learning per unit time is made smaller than that of the "requested acceleration model", and the acceleration G with respect to the accelerator stroke S and the vehicle speed V is reduced. Learn relationships. As a result, it is possible to obtain the "second reference acceleration model" that reflects the driving orientation of each driver DR. Moreover, the “second reference acceleration model” can be obtained without the driver DR operating a specific switch or the like. It is also possible to reduce the number of multilayer neural networks used.

Alternatively, with respect to the travel of the vehicle 1 in the period designated by the driver DR, the relationship between the accelerator stroke S and the acceleration G with respect to the vehicle speed V is learned. As a result, it is possible to obtain the “second reference acceleration model” that reflects how the vehicle 1 runs during the period designated by the driver DR.

Alternatively, the characteristics of the "requested acceleration model" specified by the driver DR are replaced with the "second reference acceleration model". As a result, it is possible to obtain the “second reference acceleration model” that reflects the way the vehicle 1 was driven before the time specified by the driver DR.

[0083]

As described above in detail, according to the present invention, the acceleration of the vehicle is used as teacher data to be compared,
The relationship between the acceleration of the vehicle and the operation amount of the output operation means
Learning as an acceleration model required by the driver.
Also, the output of the learned acceleration model and the standard first
Of the second reference acceleration model that reflects the driver's driving orientation (preference), and also calculates the difference between the second reference acceleration model and the first reference acceleration model. is doing. Further, based on these deviations, the relationship between the operation amount of the output operation means and the control amount of the drive source with respect to the vehicle speed is calculated as the control amount sensitivity. Then, based on the calculated control amount sensitivity, the control amount of the drive source is controlled according to the operation amount of the output operation means.

Therefore, the strength of the driver's demand for acceleration at that time, and the driver's normal driving direction (preference)
The control amount sensitivity that reflects the strength of the control source is obtained, and the control amount of the drive source is controlled with an acceleration that always meets the driver's request and with an acceleration that reflects the driver's driving direction (preference). Further, the entire relationship between the acceleration and the control amount with respect to the operation amount and the entire relationship between the control amount and the operation amount are learned as a continuous model, and the entire relation between the acceleration and the control amount with respect to the entire operation range of the output operation means becomes discontinuous. There is no.

As a result, regardless of the driver's consciousness and driving environment, it is possible to always control the driving force that suits the characteristics of the driver, and the individual driver's orientation (preference) for driving can be improved. It is possible to quickly control the driving force. Moreover, there is an excellent effect that the control of the driving force can be made continuous over the entire range of the operation amount of the accelerator pedal or the like by the driver.

[Brief description of drawings]

FIG. 1 is a conceptual configuration diagram showing a basic conceptual configuration of the present invention.

FIG. 2 is a schematic configuration diagram showing a schematic configuration of a driving force control device for a vehicle in one embodiment embodying the present invention.

FIG. 3 is a block diagram showing an electrical configuration of a throttle computer and a neuro computer in one embodiment.

FIG. 4 is a configuration diagram showing a conceptual configuration of a multilayer neural network applied to a neuro computer in one embodiment.

FIG. 5 is a configuration diagram showing a conceptual configuration of a multilayer neural network which is also applied to a neuro computer in one embodiment.

FIG. 6 is a configuration diagram showing a conceptual configuration of a multilayer neural network which is also applied to a neuro computer in one embodiment.

FIG. 7 is a configuration diagram showing a conceptual configuration of a multilayer neural network which is also applied to a neuro computer in one embodiment.

FIG. 8 is a configuration diagram showing a conceptual configuration of a multilayer neural network which is also applied to a neurocomputer in one embodiment.

FIG. 9 is a characteristic diagram showing characteristics of a “requested acceleration model” in one embodiment.

FIG. 10 is a characteristic diagram showing characteristics of a “first reference acceleration model” in one embodiment.

FIG. 11 is a characteristic diagram showing characteristics of a “second reference acceleration model” in one embodiment.

FIG. 12 is a characteristic diagram showing a characteristic of “throttle sensitivity” in one embodiment.

FIG. 13 is a flowchart showing a “learning control routine” executed by a neuro computer in one embodiment.

FIG. 14 illustrates a “requested acceleration model” in one embodiment,
It is a characteristic view explaining the relationship of a "1st reference acceleration model" and a "2nd reference acceleration model."

FIG. 15 is a flowchart showing a “throttle opening control routine” executed by a throttle computer in the embodiment.

FIG. 16 is a time chart showing changes in the acceleration signal related to the action of the low-pass filter in the embodiment.

[Description of Reference Signs] 1 ... Vehicle, 2 ... Engine as drive source, 7 ... Throttle valve, 8 ... DC motor (7 and 8 constitute control amount changing means), 10 ... Accelerator as output operating means Pedal, 11 ... Accelerator sensor as operation amount detecting means, 12 ... Acceleration sensor as acceleration detecting means, 13
... vehicle speed sensor as speed detecting means, 21 ... throttle computer constituting drive control means, 22 ... first reference acceleration model storage means, second reference acceleration model storage means, required acceleration model learning means and control amount sensitivity calculation A neuro computer that constitutes a means.

─────────────────────────────────────────────────── ─── Continuation of the front page (51) Int.Cl. ⁵ Identification number Internal reference number FI Technical display area F02D 45/00 A 7536-3G G05B 13/02 L 9131-3H 13/04 9131-3H (72) Inventor Masuharu Oshima 41, Nagachote, Nagakute-cho, Aichi-gun, Aichi Prefecture 1 of Toyota Central Research Laboratory Co., Ltd. (72) Inventor ▲ Hiroyuki Yoshida, 41th, Yokocho, Nagakute-cho, Aichi-gun, Aichi Prefecture 1 Toyota Central Research Institute

Claims (1)

[Claims] 1. A control amount changing unit for changing a control amount of a drive source mounted on a vehicle, an output operating unit operated by a driver for arbitrarily controlling an output of the drive source, and An operation amount detection unit for detecting an operation amount of the output operation unit, and controlling the output of the drive source by driving the control amount changing unit according to a detection result of the operation amount detection unit. A vehicle driving force control device configured to control a driving force of a vehicle, an acceleration detecting means for detecting an acceleration of the vehicle, a speed detecting means for detecting a speed of the vehicle, and the output operating means. First reference acceleration model storage means for storing a standard first reference acceleration model learned in advance regarding the relationship between the operation amount of the vehicle and the acceleration of the vehicle with respect to the speed of the vehicle; A second reference for storing a second reference acceleration model learned about the relationship between the operation amount of the operation means and the acceleration of the vehicle with respect to the speed of the vehicle, and the orientation (preference) for the driver's running. Acceleration model storage means, as the teacher data to be compared with the acceleration obtained by the detection of the acceleration detection means, the difference between the teacher data and the output of the means as an error signal, the error so as to reduce the error, A requested acceleration model learning means for learning the relationship between the operation amount of the output operation means and the acceleration obtained by the detection by the speed detection means as the acceleration model requested by the driver; The second reference acceleration model stored in the reference acceleration model storage means and the first reference acceleration model storage means are stored in the first reference acceleration model storage means. The deviation from the first reference acceleration model that is being calculated, the output of the acceleration model learned by the required acceleration model learning means, and the first reference acceleration stored in the first reference acceleration model storage means. A deviation from the model or the second reference acceleration model stored in the second reference acceleration model storage means is calculated, and obtained by detecting the operation amount of the output operation means and the speed detection means based on each deviation. Detected by the operation amount detection means based on the control amount sensitivity calculated by the control amount sensitivity calculation device for calculating the relationship of the control amount of the drive source with respect to the speed as the control amount sensitivity. And a drive control means for controlling the drive of the control amount changing means according to the operated amount.