JPH0781604A - Automatic travel vehicle - Google Patents

Abstract

PURPOSE:To judge travel possibility degree by conducting fuzzy theory based on respective detection values of obstacle conditions and own vehicle conditions, recognize travel environment including other vehicles by selecting one of the plural travel possible regions based on the results, and enable automatic travel while changing a traffic lane arbitrarily. CONSTITUTION:An automatic travel vehicle is provided with a CCD camera 10, a radar unit 12 consisting of a millimeter wave radar, a yaw rate sensor 14, a speed sensor 16, a steering angle sensor 18, etc., a plurality of travel possible regions are recognized in the vehicle approach direction by a controller based on these detection signals, and the conditions of an obstacle including at least location of the obstacle existing around the vehicle is detected. Fuzzy theory is conducted based on detection value of the own vehicle so as to judge travel possibility degree, either of a plurality of travel possible region is selected based on the judgement result, and control quantity of vehicle travel is calculated based on own vehicle conditions and calculated travel locus so as to make it travel in the selected travel possible region.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an autonomous vehicle,
More specifically, the present invention relates to an automated vehicle capable of automatically traveling while recognizing a traveling environment including other vehicles and arbitrarily changing lanes in a road environment where there are a plurality of lanes such as a highway.

[0002]

2. Description of the Related Art Various automatic driving techniques have been proposed in the past, for example, Japanese Patent Laid-Open No. 2-226310.
In the publication, a technique is proposed in which the steering of the vehicle is controlled by performing fuzzy inference based on lane boundaries, obstacle information, and the like extracted from a video image of the road surface on which the vehicle travels. Further, the present applicant has previously proposed a technique for recognizing a lane boundary line in an automatic vehicle in Japanese Patent Laid-Open No. 3-158976.

[0003]

As described above, various technologies have been proposed so far for automatic driving technology. However, on a road having a plurality of lanes (lanes), lanes are arbitrarily changed to automatically There was no proposal to make it run. Such a technology is extremely useful when automatically driving on a highway, etc., but in the road environment where obstacles including other vehicles exist, it determines its own behavior while automatically selecting the optimal lane. There are many problems to be solved in order to drive.

Therefore, an object of the present invention is to solve the above-mentioned problems, and in a road environment having a plurality of lanes and obstacles including other vehicles, the optimum lane is selected to stably and automatically drive. The purpose is to provide such an automated vehicle.

Further, the vehicle has a plurality of lanes, and in addition to selecting an optimal lane in a road environment where obstacles including other vehicles exist, it is possible to follow the preceding vehicle while avoiding obstacles. It is intended to provide an autonomous vehicle.

Further, it is an object of the present invention to provide an automatic vehicle which uses fuzzy inference and simplifies the inference to reduce the amount of calculation and memory used, and which improves the debugging efficiency.

[0007]

In order to solve the above-mentioned problems, the present invention is, for example, as set forth in claim 1, in a vehicle which automatically travels while recognizing the outside world, a plurality of traveling directions are possible in the traveling direction of the vehicle. At least the recognition means for recognizing the area, the obstacle state detection means for detecting the state of the obstacle including at least the position of the obstacle existing around the vehicle, the vehicle speed, and the position and direction of the own vehicle with respect to the plurality of drivable areas Own vehicle state detecting means for detecting the state of the own vehicle, traveling possibility degree determining means for determining a traveling possibility degree for the plurality of traveling possible areas by performing fuzzy inference based on the detected value, and the determined traveling possibility degree Selecting means for selecting any one of the plurality of drivable areas based on the trajectory, and trajectory calculating means for determining the traveling trajectory of the vehicle based on at least the detected vehicle state And the control amount calculating means for calculating a control amount of the vehicle traveling on the basis in order to travel the selected travelable regions on at least the detected travel locus calculated between the vehicle state, and as constituted comprising a.

[0008]

In the road environment having a plurality of drivable areas including obstacles including other vehicles, stable automatic driving can be performed while selecting an optimal drivable area. The drivable area here means the lane.

[0009]

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

FIG. 1 is an overall perspective view of an automatic vehicle according to the present invention. In the figure, the autonomous vehicle is C
It has one CD camera 10. The CCD camera 10 is fixed to the rear-view mirror mounting position above the driver's seat, and allows a monocular view of the vehicle traveling direction. Reference numeral 12 denotes a radar unit composed of a millimeter wave radar, which includes two front radars 12a attached to the front of the vehicle, three side radars 12b attached to the side of the vehicle, and two rear radars attached to the rear of the vehicle. A rear radar (not shown) comprises a total of 10 radar groups, and detects the presence of a three-dimensional obstacle such as another vehicle through reflected waves. A yaw rate sensor 14 is provided near the center of the vehicle interior.
Is provided to detect the angular acceleration about the vertical axis (z axis) of the vehicle. Further, a vehicle speed sensor 16 including a reed switch is provided near a drive shaft (not shown) of the vehicle to detect the traveling speed of the vehicle and to detect the steering angle sensor 1
8 is provided in the vicinity of the steering shaft 20 of the vehicle to detect the steering rudder angle.

A steering angle control motor 22 is attached to the steering shaft 20, a throttle actuator 24 composed of a pulse motor is attached to a throttle valve (not shown), and a brake (not shown) is further provided. A brake pressure actuator 26 (not shown in Figure 1) is attached. In this configuration, the vehicle is steered in accordance with the calculated steering angle control amount,
The throttle valve is opened and closed to adjust the vehicle speed, and the brakes are activated as necessary to automatically drive the vehicle.

FIG. 2 is a block diagram showing the above structure in more detail. The output of the CCD camera 10 is sent to the image processing hardware 30 and subjected to necessary processing, and the result is stored in the shared memory 34 via the bus 32. The image processing CPU 36 and the image evaluation CPU 38 read the stored value at every predetermined time to detect the state of the road surface on which the vehicle travels.
The output of the radar unit 12 is stored in the shared memory 34 via the radar processing circuit 40 and the bus 32. The radar evaluation CPU 42 reads the store value at every predetermined time and detects the position of the obstacle on the coordinates.

The outputs of the vehicle speed sensor 16 and the like are sent to the locus estimation CPU 44 to estimate the locus of movement of the vehicle.
The action plan decision making CPU 50 creates a target route from the store value. The movement locus of the vehicle estimated as the target route is sent to the locus tracking CPU 46, and the locus (target route) tracking control amount is determined there. Furthermore, trajectory tracking control CP
U46 calculates the steering angle control amount and outputs it to the steering angle control CPU52. The steering angle control CPU 52 is the PWM controller 5
4 and the driver 56 to control the steering angle control motor 22.
To drive. The motor drive amount is detected by the encoder 58 and feedback control is performed.

Further, the action plan decision-making CPU 50 obtains a target acceleration of the vehicle body by the speed / following control section thereof as described later,
It is sent to the vehicle speed control CPU 60. The vehicle speed control CPU 60 is an accelerator pulse motor controller 62, a driver 64
The throttle actuator 24 is driven via the brake solenoid actuator 66, and the brake pressure actuator 26 is driven via the brake solenoid controller 66 and the driver 68. The driving amount is detected via the pressure sensor 70,
The second feedback control is performed. In the above, the processing circuit of the sensor such as the waveform shaping circuit is omitted for simplification of the drawing. FIG. 3 functionally shows the block diagram of FIG.

Next, the operation of this automatic vehicle will be described with reference to the flow chart of FIG. The flow chart of FIG. 4 shows the operation performed by the action plan decision-making CPU 50 described above.

First, in S10, the image evaluation result is input. This is done by inputting the outputs of the image processing CPU 36 and the image evaluation CPU 38 described above. That is, image processing C
The PU 36 extracts road lane markings (lane boundaries, indicated by white or yellow solid lines or broken lines) from the input image as shown in FIG. 5 to obtain image plane coordinates, and projectively transforms them to obtain real plane coordinates. Then, each line component is output as point sequence data on the real plane with a geometrical relation added. Image evaluation CP
The U38 outputs detailed data such as the lane width based on it. These are input in S10.

Then, in S12, the radar evaluation result is input. This is done by inputting the output of the radar evaluation CPU 42 described above. By going through S10 and S12, the action plan decision making CPU 50 obtains the external environment information as shown in FIG.

Next, in S14, the positional relationship between the lane boundary line (or lane) and the own vehicle is obtained. Image evaluation CPU3
The same lane data from 8 is not output every time, but the recognized lane boundary line is output in real time. Therefore, the same lane boundaries and lanes are identified by the same numbers, and the positional relationship between the lane boundaries (or lanes) and the vehicle is associated. The above is shown in FIG. The specification of the lane boundary line (or lane) is described in another application (reference number A93-1025) filed on the same date, so the description will be limited to this extent.

Then, in S16, the positional relationship between the lane and the obstacle is obtained. Although illustration is omitted, this is also the same as the operation in S14, and is an operation for associating the positional relationship between the obstacle and the lane boundary line (or lane). If there are multiple obstacles, specify them by adding numbers.

Then, the process proceeds to S18 to extract the lane information, and the process proceeds to S20 to extract the obstacle information in the lane. This is a work for obtaining a feature and a feature amount by performing an appropriate process for the fuzzy inference performed subsequently.

More specifically, since the lane information is merely the coordinates of the line segment in the state of the input value, it is linearized as shown in FIG. 8 to obtain the lane width and the lane length as its features or feature quantities. . In addition, since the obstacle information is merely an obstacle existing in the coordinates in the state of the input value, as shown in FIG. 9, the features such as the distance to each obstacle or the feature amount is extracted to show the relationship with the own vehicle. Finally, as shown in the figure, the distance to each obstacle, the relative speed between the own vehicle speed and the obstacle speed, and the deviation between the target vehicle speed and the obstacle speed are obtained and used as the characteristic amount.

Subsequently, the process proceeds to S22, and the above-mentioned characteristic amount (inference parameter) is converted into a fuzzy amount. The input value in fuzzy inference basically takes a value from -1 to 1, so that the input value is normalized and the dynamic range of the normalized information amount is not lost due to an unnecessary large value. Stop in range. In the case of the example, the specific determination was made as follows. Feature quantity Fuzzy quantity Lane length 0-100m 0-1 Lane width 0-5m 0-1 Distance to each obstacle 0-100m 0-1 Relative speed between own vehicle speed and obstacle speed-100-100km / h- 1-1 Deviation between target vehicle speed and obstacle speed -100 to 100 km / h -1 to 1

At the same time, a membership function of the fuzzy set (hereinafter referred to as "fuzzy label") is set. The set fuzzy labels are shown in FIGS. In the embodiment, ZO is used as the fuzzy label.
(Zero), PS (slightly large), PM (medium size),
PB (very large), NB (very small), NM
(Small size is medium), N (small), Z (near zero), P
Nine kinds (large) were used and used properly according to the feature amount (inference parameter).

Then, in S24, the fuzzy expert system infers the drivability degree for each lane.

Explaining this, in an environment in which roads, obstacles, etc. are intricately entwined with each other, front, side, and
In order to optimally judge and act according to various situations in the back, the combination of rules is not uniquely determined by a simple rule description method, and rules increase, omissions, and inconsistencies are likely to occur. In addition, it is extremely difficult to adjust the feeling of autonomous driving and the feeling of driving of the passenger without any discomfort. A fuzzy expert system that uses fuzzy inference is considered to be the most suitable control method that is optimal and has little discomfort to passengers in such a situation. If it is a fuzzy expert system,
This is because multiple inputs and parallel simultaneous inference are possible, judgment patterns can be combined without omission, and adjustments can be made to match the human feeling simply by changing the membership function. Therefore, in the embodiment, the fuzzy expert system is used.

By the way, even if the fuzzy expert system is used, the number of rules becomes huge because of the large number of input conditions. Therefore, the number of rules was reduced by adopting the method described below.

FIG. 12 shows a fuzzy expert system in the embodiment. In the embodiment, the fuzzy reasoning is composed of primary reasoning and secondary reasoning. In the first-order inference, the drivability degree of each lane is inferred from the characteristic amount of the road by using the lane drivability degree rule, and 2 from the characteristic amount of the obstacle.
The degree of obstacle runnability is inferred using certain types of obstacle risk determination rules.

In the second-order inference, traveling is performed based on the traveling possibility inferred through the lane traveling possibility rule and the traveling possible degree obtained through two kinds of obstacle danger judgment rules (collectively referred to as obstacle danger degree rules). The drivability degree of each lane is totally inferred using the possibility determination rule.

Since the feature amount (inference parameter) itself does not differ depending on the lane, the same rule group is used to infer each lane. However, it goes without saying that the inference parameter to be detected differs depending on the lane because the inference parameter to be detected differs depending on the lane.

FIGS. 13 to 16 show these three types of primary inference rules and one type of rule group for secondary inference. As shown, they all consist of 16 rules whose antecedent part consists of two fuzzy labels.

The inference method of the present invention (hereinafter referred to as "multiple inference") will now be described. In this multiple inference, as shown in FIG. 17, the consequent part of the primary inference rule is used as the antecedent part of the secondary inference. I decided to use it. More specifically, FIG.
As shown in FIG. 16 and FIG. 18, a rule having the maximum grade value of each fuzzy label is selected for the consequent part of the primary inference rule, and the grade value (indicating weight) is used as the representative value of the fuzzy label. It was applied to the fuzzy label in the antecedent part of the secondary inference rule, and the final output of the secondary inference rule was obtained and used as the output of the fuzzy expert system.

That is, as shown in FIG. 19, the usual fuzzy inference is rule R1 (for convenience of explanation, the usage rule is R
1 and R2), the minimum value of the membership values of the antecedent fuzzy labels A11, 12 is selected to obtain the grade value ω1 of the antecedent fuzzy label B1.
The same operation is performed for the rule R2 to obtain the grade value ω2 of the consequent part fuzzy label B2, the union B0 of the two is obtained on the domain, and the center of gravity y0 is obtained to obtain the rule group R2.
The final output of 1 and R2. As a result, when the fuzzy inference is continuously performed as shown in FIG. 12, the number of rules becomes enormous as shown in FIG. 20 according to the usual method.

Therefore, in the embodiment, as shown in FIG. 13 and the like, a fuzzy label having the maximum grade value is selected and the fuzzy label is represented. That is, in the case of FIG. 13, Z is used as the fuzzy label.
Although four types of O, PS, PM and PB are used, the consequent part has a plurality of values for the same fuzzy label. For example, in terms of PM, it has three values. The three values of PM have a magnitude relationship as shown in the lower part of the figure.

Looking at the magnitude relation of the three values of PM, the largest one among them is that the fuzzy label (that is, PM) has the highest degree of satisfaction of the rule.
It can be considered that the fuzzy label can be represented by the one with the maximum grade value. Therefore, in the present invention, the maximum grade value is obtained for each fuzzy label, and the grade value is substituted for the fuzzy label in the antecedent part of the secondary inference rule.

The grade value means a weight in fuzzy inference. In other words, in the first inference, only the weight of each fuzzy label is obtained and the final output on the domain is not obtained, and the weight is first used in the second inference. The final output on the domain is calculated. The final output of the secondary inference rule thus obtained is shown at the end of FIG. Incidentally, the final output is generally performed by obtaining the center of gravity of the union of the consequent part membership values as described above, but it takes processing time.
Here, a simple center of gravity calculation called the so-called singleton method was used.

As a result, as shown in FIG. 20, it is possible to greatly reduce the number of rules, and as a result, it is possible to reduce the calculation amount or the used memory capacity, and it is possible to improve the debugging efficiency. .

FIG. 21 shows an output example of the drivability degree obtained by the inference of the fuzzy expert system described above. The drivability is determined as a real number from 0 to 1 for each lane as shown in the figure. The illustrated example is a case of a road environment in which there are a total of three lanes on the left and right around the own lane. The output is performed for all the recognized lanes, and the output value of the lane which is recognized at the previous inference time and is not recognized at the inference time becomes 0.

In the example, since each lane was evaluated in this way, obstacles exist at various positions in each lane,
Moreover, it was possible to accurately evaluate the complicated road conditions that change from moment to moment. As shown in FIG. 12, one of the lanes is then selected based on this, but the selection work can be simplified and can be performed smoothly.

Returning to the flow chart of FIG.
Proceed to step 26 to determine the target lane. Fuzzy reasoning is not used for this.

FIG. 22 is a flow chart showing the selection work. Before entering the description of FIG. 23, the selection work will be briefly described with reference to FIG. 23. Until the drivability degree of the own lane becomes equal to or lower than the threshold old value, the drivability degree of the other lane is larger than that. I did not change the lane. When the vehicle becomes less than the threshold old value, a lane with a driving possibility greater than the threshold old value is selected, and once selected, even if the driving possibility of the selected lane becomes less than the threshold old value, the selected lane is selected. And the difference in the degree of drivability in other lanes are set values (set appropriately)
Unless you exceed the limit, you will not change lanes. That is,
Hysteresis is provided to prevent control hunting.

Explaining the above with reference to FIG. 22, first, in S100, it is judged whether or not the travelable degree of the own lane is less than or equal to the threshold value. If the result is negative, the routine proceeds to S102 to maintain the own lane, and S104 Go to and replace the target lane with your own lane (however, in this case, your own lane is replaced with your own lane). When the result in S100 is affirmative, the program proceeds to S106, in which it is determined whether or not the drivability in the right lane is greater than or equal to the threshold value.

When the result in S106 is affirmative, the program proceeds to S108, in which it is determined whether or not the difference between the travelable degree in the own lane and the travelable degree in the right lane is smaller than the set value. This is because if there is no great difference in the degree of drivability even if the lane is changed to the right lane, the lane change is of little significance, and therefore it is determined. Therefore, when the determination is negative, the difference in the travelable degree is not small, and thus the process proceeds to S110, where the target lane is the right lane, and the process proceeds to S104 where the target lane is replaced with the own lane.

On the other hand, if it is determined in S108 that the difference in the degree of travelability from the right lane is smaller than the set value, S112.
Then, it is judged whether or not the degree of drivability in the left lane is equal to or more than the threshold old value. When the result in S106 is negative, there is no further consideration for the right lane, so the process immediately jumps to S112.

If it is determined in S112 that the left lane is not more than the threshold value, it is meaningless to change to the left lane. Therefore, the process proceeds to S102 to maintain the own lane and is affirmed in S112. When S114
Then, it is determined whether the difference between the driving lanes of the own lane and the left lane is smaller than the set value, and when the determination is affirmative, S1
When it is denied, the program proceeds to 02 to maintain the own lane, and when the result is negative, the program proceeds to S116 to set the left lane as the target lane. It should be noted that the reason why the judgment is made from the right lane is that, in principle, overtaking is done from the right side when passing by law.

Returning to the flow chart of FIG.
Proceed to step 28 to plan a target route for the determined target lane.

Referring to FIG. 24, when the target lane is the own lane, points x and y are set on the lane every 2.5 m in the center of the lane. Here, x and y indicate positions on coordinates as shown in FIG. By planning a target route in the center of the lane, you can drive safely and automatically.
In this embodiment, the X axis of the coordinates is in the vehicle traveling direction.

When the target lane is another lane, a reference line is obtained in the center of the lane, an appropriate position (target lane change, point) (a) is selected on the reference line, and the own position is selected from the position (a). A line segment (b) to the vehicle position is obtained, and points x and y are set every 2.5 m for the line segment. The position (a) is obtained by the distance L obtained by multiplying the own vehicle speed V by the scheduled lane change (lane change) time T. After the lane change, the target route is planned along the central reference line as in the case of traveling in the own lane.

Furthermore, if there is an obstacle near the position (a), there is a possibility of contact, so the target position is the position obtained by subtracting 5 m from the position of the obstacle, and the line segment (b ) And perform the same processing.

Next, in S30, the trajectory estimation result is input. As described above with reference to FIG. 2, the locus estimation CPU 44 estimates the traveling locus of the own vehicle from the output of the yaw rate sensor 14 as shown in FIG.
The stored value is read out and the present position of the vehicle is estimated in this step.

Then, in S32, the delay time of the target route is corrected.

That is, the target route is planned in S28, and the traveling locus of the own vehicle is input in S30 to proceed along the route, but it takes some processing time to determine the target lane,
My car is obviously moving in the meantime. Therefore, FIG.
As shown in, the position of the host vehicle should be different between when the image is input and after the processing time has elapsed.Therefore, if the target route is not determined in consideration of the moved position, there will be an error in the target route tracking control. Occurs. Therefore, in this step, as shown in FIG. 26, the coordinates are changed so that the position of the own vehicle is placed at the origin position of the coordinates,
Delay correction of the target route (indicated by a sequence of points).

Then, in S34, the target traveling speed (acceleration) is calculated by the fuzzy control system.

FIG. 27 is a block diagram showing this, and FIG. 28 is a block diagram showing details of the acceleration controller therein. Further, FIG. 29 shows a fuzzy membership function used for inferring the target acceleration. In FIG. 27, the fuzzy inference unit performs fuzzy inference using the distance to the obstacle, the relative speed of the own vehicle speed and the obstacle speed, and the own vehicle speed as input parameters to calculate the target acceleration. The acceleration is represented by the first-order difference value of the vehicle speed. Moreover, the fuzzy inference uses the conventional method instead of the multiple inference described above.

In the acceleration controller shown in detail in FIG. 28, the deviation between the target acceleration and the actual acceleration is obtained, and the throttle opening or the brake pressure is PID-controlled to the target value according to the deviation. With such a configuration, it is possible to automatically travel at the target acceleration while keeping the distance to the front vehicle at a predetermined value, and it is possible to accelerate or decelerate or stop according to the movement of the front vehicle.

Finally, in S36, the trajectory tracking control CPU shown in FIG.
46, steering angle control CPU 52, vehicle speed control (acceleration control) C
The target route and the target speed (acceleration) are output to the PU 60, and the process ends.

30 to 33 show the actual vehicle running test results for the above. FIG. 30 shows the case of following a preceding vehicle traveling at a constant speed, and it can be seen that the vehicle is accurately following the preceding vehicle while maintaining a predetermined separation distance. Figure 31
Fig. 3 shows a comparison between the case where the travelable degree at that time is inferred by the conventional method and the case where the inference is performed using the multiple inference proposed by the present invention. As shown in the figure, there was almost no difference between the two results.

FIG. 32 is a test data diagram similar to that of FIG. 30 and shows a case where the vehicle is stopped in response to a vehicle in front. In addition, in FIG.
The inference of the drivability degree at that time is shown in comparison between the conventional method and the case of using the multiple inference proposed in the present invention for the obstacle detection point. Similar to FIG. 31, there is almost no difference between the two in this test result.

In this embodiment, the concept of the degree of lane travelability is used as described above, and it is determined by fuzzy reasoning. Therefore, it is always optimal even in a road environment in which a plurality of lanes are parallel and obstacles are scattered. You can select a different lane and drive stably. In addition, since the lane is selected by inferring the drivability for each lane based on the presence or absence of an obstacle, there is no danger of contact with the obstacle.

Further, since the degree of running possibility for each lane is determined through fuzzy inference, driving control suitable for the passenger's feeling is possible, and multiple inference is used in the fuzzy inference. The amount of calculation and the amount of memory can be reduced, and the debugging efficiency also improves.

Furthermore, since the degree of drivability for each lane is determined through fuzzy reasoning, and the target lane is then determined by comparison with the threshold old value, the lane can be changed by raising or lowering the threshold old value. The frequency and the like can be adjusted, and the vehicle can travel more optimally according to the road environment.

In the above description, the fuzzy inference multiple inference is connected in two stages, but three or more stages may be connected as long as the inference parameters do not interfere with each other.

In addition to the fuzzy inference parameters shown in the embodiment, various parameters such as lane curvature, obstacle speed, own vehicle speed, and target vehicle speed may be used. Further, in the embodiment, only the obstacle existing ahead of the obstacle is used as the fuzzy inference parameter, but the obstacle existing behind the obstacle may be used as the fuzzy inference parameter.

Furthermore, although the visual sensor is monocular, the front radar may be omitted by using binocular vision.

[0064]

According to the first aspect of the present invention, the optimum lane can always be selected and the vehicle can travel stably even in a road environment where a plurality of lanes are parallel to each other and obstacles exist.

According to the second aspect, in addition to the above-mentioned effects, it is possible to select an optimum vehicle speed and travel while reliably avoiding an obstacle.

According to the third aspect of the present invention, it is possible to always select the optimum lane even in a road environment in which a plurality of lanes are parallel to each other and obstacles exist, and along the target locus toward the selected lane. And can drive accurately.

According to the present invention, a plurality of lanes are parallel to each other, the optimum lane can always be selected even in a road environment where obstacles exist, and the vehicle runs in the center of the lane. You can drive to.

According to the fifth aspect of the present invention, fuzzy inference is used to obtain a control value required for automatic driving, and control matching the human feeling is easily possible.
The calculation amount of inference and the memory capacity used can be reduced, and the debugging efficiency can be improved.

[Brief description of drawings]

FIG. 1 is an explanatory perspective view generally showing an automatic vehicle according to the present invention.

FIG. 2 is a block diagram showing in detail the sensor shown in FIG. 1 and its processing.

3 is a functional block diagram of the configuration of the block diagram of FIG. 2;
It is an explanatory view similar to.

FIG. 4 is a flow chart showing an automatic traveling operation by the operation of the action plan decision-making CPU of the block diagram of FIG.

FIG. 5 is an explanatory diagram of image processing for explaining the image evaluation result of the flow chart of FIG. 4;

FIG. 6 is an explanatory diagram showing external environment information obtained by inputting image evaluation results and radar evaluation results in the flow chart of FIG. 4;

7 is an explanatory diagram illustrating a positional relationship between a lane boundary line and a self-position (self-vehicle) in the flow chart of FIG. 4;

FIG. 8 is an explanatory diagram showing characteristic amounts of lanes in the flow chart of FIG. 4;

FIG. 9 is an explanatory diagram showing feature quantities of obstacles in the flow chart of FIG. 4;

10 is an explanatory diagram showing a membership function (fuzzy label) set for the feature amount of the lane of FIG.

11 is an explanatory diagram showing a membership function (fuzzy label) set for the feature quantity of the obstacle in FIG. 9. FIG.

FIG. 12 is an explanatory diagram generally showing target lane determination performed by the driving degree inference for each lane and the inference value by the fuzzy expert system of the flow chart of FIG. 8;

13 is an explanatory diagram showing a lane traveling possibility degree rule group used in the fuzzy expert system of FIG. 12;

14 is an explanatory diagram showing an obstacle risk determination rule group used in the fuzzy expert system of FIG.

15 is an explanatory diagram showing another obstacle risk determination rule group used in the fuzzy expert system in FIG. 12. FIG.

16 is an explanatory diagram showing a total runnability degree rule group for secondary inference used in the fuzzy expert system of FIG. 12;

FIG. 17 is an explanatory diagram showing a technique called multiple inference used in the present invention.

FIG. 18 is an explanatory diagram more specifically showing the method of multiple inference used in the present invention.

FIG. 19 is an explanatory diagram showing fuzzy inference according to a conventional technique.

FIG. 20 is an explanatory diagram showing the reduction degree of the number of rules by the multiple inference used in the present invention as compared with the fuzzy inference by the conventional technique.

FIG. 21 is a timing chart showing the drivability degree inferred from the flow chart of FIG.

22 is a subroutine flow chart for explaining the target lane determination work of the flow chart of FIG.

FIG. 23 is a timing chart explaining the target lane determination work of the flow chart of FIG. 4;

FIG. 24 is an explanatory diagram showing a work of planning a target route for a target lane in the flow chart of FIG. 4;

FIG. 25 is an explanatory diagram illustrating a trajectory estimation result of the flow chart of FIG. 4;

FIG. 26 is an explanatory diagram showing a delay time correction operation for the target route in the flow chart of FIG. 4;

FIG. 27 is a block diagram of target traveling speed (acceleration) control of the flow chart of FIG.

28 is a block diagram showing details of the acceleration controller in FIG. 27. FIG.

FIG. 29 is a parameter used in fuzzy inference of the target acceleration in the control of FIG.

FIG. 30 is a test data diagram of following vehicle traveling in the automatic traveling operation according to the present invention.

FIG. 31 is an explanatory view showing a case where the travelable degree is inferred by the conventional method in the test of FIG. 30 and a case where it is inferred by multiple inference according to the present invention in comparison.

FIG. 32 is an actual vehicle running test data diagram similar to FIG. 30.

FIG. 33 is an explanatory diagram showing comparison of inference results as in the case of FIG. 32.

[Explanation of symbols]

 10 CCD camera 12 radar unit 14 yaw rate sensor 16 vehicle speed sensor 18 steering angle sensor 30 image processing CPU 36 image evaluation CPU 42 radar evaluation CPU 44 trajectory estimation CPU 46 trajectory tracking control CPU 50 action plan decision making CPU 52 steering angle control CPU 60 vehicle speed Control (acceleration control) CPU

─────────────────────────────────────────────────── ─── Continuation of front page (51) Int.Cl. ⁶ Identification code Internal reference number FI Technical display location G08G 1/16 C 7531-3H // G05B 13/02 N 9131-3H B62D 101: 00 113: 00 137: 00

Claims (5)

[Claims] 1. A vehicle that automatically travels while recognizing the outside world, comprising: a. Recognition means for recognizing a plurality of drivable areas in the traveling direction of the vehicle, b. Obstacle state detecting means for detecting the state of the obstacle including at least the position of the obstacle existing around the vehicle, c. Vehicle state detection means for detecting the state of the vehicle including at least the vehicle speed and the position and direction of the vehicle with respect to the plurality of drivable areas; d. Drivability degree determining means for determining a drivability degree for the plurality of drivable areas by performing fuzzy inference based on the detected value, e. Selecting means for selecting one of the plurality of drivable areas based on the determined drivability degree; f. Trajectory calculating means for obtaining a traveling locus of the own vehicle based on at least the detected state of the own vehicle, and g. A control amount calculation means for calculating a control amount of vehicle traveling based on at least the detected state of the own vehicle and the calculated traveling locus for traveling in the selected travelable region is provided. Self-driving vehicle. 2. h. At least the moving locus calculating means for obtaining a moving locus of the obstacle based on the detected state of the obstacle, the control amount calculating means is at least the obtained locus of the own vehicle and the moving locus of the obstacle. The automatic vehicle according to claim 1, wherein the control amount of the vehicle traveling is calculated based on the vehicle traveling amount. 3. The control amount calculation means comprises means for correcting a time delay until the vehicle starts moving to the selected travelable area. Self-driving vehicle. 4. The automatic travel according to claim 1, wherein the control amount calculation means includes means for determining an optimum position of the vehicle with respect to the selected travelable area. vehicle. 5. A vehicle that automatically travels while recognizing the outside world, comprising: a. At least means for recognizing a travelable area in which the vehicle travels and detecting a parameter indicating a state of the travelable area, b. Means for detecting at least a parameter indicating the traveling state of the vehicle, c. A first fuzzy inference means for setting a fuzzy production rule using the detected parameters in the antecedent part, performing fuzzy inference, and then obtaining the weight of the antecedent part; d. The consequent part is used as the antecedent part to set the nth (n ≧ 2) fuzzy production rule, and the nth fuzzy inference is performed to determine the consequent from the weight and the position of the center of gravity of the consequent part of the nth rule. N fuzzy inference means for obtaining the output of the section, and e. An automatic traveling vehicle comprising: a control amount calculating means for calculating a control amount of vehicle traveling based on the obtained output of the consequent portion.