JP3147541B2 - Obstacle recognition device for vehicles - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an obstacle recognition device for a vehicle, and more particularly, to a vehicle warning device that issues a warning to a driver so that the vehicle can travel safely without collision with an obstacle. The present invention relates to an obstacle recognition device for a vehicle that is suitable as an obstacle recognition device or an obstacle detection device required for a device that automatically controls a vehicle speed. Further, the vehicle obstacle recognition device of the present invention is particularly suitable as a vehicle obstacle recognition device for a safe traveling device during low-speed traveling such as parallel parking or garage parking in a transport vehicle or an automobile in a factory.

[0002]

2. Description of the Related Art Conventionally, there has been known an obstacle recognizing device for a vehicle which is used for detecting an obstacle which prevents smooth running of a vehicle, and for warning or decelerating or stopping the vehicle. ing. This vehicle obstacle recognition device projects ultrasonic pulses or pulse-like microwaves using a sensor that projects ultrasonic waves or microwaves, from the time until the reflected wave reflected by the obstacle returns. This is to grasp the distance to an obstacle within a range where the ultrasonic waves or the microwaves reach. Further, there is also a device that uses light to irradiate light in the same manner as described above, detects reflected light from an obstacle, and detects the distance to the obstacle. A disadvantage common to these sensors is that as the position of the obstacle moves away from the vehicle, the intensity of ultrasonic waves, light and the like attenuates, making it difficult to detect the obstacle. There is also a disadvantage that a misperception occurs due to the multiple-reflected medium.

Japanese Patent Laying-Open No. 57-175442 discloses a technique in which, when it is detected that an obstacle is present within a predetermined distance near a predetermined vehicle, the predetermined distance is held on a display unit and displayed. ing. This technology stores and holds detection information when an approaching obstacle is detected in order to alert and display an obstacle that cannot be detected due to entering a nearby blind spot. Similar to the invention, but different from the invention.
Further, no measure is taken against the problem that reliability is reduced when an obstacle is detected by a single sensor signal.

Japanese Patent Laying-Open No. 57-182573 stores a plurality of distance information output at predetermined time intervals,
When each distance information is within a predetermined range centered on a distance value determined by the plurality of distance information, by outputting the distance value as distance information, when detecting an obstacle using ultrasonic waves, A technique for preventing an erroneous detection of a distance and obtaining an accurate distance has been disclosed. However, it is a technique to prevent malfunction due to noise by repeating sensing within a time short enough to neglect the movement of the vehicle and selecting one with a similar sensed value, aiming to improve the reliability of sensing Is common to the present invention, but differs from the present invention, but does not include a function of comprehensively judging the sensing result after moving. Therefore, the application is limited only to a sensor that can be repeatedly measured in a short time, and the reliability is poor because the past results are not comprehensively determined even when the vehicle moves.

Japanese Patent Application Laid-Open No. 58-221111 discloses a method in which a distance measuring means and a two-dimensional display means are provided, and information corresponding to the distance between the vehicle and the obstacle measured by the distance measuring means is displayed on the two-dimensional display means. At the same time, there is disclosed a technique of updating display information of a two-dimensional display means to different display coordinates at a predetermined timing according to a speed of a vehicle, and displaying a positional relationship between the vehicle and an obstacle on the two-dimensional coordinates. I have. According to this technology,
The driver can surely know the positional relationship between any part of the vehicle and the obstacle on the two-dimensional coordinates, and can visually confirm the vehicle speed information and the obstacle simultaneously. However, it is a technology that simply links the display of coordinates and the movement of the car, and only shows a method that is used when traveling near straight ahead. There is no description of how to obtain more reliable information by fusing the information.

As described above, in the prior art, inconveniences such as oversight of obstacles and erroneous recognition cannot be avoided, and this is an obstacle to improving the reliability of the system and speeding up unmanned traveling.

SUMMARY OF THE INVENTION The present invention has been made to solve the above problems, and has as its object to reduce the frequency of occurrence of oversight or erroneous recognition which is inevitable from the characteristics of the obstacle sensor itself.

[0008]

SUMMARY OF THE INVENTION In order to achieve the above object, the present invention provides an obstacle sensor for detecting an obstacle around a vehicle, a movement amount output means for outputting a movement amount of the vehicle, and a plurality of small areas. Storage area for storing a probability value representing at least one of a probability that an obstacle exists and a probability that an obstacle does not exist for each of the small areas, based on the output of the obstacle sensor. First probability updating means for updating the probability value stored in the storage means, and displacement and azimuth of the vehicle stored in the storage means based on changes in the displacement and azimuth of the vehicle obtained from the output of the movement amount output means. And a second updating means for updating the probability value of the small area before the change of the vehicle has occurred to the probability value of the small area after the change of the displacement and the azimuth of the vehicle has occurred.

[0009]

According to the present invention, for a region around a vehicle divided into a plurality of small regions, a storage means for storing a probability value for each small region is provided, and information on the presence or absence of an obstacle for each small region around the vehicle is provided. Probability value representing the probability that an obstacle is present, using the output of an obstacle sensor that detects obstacles around the vehicle as
The storage means stores the probability value indicating the probability that an obstacle does not exist or the probability value indicating the probability that an obstacle exists or does not exist. In the present invention, when the existence or non-existence of an obstacle is represented by a probability, a so-called Bayes probability can be applied. However, if the expression is represented by a Bayes probability, the unknown cannot be accurately represented. On the other hand, it is effective to apply the Dempster-Shafer probability including the unknown as an element of the probability because the unknown can be accurately expressed. The first probability updating means updates the probability value stored in the storage means based on the output of the obstacle sensor. Thereby, the probability value is updated when the obstacle sensor measures. The second updating means is configured to store, based on the displacement and azimuth change of the vehicle obtained from the output of the movement amount output means, the probability of the small area before the change in the displacement and azimuth of the vehicle stored in the storage means. The value is updated to the probability value of the small area after the displacement of the vehicle and the change of the azimuth have occurred. As described above, since the probability value is updated based on the change in the displacement and the azimuth of the vehicle, even when the vehicle moves, the probability value when the small area is always located at a specific position viewed from the reference of the vehicle is stored. Can be done. Further, by using the output of the obstacle sensor output at a time interval, a probability value obtained by integrating the stored probability value and the probability value obtained from the obstacle sensor output can be stored in the storage unit. The stored probability values serve as more reliable information on the existence or non-existence of an obstacle, reflecting the history observed in the past. That is, the accuracy is increased because the obstacle is present at the same place or is expressed as an overall probability with reference to the result of the past observation that the obstacle was not present. Therefore, obstacles can be correctly recognized even in situations where oversight and misidentification cannot be avoided with only a single sensor signal, and a more reliable system can be used in combination with obstacle display devices and alarm devices. Can be configured.

As described above, according to the present invention, it is possible to perform more accurate display and alarm as compared with the case where an obstacle is displayed and alarmed based on a single sensor signal. Thus, a highly reliable system for safety can be constructed.

[0011]

[Description of Other Inventions] Next, other inventions will be described while explaining in detail the principle of the present invention (first invention) configured as described above. When there is an obstacle around the vehicle, the obstacle sensor cannot always detect the obstacle. The obstacle sensor detects an obstacle in a limited area and outputs a signal, but the output is not always correct. However, a more accurate judgment can be expected if the judgment is made based on the results of several observations by the obstacle sensor. FIG. 3 shows a simple example. When the vehicle A observes an obstacle sensor provided with an isosceles triangular detection area while moving, no obstacle is detected by one observation. If not, it is not enough to judge that there is really no obstacle for the above reason. However, if it is possible to obtain an observation result that an obstacle does not exist a plurality of times, for example, three times or more, it can be considered that it is almost safe to determine.
At this time, depending on the position, the area E ₃ observed three times by the obstacle sensor, the area E ₂ observed twice, and the area E observed once
_1. Observation of an obstacle can be divided into an area E ₀ where the observation is performed 0 times, and the degree to which safety is confirmed differs for each area. In FIG. 3, a region E ₃ shown by oblique lines is a region in which the results of three times observed along with the movement shown by the arrow are integrated and can be determined to be safe. When judging whether or not an obstacle exists at a certain position by integrating a plurality of obtained outputs of the obstacle sensor, the ambiguity of the judgment differs depending on the position. To express this accurately, depending on its position,
A method of stochastically expressing an event that an obstacle exists or does not exist is effective.

Now, the obstacle detection output of a certain obstacle sensor is considered as stochastic information on the presence or absence of an obstacle in the detection area of the obstacle sensor, and the output of another obstacle sensor (the same obstacle) is considered. Output at different times using an object sensor,
Alternatively, by taking into account the output of the detection area of the different sensors) and stochastically expressing it and determining the presence or absence of an obstacle with respect to the detection area, based on the output of a single obstacle sensor You will be able to make more reliable and reliable decisions than if you decide. In addition, it is possible to show the certainty of the basis of the judgment.

If a plurality of observation results are processed and stacked in this way, even the information of the obscure obstacle sensor becomes more and more reliable, so that it is expected that a reliable safety judgment can be made. it can. The point of the present invention is, with respect to the output of an obstacle sensor attached to a moving vehicle,
An object of the present invention is to apply a technique for improving reliability as described above.

The problem with applying the above technique is that the obstacle sensor attached to the vehicle moves with the movement of the vehicle. In general, they move with respect to the ground with the passage of time, and do not match the detection areas of the obstacle sensors at other points in time and the other obstacle sensors. It is impossible to match.

To solve this problem, the following (1) to
The method (4) is adopted. (1) As shown in FIG. 4, a region R around the vehicle A is divided into a plurality of small regions r, for example, in a lattice shape.

(2) The information on the presence / absence of an obstacle output from the obstacle sensor is used as a probability that an obstacle exists in each of the small areas and a probability that no obstacle exists.

(3) Probability that an obstacle exists for each small area,
A probability value representing the probability that an obstacle does not exist is stored in a format that can be compared with a small area.

When information on the presence / absence of an obstacle output by the obstacle sensor is newly obtained, the information on the presence / absence is used as a probability that an obstacle exists for each small area and a probability that an obstacle does not exist. With reference to the probability value stored in the small area, the probability of the existence of the obstacle, the probability of the absence of the obstacle, or the probability of both of them obtained by integrating the outputs of the obstacle sensors for the small area obtained so far is calculated. Calculate and re-store that probability value.

The reliability of obstacle detection is improved by successively repeating the above processing. FIG. 5 shows an example of the processing, in which a probability value indicating whether or not an obstacle is likely to exist for each small area is stored, and the probability value is calculated each time a new obstacle sensor observation result is obtained. By repeating the process of updating the value, a highly reliable recognition result of the presence of the obstacle can be stored.

When the stored probability value is updated based on the new observation result of the obstacle sensor, the relationship between each small area corresponding to the stored probability value and the detection area of the obstacle sensor is always determined. It needs to be clarified. That is, it is necessary to clarify the relationship between the detection area of the sensor and the stored small area, which changes with the movement of the vehicle. This can be realized if the amount of movement of the vehicle is constantly grasped.

By the way, as a method of obtaining a small area for storing a probability value, (a) a method of taking a small area fixed to the ground, and (b) a small area having a fixed positional relationship with the vehicle. There are two methods, a method of taking an area. In the method (a), it is necessary to prepare a storage area for storing the probability value for the entire range in which the vehicle moves, and there is no problem for a vehicle having a limited movement range. The area is enormous and not practical. Also,
The method (b) is always stored at a position with respect to the vehicle, and it is sufficient to secure a storage area of a size necessary for determining safety. Therefore, the method (b) is different from the method (a) described above. There is an advantage that the storage area can be small, and it is easy to correspond to the output of the obstacle sensor fixed to the vehicle. When the method of (b) is adopted as a method of taking the small area,
Assuming that the obstacle does not move with respect to the ground, the information about the obstacle stored for the small area fixed in the positional relationship with respect to the vehicle is not changed by the movement of the vehicle. It will show obstacle information about the area. In other words, an obstacle that does not move with respect to the ground has moved in an area on the basis of the vehicle. Therefore, among the stored probability values, it is necessary to move the probability value of a small area to the corresponding small area probability value in accordance with the movement of the vehicle. That is, as shown in FIG. 6, the vehicle is facing the direction of the azimuth θn at the position of the coordinates (Xn, Yn) at a certain time, and the coordinates (Xn + 1,
When the vehicle moves in the direction of the azimuth θn + 1 at the position of (Yn + 1), the probability value of the position of the coordinates (X, Y) needs to be expressed as a probability value of a different position when the vehicle moves in the coordinate system fixed to the vehicle. is there. The process of moving the probability value due to the movement of the vehicle can be expressed by a combination of so-called translation and rotation in the geometric conversion process, and an affine transformation often used in so-called image processing may be performed. That is, by treating the probability value for each small area as two-dimensional data as in the case of image processing, affine transformation can be realized at high speed using a processor for image processing. Another point of the present invention is that the probability value before the change of the vehicle displacement and the azimuth stored in the storage means before the change of the vehicle is generated based on the change of the vehicle displacement and the azimuth based on the affine transformation, for example. Is updated to the probability value of the small area after the displacement and the change of the azimuth have occurred.

When the existence of an obstacle is represented by a probability, it is common to use a so-called Bayesian probability. However, the Bayesian probability is represented by an unmeasured area, that is, an unknown area. However, there is a disadvantage that it is not accurate. That is, in the Bayesian probability, the unknown area is expressed as 0.5 with the probability that an obstacle is present and 0.5 with the probability of not being an obstacle, so it is difficult to intuitively grasp how highly reliable the measurement was performed. There is a problem. Rather, it is easier to understand if both the existence probability and the non-existence probability are expressed as zero. Since the probability is a constraint that the sum of the probability values of all events is set to 1, Bayesian probability cannot be accurately represented because the unknown region is equally allocated to the existence and the non-existence. On the other hand, if an event that exists or is not present is prepared, and if unknown, the probability of Dempster-Shafer assigning 1 to this event can be accurately expressed. That is, using the Dempster-Shafer probability instead of the general probability, the output of the obstacle sensor is used as a probability value of three events of the presence, absence, and unknown (existence or absence) of the obstacle. By expressing, it is possible to more accurately recognize the presence of an obstacle. In addition, since there is a restriction that the sum of the probability values of the three events is 1, it is sufficient that the probability values of the two events are substantially known.

According to the second invention based on the above principle, FIG.
As shown in the figure, an obstacle sensor 10 for detecting an obstacle around the vehicle, and a movement amount output means 40 for outputting a movement amount of the vehicle.
And the probability that an obstacle exists or does not exist in a small area having a fixed positional relationship with respect to the vehicle is stored as a probability value corresponding to the small area, and the probability value is read and written according to an input. A storage unit 210, a probability value deriving unit 202 that receives an output of the obstacle sensor 10, and outputs a probability value for each of a plurality of small regions in a fixed positional relationship with respect to the vehicle based on the input; A probability value that is connected to the probability value deriving means 202 and the storage means 210 and that the obstacle is present or absent in the small area output by the probability value deriving means 202 and a probability value read from the storage means for the corresponding small area; A probability calculating means for calculating a probability value obtained by taking both into account and outputting a write signal to the storage means so as to write this value as a probability value for the corresponding small area. And 203 are connected to the movement amount output unit 40 and the storage unit 210, the information of the change of the displacement and the azimuth of the vehicle obtained from the output of the movement amount output unit 40,
A conversion parameter is calculated based on a predetermined formula, a probability value stored in the storage unit 210 is read, and a value obtained by performing an affine transformation according to the calculated parameter is written in the storage unit so as to be written as a storage value. An affine transformation unit 205 that outputs a signal, calculates the change in the probability value of each small area with the movement from the probability value of the presence of an obstacle in another area, and updates the probability values of all the small areas, It is composed of

According to the second aspect, the probability value corresponding to each small area stored in the storage means is affine-transformed based on the signal of the movement amount measurement means, and as a result, the stored probability value corresponds to the probability value. The small area is always an area at the same position with respect to the vehicle regardless of the movement of the vehicle. Therefore,
The output of the in-vehicle obstacle sensor output by the probability derivation means outputs a probability value regarding the presence or absence of an obstacle for each small area around the vehicle and a small area corresponding to the probability value stored in the storage means. However, the same area can always be set at a specific position viewed from the vehicle reference. Therefore, the output of each obstacle sensor obtained after the passage of time is stored in the storage means which stores the result obtained by integrating the outputs of the obstacle sensors obtained so far for the area by the probability calculation means as a probability value. The existence probability of the obstacle in the small area which is integrated with the value and reflects the output of the new obstacle sensor can be stored in the storage means. In other words, the value stored by the storage means becomes more reliable probability information on the existence of an obstacle reflecting the history of the result measured by the obstacle sensor in the past, and the presence or absence of the obstacle was in the same place. A highly reliable probability value is obtained as the probability of comprehensive existence with reference to the result of the past observation. Therefore, even in situations where oversight and misunderstanding cannot be avoided with only a single sensor signal, correct judgment can be made.Using this obstacle recognition device as an element of an obstacle display device or an alarm device, A highly reliable system can be configured.

In the third invention, as shown in FIG.
An obstacle sensor 10 for detecting an obstacle around the vehicle, a movement amount output means 40 for outputting a movement amount of the vehicle, and a probability and existence of the obstacle in a small area having a fixed positional relationship with respect to the vehicle. The storage means 210 stores a probability value representing a probability of not being performed as a probability value corresponding to a small area, and reads and writes the probability value in accordance with an input.
A probability value that outputs the probability value representing the probability that an obstacle exists for each of a plurality of small regions that are in a fixed positional relationship with respect to the vehicle based on this input and the probability that an obstacle does not exist The deriving means 202, the probability value deriving means 202 and the storage means 210 are connected, and the probability value deriving means 2
02, a probability value indicating a probability that an obstacle exists in the small area and a probability value indicating that no obstacle exists in the small area, and a probability value read from the storage unit with respect to the corresponding small area, an obstacle obtained by taking both into account is obtained. A probability calculating means 203 for calculating a probability value representing the existence probability and the non-existence probability and outputting a write signal to the storage means so as to write this value as a probability value for the corresponding small area; 40
The conversion parameter is calculated based on a predetermined equation from information on the displacement of the vehicle and the change in the azimuth output from the movement amount measurement output stage 40, and the value stored in the storage means. And outputs a write signal to the storage means 210 so as to write a value obtained by performing an affine transformation according to the calculated parameter, and changes in the probability of the existence of the obstacle in each small area due to the movement. Affine transformation means 205 which calculates the probability of the presence of obstacles in all the small areas by calculating from the probability of the presence of obstacles in other areas
And, it consists of.

When the existence of an obstacle is represented by a probability, if it is represented by a so-called Bayesian probability, there is a drawback that the unknown cannot be accurately represented. On the other hand, it is effective to apply the Dempster-Shafer probability including unknown as a probability element. According to the third aspect, the probability stored in the storage means can have two values, that is, the probability that an obstacle exists and the probability that it does not exist. Both can be represented by 0.
Therefore, the probability value corresponding to the small region of the region that the obstacle sensor has never observed is 0 for both the probability of existence and the probability of non-existence, and it is accurate that the blind spot of the obstacle sensor was obtained. Can be expressed as

As described above, the display of an obstacle based on a single sensor signal and the more accurate display as compared with the case where an alarm is issued,
Since an alarm can be issued, a highly reliable system for safety can be constructed. Also, unknown areas,
It is important as an area where safety has not been confirmed.
In this invention, this can be positively recognized.

[0028]

DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, the present invention will be described in detail based on specific embodiments. FIG. 7 is a block diagram showing a configuration of a vehicle to which the obstacle recognition device according to one specific embodiment of the present invention is applied. The vehicle (for example, an automatic guided vehicle) drives the left and right drive wheels 50 and 51 to perform traveling such as straight traveling, turning, and spin turning. An obstacle sensor 10 is attached to this vehicle, and the output signal is processed by the arithmetic unit 20 to make a more reliable safety judgment. On the other hand, the arithmetic unit 20 is connected to encoders 52 and 53 for measuring the amount of rotation of the left and right drive wheels, so that the amount of movement of the vehicle can be grasped. It is also connected to a CRT 61 so that the situation of obstacles around the vehicle can be monitored on the image plane. The vehicle has driven wheels 60, 61 that can rotate freely.
Is attached.

The arithmetic unit 20 includes a light projector 100 for projecting a light beam and a light receiver 10 having sensitivity to light.
1 and an obstacle sensor 10 composed of a sensor driver 102 for driving the light emitter 100 and the light receiver 101 are connected. The light emitter 100 and the light receiver 101 of the obstacle sensor 10 are attached at a predetermined distance in the vehicle width direction at a position in front of the vehicle in the traveling direction, and the arithmetic unit 20 uses the output of the obstacle sensor 10 as a reference for the vehicle. Avoid collision with obstacles by controlling speed.

FIGS. 8 and 10 show the obstacle sensor 10 and
3 shows details of a detection area of the obstacle sensor 10.
The floodlight 100 of the obstacle sensor 10 is as shown in FIG.
And 21 light-emitting elements E1 to E21 in an array.
The light receiver 101 has 24 arrays of light receiving elements R1 to R1.
R24. The detection area ahead in the traveling direction is
504 where the light projecting direction of 21 intersects the light receiving direction of 24
Intersection area (A_1,1~ A _{24, 21}504 areas)
Light emission of the projector 100 by the sensor driver 102
Control the switching between the direction and the light receiving area of the light receiver 101.
To detect the presence or absence of obstacles in all intersection areas.
It is configured as follows. That is, the sensor driver 102
Therefore, the light emitting element E₁~ E_{twenty one}Any one of the light-emitting elements
From the light receiving element R₁~ R_{twenty four}Reflected light
Judgment as to whether or not light is received is required for all light emitting elements.
Obstacles in all intersection areas
Or not.

FIG. 9 shows the principle of measurement by the obstacle sensor. An intersection area (shown by oblique lines) where the light beam emitted by the light emitting element of the light projector 100 and the light receiving area of the light receiving element of the light receiver 101 intersects. When a _{1, 1} to a _{24, 21} obstacles to any one region) is present in the area, the light beam emitting element is projected light projector 100 is reflected by the obstacle, the reflected light photodetector 101 Is observed in the light-receiving element. In contrast,
When no obstacle exists, the light beam emitted by the light emitting element of the light projector 100 is not observed in the light receiving element of the light receiver 101. Therefore, the presence or absence of an obstacle in the intersection area can be detected by determining whether or not the reflected light has been observed. The obstacle sensor 10 according to the present embodiment detects an obstacle based on this principle, and the light beam emitted from the light emitting element of the light projector 100 is reflected by the obstacle and detected by the light receiving element of the light receiver 101. At this time, the sensor driver 102 outputs to the arithmetic unit 20 a digital signal indicating the area number of the intersection area where the obstacle is detected. This area number is 1 to
A numerical value of 504 or a number of (1, 1) to (24, 21) can be used. The numbers of the light emitting elements and the light receiving elements are not limited to the above numbers, but can be increased or decreased as needed.

The arithmetic unit 20 receives a signal from the obstacle sensor 10 and functions as a signal processing unit of the present embodiment. FIG.
FIG. 1 is a block diagram showing details of functions of the arithmetic device 20. The signal from the obstacle sensor 10 is a number (1 to 1 for the projector) specifying an area when an obstacle is detected in each intersection area between the light projecting beam and the light receiving area.
21 (a pair of numbers 1 to 24 for the light receiver) are provided as output. This signal is output to the probability value deriving means 2
02. The probability value deriving means 202 performs the following two processes (1) and (2). (1) The entire intersection area between the light beam and the light receiving area of the obstacle sensor is classified into three areas: an area where an obstacle is likely, an area where an obstacle is unlikely, and an unknown area. (2) From the result of the above (1), a small area described in the storage means (hereinafter referred to as a cell, but does not coincide with the intersection area)
The existence probability value and the non-existence probability value of each obstacle are set. FIG. 12 shows an example of the positional relationship between the intersection area and the cell.

FIG. 13 shows the classification of each intersection region in the above (1) into the region of the intersection. That is, the area output by the obstacle sensor is classified into the area, and the area on the same light beam and closer to the sensor than the area where the obstacle is detected is classified into the area, and the area Alternatively, an area on the light receiving area that is farther than the area where the obstacle is detected is classified into an area. In the case where no output was obtained in any area on a certain light beam, all the areas on the light beam were classified into the area. Since the number of each intersection area is predetermined as shown in FIG. 10, the classification of (1) can be easily performed by comparing the sizes of these numbers. The number of each intersection area may be set such that the number increases as the distance between the light emitting beam and the light receiving area increases.

Next, the probability value deriving means 202:
The probability value corresponding to the small area stored in the storage means 210 is set based on the result of the classification into the area. First, the small area stored in the storage means 210 is divided as shown in FIG.

That is, 12.8 m with respect to the vehicle
The four areas are defined as 512 × 512 square small areas (cells
(25 mm x 25 mm). Based on the classified results, the setting of the probabilities for these cells with respect to the sensing results is performed as follows. The position of each intersection area of the obstacle sensor with respect to the vehicle is measured in advance, and the correspondence between these intersection areas and cells for storing probability values is stored in the probability value deriving means 202 as a table shown in FIG. Keep it. In addition, the reliability (probability value) for the detection result of each intersection area is similarly stored in this table.

The setting of the probability value of each cell by the probability value deriving means 202 is performed according to the following procedures (1) and (2).

(1) Initialization (existence probability Psce of each cell)
Is set to 0 and the non-existence probability Pscn is set to 0.) (2) With reference to the table of FIG. 14 and the intersection areas classified into the areas, the cells included in each intersection area are obtained, and the following expression is obtained. Therefore, the value obtained by the operation is set in each cell.

Cell Psce = Ps / N, Pscn = 0, where the intersection area is classified as an area Cell Psce = 0, Pscn = Ps, where the intersection area is classified as an area Psce = 0, Pscn = 0, where Ps: reliability of the detection result of the intersection area,
N: Total number of cells included in the intersection area.

Then, the probability value deriving means 202 outputs the probability value of each cell set as described above.

The probability calculating means 205 includes a probability value deriving means 2
02 output by each cell (existence probability Psc
e, non-existence probability Pscn) and the probability value (existence probability Pce (n), non-existence probability Ps) for each cell stored in the storage means
cn (n)) and the probability values (existence probability Pce (n + 1), non-existence probability Pcn (n)
+1)) is calculated and stored in the storage location of the corresponding cell in the storage means. These operations are performed for each cell, and
The integration of the probability values is performed by the arithmetic unit 20 according to a program based on the following equation.

[0041]

(Equation 1)

On the other hand, the moving amount output means 203 counts the pulses output by the encoders 52 and 53 attached so as to measure the rotation of the left and right driving wheels, and obtains the moving amount of the wheel every predetermined time. That is, the displacement Δ1,
The azimuth change Δθ is calculated and output.

That is, Δθ = θ (n + 1) −θ (n) = (Dl−Dr) / Tread Δ1 = (Dl + Dr) / 2 where Dl and Dr are the movement amounts of the left and right drive wheels, Tre
ad is the distance between the left and right drive wheels. If the arithmetic unit 20 outputs a wheel movement amount command value, the movement amount can be output based on the command value, so that the movement amount output unit 203 can be used.

The affine transformation means 205 refers to the movement amount output from the movement amount output means 203 and the probability value of each cell stored in the storage means 210 to find a small area around the vehicle that changes with movement. The movement of the probability value of the corresponding cell is obtained by calculation, and this probability value is updated. In general, the movement of the position of a certain point by traveling expressed by a combination of rotation and translation movement is given by the following equation.

[0045]

(Equation 2)

That is, the position of the point (X (n), Y (n)) in the vehicle-fixed reference coordinate system at a certain point in time is changed in the vehicle-fixed reference coordinate system after the vehicle has changed with running. (X
(N + 1), Y (n + 1)). Conversely, the probability value of the cell at the position of the point (X (n + 1), Y (n + 1)) after the movement corresponds to the probability value of the position of the point (X (n), Y (n)) before the movement. . The affine transformation unit refers to the probability value of each cell after movement with reference to the movement amount output by the movement amount output unit 203 to find a corresponding cell before movement, and moves the probability value to the corresponding cell after movement. The process to be performed is executed for all the cells after the movement. In this embodiment, in order to increase the accuracy of the conversion, the probability value of the point at the position (X (n), Y (n)) before the movement is obtained by interpolating the probability values of the four surrounding cells with four points. Increased accuracy. This processing is performed using a dedicated processor called a so-called DSP independently of the CPU. 5
The data of 12 × 512 cells is stored in a storage area that can be processed by the DSP, and an affine transformation coefficient corresponding to the moving amount of the vehicle is set. can do.

The storage means 210 includes an existence probability storage means 211 for storing a probability value of an obstacle for each cell and a non-existence probability storage means 212 for storing a non-existence probability value. Updating the probability value based on the output of the object sensor 10 is a program executed by the CPU,
Since the process of updating the value by affine transformation based on the output of the moving amount output means is executed by a so-called DSP, a memory that can be accessed at high speed by both is used.

Further, in the embodiment, in order to display a recognition result indicated by the probability value, a display is connected to a so-called frame memory (image memory), and the contents of the memory are stored in a CRT.
30 is displayed in color so that it can be confirmed.

FIG. 15 shows an example of a hardware configuration of an arithmetic unit for performing the above processing. This hardware includes a counter for inputting signals from encoders 52 and 53 attached to the left and right driving wheels and capturing the rotation amount of each wheel at predetermined time intervals to a CPU, and a PIO for capturing the output of an obstacle sensor. Board 40, DSP board 50 for executing affine transformation at high speed, image memory board 6 for displaying the stored probability value of each cell
0, probability value deriving means 202, probability calculating means 20
3. Movement amount measuring means 40, CPU board 7 for executing each program for controlling affine transformation by DSP
0. As hardware for realizing the storage means 210, r of the color image memory board 223 is used.
The ed pixel (red pixel) is assigned to the obstacle existence probability, and the green pixel (green pixel) is assigned to the obstacle absence probability. Thus, on the CRT screen, the area where an obstacle is likely to be red is displayed, the area which is likely to be safe is displayed in green, and the unclear area is displayed in black, so that a display which is intuitive to humans can be performed.

These boards are also called VME
Operating on the bus, the CPU and DSP can access the memory of the image memory board. The acquisition of the signal of the obstacle sensor and the affine transformation corresponding to the movement are performed asynchronously, and each program of the probability value deriving means, the probability value calculating means, the moving amount output means, and the control of the affine transformation by the DSP is: So-called real-time O
It is executed while communicating with each other under the management of S (operating system).

According to the above configuration, the observation result obtained by the obstacle sensor is converted into a probabilistic representation of whether or not an obstacle is present in accordance with the position represented by the vehicle-based coordinate system. By storing the probability value obtained by integrating the data of the same probability expression obtained up to the observation point and the newly obtained data stored in the A highly reliable recognition result reflecting the history can be obtained. Also, regardless of movement, the stored probability value is always in the vehicle-based coordinate system,
A change in the movement and direction of the vehicle is obtained from the rotation amounts of the left and right wheels, a translation parameter for translational rotation of a probability value associated with the movement is set, and this can be performed at high speed. By storing this probability value in the image memory, it becomes possible to always display the recognition result of the surrounding obstacles on the CRT based on the vehicle. In the above-described embodiment, the description has been given of the vehicle that realizes the steering by the rotation difference between the left and right drive wheels. However, the same configuration is also possible in a four-wheel configuration having so-called automobile-type steered wheels. That is, as shown in FIG. 16, the moving amount D (ZL) H means 40
A goniometer 80 for measuring the steering angle δ of the steered wheels, and encoders 90 and 9 for measuring the moving amounts Dl and Dr of the left and right rear wheels.
The same effect as in the above embodiment can be obtained by using a calculation means for calculating the displacements and changes in azimuth angle Δ1 and Δθ in the same manner as in the above embodiment by using the following formulas.

Δ1 = (Dl + Dr) / 2 Δθ = θ (n + 1) −θ (n) = Δ1 · tan (δ) / WB WB: Distance between front and rear wheels A three-wheel vehicle in the same manner as described above. Needless to say, the present invention can be easily applied to such a case.

Further, in the above embodiment, an example is shown in which a sensor using light is used as the obstacle sensor, but the present invention is not limited to this sensor. The probability value deriving means of the present invention can be configured according to the type of the sensor, as long as the detection range can be specified for all of the obstacle sensors.
FIG. 17 shows an example of how to assign probabilities in the probability value deriving means in the case of an ultrasonic sonar as another example of the sensor. That is, the sonar outputs the distance to the point where the emitted ultrasonic wave returns in the shortest time. In addition, the spread of the emitted ultrasonic waves can be measured in advance.
Interpretation of the detection area according to the extent as shown in Fig. 7 where there is an obstacle somewhere in the horizontally long area according to the distance resolution, the area closer to it has no obstacle, and the area farther away is unknown. Can be. It is obvious that probability value deriving means can be similarly configured based on this interpretation. In addition, it is obvious that the present invention can be similarly applied to a microwave radar and a laser radar other than the ultrasonic sonar.

In the above-described embodiment, an example has been shown in which the advantage is exhibited by displaying a highly reliable recognition result to a human. Based on the recognition result, a collision with an obstacle is determined. If an alarm is issued to avoid this, it is possible to provide a reliable alarm with less false alarm. That is, the course of the vehicle is predicted, and whether or not there is an obstacle in the predicted area is determined based on the probability of the existence of the obstacle and the probability of non-existence for each small area (cell) stored in the storage means included in the course. Therefore, the calculation can be performed with high reliability. That is, the fact that the traveling course is safe can be represented by the probability that no obstacle exists in all areas in the course, and can be represented by the minimum value of the probability that no obstacle exists in that area. On the other hand, when observations are insufficient, it can be said that there is an obstacle, even if the probability of the presence of an obstacle in each small area and the probability of the absence of an obstacle being low are not safe. There is no need to stop, so it is sufficient to observe slowly. Conversely, it is necessary to stop when it is clearly determined that an obstacle is present, which is when there is a high probability that an obstacle exists somewhere in a small area in the traveling course. You can think. The traveling course can be predicted from the vehicle speed, the steering angle, and the like. In addition, if this course is determined, the small area included therein can be specified, so that the probability that no obstacle exists in the entire traveling course and the probability that the obstacle exists somewhere in the traveling course can be calculated. Means If there is a sufficiently high probability that there will be no obstacles on this course, there will be no alarm; if there is no obstacle on the course, there will be no warning. If the value is higher, an alarm such as a stop can be output with high reliability. Therefore, if the recognition device of the present invention is connected to an alarm device, it is possible to output an alarm capable of appropriately responding to the situation without false alarm.

[0055]

As described above, according to the present invention, at least one of the probability values regarding the presence and absence of an obstacle is updated based on the output of the obstacle sensor and the output of the moving amount output means. Therefore, it can be expressed as a comprehensive probability by reflecting the results of past observations, and this can reduce the frequency of unavoidable oversight and misidentification of obstacles and perform accurate obstacle recognition The effect is obtained.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a second invention.

FIG. 2 is a block diagram showing a third invention.

FIG. 3 is a diagram showing a state where judgment is made based on several observation results of an obstacle sensor.

FIG. 4 is a diagram showing a number of small areas provided around the vehicle.

FIG. 5 is an explanatory diagram illustrating a state in which a stored probability value is updated.

FIG. 6 is a diagram showing a state in which a probability value is affine-transformed.

FIG. 7 is a block diagram showing a vehicle according to the first embodiment of the present invention.

FIG. 8 is a diagram showing an obstacle sensor and a detection area.

FIG. 9 is a diagram illustrating the measurement principle of the obstacle sensor.

FIG. 10 is a diagram showing details of FIG. 8;

FIG. 11 is a functional block diagram showing an arithmetic unit.

FIG. 12 is a diagram showing a relationship between a cell and an intersection area.

FIG. 13 is a diagram showing an existing area, an non-existent area, and an unknown area of an obstacle.

FIG. 14 is a diagram showing a table indicating a relationship between an intersection area and a cell;

FIG. 15 is a block diagram showing a configuration of hardware for performing color display.

FIG. 16 is a block diagram similar to FIG. 7 of a four-wheel vehicle.

FIG. 17 shows the presence of an obstacle by the ultrasonic sonar;
FIG. 3 is a diagram showing a non-existent area.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 10 Obstacle sensor 40 Moving amount output means 30 Display device 31 Alarm device

──────────────────────────────────────────────────続 き Continuation of front page (56) References JP-A-3-164346 (JP, A) JP-A-59-109884 (JP, A) JP-A-62-79385 (JP, A) JP-A-6-79385 150195 (JP, A) (58) Fields investigated (Int. Cl. ⁷ , DB name) G01S 17/00-17/95 G01S 13/93 G01S 15/93

Claims (1)

(57) [Claims] 1. An obstacle sensor for detecting an obstacle around a vehicle, a movement amount output means for outputting a movement amount of the vehicle, and an obstacle around a vehicle divided into a plurality of small regions. Storage means for storing a probability value representing at least one of a probability that an object exists and a probability that an obstacle does not exist, and first probability update means for updating the probability value stored in the storage means based on an output of the obstacle sensor Based on the displacement and azimuth change of the vehicle obtained from the output of the movement amount output means, the probability value of the small area before the change of the displacement and azimuth of the vehicle stored in the storage means is generated. Second updating means for updating the probability value of the small area after the change of the azimuth angle has occurred, and an obstacle recognition device for a vehicle.