JPH07121235A - Position self-recognizing method for moving robot - Google Patents

Abstract

PURPOSE:To enable the robot to recognize its position correctly at all times even when performing operation accompanied by long-distance movement. CONSTITUTION:The robot moves while recognizing its position, and also predicts an error in position recognition at the same time (step 1). An image of a peripheral body is stored and the recognition position is stored at the same time (steps 2 and 3). When the predicted error reaches a reference error, the position which is closest to a current position among points where the peripheral body is stored before is selected (steps 4 and 5). The image of the peripheral body which is stored at the selected point is moved in the direction wherein the difference from the peripheral state of the current point decreases (step 6). When the selected position is reached, the recognition of the position of the robot is replaced with the position recognition of the selected position.

Description

Detailed Description of the Invention

[0001]

The present invention relates to a mobile robot (hereinafter,
Simply referred to as a robot), and a method of recognizing its own position in an area where the robot moves.

[0002]

2. Description of the Related Art When a robot moves in an environment and performs various tasks, it is necessary to always recognize its own position in the environment. As a method to enable this recognition,
Conventionally, there is dead reckoning. Dead reckoning means that when the robot starts from a certain reference point and moves, for example, if it moves by wheels, the value obtained by multiplying the rotation speed of the wheel by the circumference of the wheel is regarded as the movement distance, and the reference point This is a method of recognizing one's position with respect to. If the direction is changed during the movement, the rotation angle can be known based on the difference between the rotation speeds of the left and right wheels, for example. The robot can always recognize its own position by accumulating the rotation angle and the moving distance each time the direction is changed.

[0003]

However, position recognition based on dead reckoning has a problem in that an error increases as the robot moves. That is, in dead reckoning, it is necessary to first measure the circumferential value of the wheel in advance, but the measured value itself may include an error, and the moving surface and the wheel may slip when the robot moves. The error in position recognition caused by these causes is very small if the moving distance is short,
If the moving distance becomes long, the error gradually accumulates and becomes a non-negligible error. Therefore, if the robot tries to recognize its own position based on dead reckoning when moving in the environment and performing various tasks, position recognition becomes inaccurate in tasks involving long-distance movement. There is a problem that it becomes difficult to carry out work.

An object of the present invention is to solve the above-mentioned problems caused by the conventional method for recognizing the self-position of the robot, and to always correctly recognize the self-position even when the robot performs work involving long-distance movement. The object of the present invention is to provide a method for recognizing the position of a mobile robot, which makes it possible to perform work as a result.

[0005]

A method of recognizing a self position of a mobile robot according to the present invention moves while recognizing a rough self position by using dead reckoning, and at the same time, an error in position recognition caused by dead reckoning. And when a predetermined point is reached during the movement, the image of the object existing around itself is projected on a virtual circumference centered on the point, and the image of the projected object is projected. The step of storing the information that characterizes, and at the same time, the position recognized by dead reckoning, and when the prediction error of position recognition increases to a predetermined reference value, the amount that characterizes the image of the surrounding object previously. Of the points stored in, selecting the closest point to the current point,
Repeating the operation of moving a small distance in a direction in which the difference between the information at the selected point and the information at the current point decreases until the difference becomes zero, and when the difference becomes zero, Recognizing the location of that point as the stored location of the selected point.

[0006]

A feature of the present invention is that, as a method for the robot to recognize its own position in the environment, in addition to the dead reckoning described above, the relative positional relationship between the surrounding objects and itself is used. . Here, the relative positional relationship between the objects existing around the robot and the robot refers to the position of the object with respect to the robot at each time point, such as the distance and direction of each object viewed from the robot. Regarding the use of the relative positional relationship between the surrounding objects and the robot,
The present invention has the following two features.

First, the use of this relative positional relationship stores the images of surrounding objects projected by a robot on a virtual circumference centered on itself at various points in the environment. This is done by comparing the stored image with the current surrounding conditions.

Secondly, when comparing the already stored image of the surrounding object with the current surrounding situation, first, the stored image at the point closest to the current position is selected, and the image and the surrounding situation are selected. If the two are in agreement with each other by moving in a direction in which the difference between and decreases, the position of the point where the image is stored is recognized as the current position.

Of the above-mentioned two features concerning the use of the relative positional relationship between the surrounding objects and the self, the more important feature is the second feature. In other words, if the robot's surroundings move to a point that matches the previously stored image of the surrounding object, the position recognition error due to dead reckoning will occur at the point where the image of the surrounding object was previously stored. The error included in the position recognition can be reduced. Therefore, by appropriately performing the operation indicated by the second feature in the process of the robot moving in the environment, it is possible to prevent the position recognition error caused by dead reckoning from accumulating, and as a result, the position recognition of the robot is performed. It is possible to always keep the included error to a small value. The first feature enables the robot to store the surrounding situation with a small amount of memory even in a complicated environment. As a result, the present invention can be easily implemented.

As information representing the relative positional relationship between the object existing in the surroundings and the robot, it is the angle formed by the viewing angle of the projection viewed from the center of the circle and the reference direction of the bisector of the viewing angle. Directional angles can be used.

In this case, the projections on the virtual circumference at the selected point are associated with the projections on the virtual circumference at the current point that have the closest directional angles. Then, for each projection on the virtual circumference at the current point, the direction is the direction indicated by the bisector of the viewing angle when the projection is viewed from the center of the circle, and the direction is the viewing angle of the projection, A diameter vector in which the difference between the corresponding viewing angles on the virtual circumference at the selected point is in a decreasing direction, and the absolute value of the magnitude is proportional to the difference between the two viewing angles,
The direction is orthogonal to the bisector of the viewing angle when the projection is viewed from the center of the circle, and the orientation is the direction angle of the projection and the corresponding direction angle on the virtual circumference at the selected point. The angle vector is such that the difference between the two decreases and the absolute value of the magnitude is proportional to the difference between the two directions. In this way
A distance vector proportional to the size of the combined vector in the direction indicated by the combined vector, by combining the diameter vector and the angle vector obtained for all projections on the virtual circumference at the current point. Just move.

[0012]

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

FIG. 1 is a flow chart showing a method for recognizing a self position of a mobile robot according to an embodiment of the present invention, and FIG.
It is a detailed flowchart of step 6 in the inside.

Regarding the configuration example of the robot, the previously filed “method for learning environment of mobile robot” (Japanese Patent Application No. 4-
No. 66317).
A description of this configuration example is omitted. It is assumed that the robots in the following embodiments further have a gyro compass in addition to this basic configuration example, and can always recognize a certain direction (for example, north) by this gyro compass.

FIG. 3 is a view from above showing how the robot moves in the environment bounded by the boundary line 21. Objects 22-26 are present in the environment. The robot 27 departs from the point A and moves in the environment. The robot 27 recognizes the position in the environment using the starting point A as a reference point (for example, the robot 27 recognizes its own position as a coordinate value on a two-dimensional orthogonal coordinate axis, and regards the point A as the origin of the coordinate axis). . The robot 27 moves while recognizing its rough position using dead reckoning,
At the same time, a position recognition error caused by dead reckoning is predicted (step 1). The expected value of this error is hereinafter referred to as the expected error. The prediction error increases in proportion to the moving distance (the proportionality coefficient is determined by measuring the relationship between the moving distance and the error in advance).

The robot 27 stores images of surrounding objects each time it passes through each of the points B to G while moving, and at the same time stores the recognized position (based on dead reckoning) at each point ( Steps 2 and 3). A specific example of a method of selecting a point for storing an image of a surrounding object and a method of storing an image will be described later. The robot 27 reaches the point H after passing the point G. The robot 27 knows that the predicted error at this point H has increased to a predetermined reference value (called a reference error), and the robot 27
Of the points where the images of surrounding objects were previously stored, the point closest to the current point is selected (steps 4 and 5). As a result, the point E is selected.

The robot 27 moves in a direction in which the difference between the image of the surrounding objects stored at the selected point E and the current surrounding situation decreases, and reaches the point E where this difference becomes zero (step 6). ). A specific example of this operation will be described later. At this time, the robot 27 replaces its own position recognition with the position recognition previously stored at the same time as the image of the surrounding object at the point E (step 7). As a result, the prediction error of the robot 27 is reduced to the value previously recognized at the point E.

By repeating this operation (step 8), the robot 27 can keep the predicted error of position recognition due to dead reckoning below the reference error even when performing work involving long-distance movement. .

FIG. 4 is a diagram showing how the prediction error increases as the robot 27 moves. The horizontal axis represents the movement distance of the robot 27, and the vertical axis represents the prediction error. On the horizontal axis, symbols A to H indicate movement distances corresponding to points A to H in the environment where the robot 27 passes. The prediction error of the robot 27 is the starting point A.
Is zero at and increases in proportion to the distance traveled.
As described above, since the prediction error reaches the reference error at the point H, the robot 27 selects the point E from the points previously storing the images of the surrounding objects and reaches the point. At this time, the prediction error of the robot 27 is reduced to the value previously recognized at the point E.

Next, the robot 27 moves in a direction in which the method of selecting a point for storing the image of the surrounding object, the method of storing the image, and the difference between the stored image and the surrounding situation are reduced, and this difference is zero. We will give examples of how to reach each point.

First, a method of storing an image will be described. Figure 5
Indicates that the robot 27 is in the surrounding area at point E in FIG.
It is the figure which showed the example which memorize | stores the image of an object. At this point E
The robot 27 uses the objects 23, 2 existing in the surroundings.
The images of 4 and 25 are virtual circumferences (hereinafter,
Project on 28 (calling)
Use a Levi camera). For example, the image of the object 23 is a virtual
It becomes a projection 29 on the circumference 28. This projection 29 is a robot
Constant direction 30 indicated by the gyroscope of the toe 27
Angle θ with reference to (for example, north)₁ And α₁ By
Represented. Where α₁ Robot 27
Viewing angle, θ₁ Is the viewing angle α₁ The bisector 31 of is the direction
It is an angle formed with 30 (hereinafter, the former is the viewing angle, the latter is
Called direction angle). The robot 27
Position recognized by dead reckoning (in this embodiment
And the coordinate values (x, y) in the xy Cartesian coordinate system
, N, the number of objects existing around (at point E
Is n = 3), and the above θ ₁ , Α₁ Each object
The set of values obtained for is collectively stored (in the following,
These values are used as position information, number information and angle information, respectively.
It is called information and is collectively called point information. Especially for storage devices
The stored spot information is called stored spot information). Ground
The point information at the point E is {x, y}, 3, {θ₁ , Α
₁ }, {Θ₂ , Α₂ }, {Θ₃ , Α₃ } Becomes. However
The angle information representing the objects 24 and 25, respectively.
{Θ₂ , Α₂}, {Θ₃ , Α₃ } Was set. Robot 27
At multiple points in the environment (points A to G in the case of Fig. 3)
In addition, we collect and record the spot information at each spot.
Store information about storage points.

FIG. 6 shows that the robot 27 moves in a direction in which the difference between the previously stored image and the current surrounding situation decreases.
It is a figure which shows the example of the method of reaching the point where this difference becomes zero. Now, it is assumed that the robot 27 is located at the point H in FIG. As mentioned above, the robot 2
7 knows that the prediction error caused by dead reckoning at this point has reached the reference error, and selects the point closest to the current position among the points previously storing the images of the surrounding objects. This selection is performed by recognizing the position information included in all the stored position information at the position H {x ', y'}.
By comparing with. As a result, the point E is selected.

The robot 27 is similar to the case at the point E, and the surrounding objects 23, 24 and 25 at the point H.
The image of is projected onto the virtual circumference 32. In this case, for example, the image of the object 23 becomes a projection 33 on the virtual circumference 32.
This projection 33 is projected in the fixed direction 30 indicated by the gyroscope.
Is represented by angles θ ′ ₁ and α ′ ₁ . As a result, the point information at the point H is {x ', y'},
_{3, {θ '1, α} ' 1}, {θ '2, α' 2}, {θ '3, α' 3}
Becomes However, the angle information of the objects 24 and 25 at the point H is (θ ′ ₂ , α ′ ₂ ), (θ ′ ₃ , α ′ ₃ ), respectively.
And In FIG. 6, the virtual circumference 28 at the point E and the projections of the objects 23, 24, 25 are shown in an overlapping manner.

Next, the robot 27 associates the projections on the virtual circumference 28 with the projections on the virtual circumference 32 (step 11). This operation is performed by selecting, for each projection on the virtual circumference 28, the projection having the closest direction angle from the projections on the virtual circumference 32 among the angle information. As a result, for example, the projection 3 with respect to the projection 29
3 are combined. The robot 27 obtains two types of movement vectors (called radius vector and angle vector) for each combination of projections (step 12). Here, the movement vector is a vector indicating that if the robot moves in the direction indicated by the vector, the difference in angle information between the associated projections decreases. The magnitude of the movement vector represents the degree of difference between the pieces of angle information.

Of the two types of movement vectors, the diameter vector indicates the movement direction in which the difference in the viewing angle among the angle information is reduced. The radius vector is a vector oriented in the direction indicated by the bisector of the visual angle at which the robot 27 looks at the object. For example, the vector 35 directed in the direction indicated by the bisector 34 of the viewing angle when the object 23 is viewed is the diameter vector corresponding to the projection 33. The size of the diameter vector is obtained based on the relative relationship of the viewing angles of the combined projections. For example, if the viewing angle α ′ ₁ of the projection 33 is smaller than the viewing angle α _{1 of the} projection 29, this means that the distance to the object 23 at the point H is larger than the distance at the point E. Therefore, the robot 27 can reduce the difference between the viewing angles by moving in the same direction as the direction of the bisector 34. That is, the diameter vector 35 needs to have a positive size. Further, the absolute value is obtained, for example, by multiplying the difference α ′ ₁ −α ₁ between the two viewing angles by a certain proportional coefficient.

On the other hand, the angle vector indicates the moving direction in which the difference in the direction angle decreases. The angle vector is a vector that is orthogonal to the bisector of the visual angle at which the robot 27 looks at the object and is oriented in a direction in which the angle (hereinafter, referred to as an absolute angle) with respect to the constant direction 30 increases. For example, if the viewing angle of the object 23 is 2
A vector 36 orthogonal to the bisector 34 is an angle vector corresponding to the projection 33. The size of the angle vector is obtained based on the relative relationship between the direction angles of the combined projections. For example, if the directional angle θ ′ ₁ of the projection 33 is larger than the directional angle θ _{1 of the} projection 29, it means that the absolute angle of the object 23 at the point H is larger than the value at the point E. Therefore, the robot 27 divides the bisector 3
By moving in the direction in which the absolute angle of 4 decreases, it is possible to reduce the difference between the angles in both directions. That is, the angle vector 36 needs to have a positive size. Also, the absolute value is, for example, the difference θ ′ ₁ −θ ₁ between the two-direction angles,
Multiply by a certain proportional coefficient.

By the above operation, the radius vector and the angle vector are obtained for each combination of the projections on the virtual circumference 28 and the projections on the virtual circumference 32. The robot 27 obtains a vector (referred to as a combined movement vector) obtained by combining all of these vectors (step 1).
3). Next, the robot 27 moves in the direction indicated by the combined movement vector by a slight distance proportional to the magnitude of the combined movement vector (step 15). After this slight movement, the robot 27 repeats the above operation again to obtain a new combined movement vector (step 11).
~ 13). If the robot 27 reaches the point E after this series of movements, the projection on the virtual circumference at that point will all coincide with the projection on the virtual circumference 28. Therefore, the radius vector and angle vector obtained from the combination of the projections are all zero (step 1
4). The robot 27 recognizes that the point E has been reached because all of these vectors have become zero.

Finally, a method of selecting a point where the robot 27 stores the images of surrounding objects will be described based on the example of movement of the robot shown in FIG. The robot 27 first stores the image of the surrounding object at the point A and uses it as the storage point information. Next, move along the route shown in FIG. 3 and stop appropriately in the process (for example, stop again at a point where the moving distance from the point where the vehicle stopped last time exceeds a certain value).
Get point information. Each time the robot 27 obtains new spot information, it compares it with the stored spot information at the spot A. If the difference between the two exceeds a certain level (for example, if the above composite movement vector is obtained based on the difference between the two and the magnitude exceeds a certain value), the point information at that point is stored and stored. Use as point information. FIG. 3 shows that new storage location information has occurred at location B. Next, the robot 27 moves while comparing the stored spot information at the spot B with the spot information obtained on the moving route, and moves to the spot C.
In, get new memory point information. Hereinafter, the same operation is repeated.

[0029]

As described above, according to the present invention, the dead
In addition to reckoning, the robot recognizes its own position using the relative positional relationship between the surrounding objects and the robot, and stores the relative positional relationship when the predicted position recognition error increases to a reference value. Select the closest point to the current point among the existing points, move to the direction in which the error in the relative positional relationship at both points decreases, and when the relative positional relationship matches, change the position of that point to the current point. By recognizing the position of a point, even when performing a work involving a long-distance movement of a robot, it is possible to always correctly recognize its own position in the environment, and as a result, it is possible to perform the work.

[Brief description of drawings]

FIG. 1 is a flowchart showing a method of recognizing a self position of a mobile robot according to an embodiment of the present invention.

FIG. 2 is a detailed flowchart of step 6 in FIG.

FIG. 3 is a schematic diagram showing how a robot 27 moves in an environment.

FIG. 4 is a schematic diagram showing how the prediction error increases as the robot 27 moves.

5 is a schematic diagram showing an example in which the robot 27 stores images 23, 24, 25 of surrounding objects at a point E in FIG. 1. FIG.

FIG. 6 is a schematic diagram showing a method in which the robot 27 moves in a direction in which the difference between the previously stored memory location information and the current location information decreases, and reaches a point where the difference becomes zero.

[Explanation of symbols]

1-8, 11-15 Step 21 Boundary line that delimits environment 22-26 Object 27 Robot 28,32 Virtual circumference 29,33 Projection 30 Constant direction indicated by gyro compass 31,34 Bisecting line 35 Diameter vector 36 Angle vector A to H points α ₁ , α ′ ₁ viewing angle θ ₁ , θ ′ ₁ direction angle

Claims (2)

[Claims] 1. A method for a mobile robot to recognize its own position in an area in which the mobile robot moves, and moves while recognizing the rough position of itself using dead reckoning, and at the same time a position caused by dead reckoning. The step of predicting the recognition error, and when reaching a predetermined point during the movement, the image of the object existing around itself is projected on a virtual circle centered on that point and projected. Storing the information that characterizes the image of the object and at the same time storing the position recognized by dead reckoning, and when the prediction error of position recognition increases to a predetermined reference value, the image of the surrounding object is previously stored. Selecting a point closest to the current point among the points storing the amount to be characterized, the information and the current point at the selected point Repeating the operation of moving a slight distance in the direction in which the difference of the information decreases until the difference becomes zero, and when the difference becomes zero, the position of the point is stored in the selected point. A method for recognizing a self-position of a mobile robot, the method including the step of recognizing the position of the mobile robot. 2. The information is a direction angle, which is an angle formed by a viewing angle of the projection viewed from the center of a circle and a direction in which a bisector of the viewing angle is a reference, and at the selected point. For each projection on the virtual circumference, the projections on the virtual circumference at the current location that have the closest directional angles are associated, and the projections on the virtual circumference at the current location are associated. For each projection,
The direction is the direction indicated by the bisector of the viewing angle when the projection is viewed from the center of the circle, and the orientation is the viewing angle of the projection and the corresponding viewing on the virtual circumference at the selected point. The direction in which the difference in angle is reduced, the radius vector whose absolute value of the magnitude is proportional to the difference in both viewing angles, and the direction are orthogonal to the bisector of the viewing angle when the projection is viewed from the center of the circle, The orientation is an orientation in which the difference between the direction angle of the projection and the corresponding direction angle on the virtual circumference at the selected point decreases.
The angle vector whose absolute value is proportional to the difference between the angles in both directions is obtained, and the radius vector and angle vector obtained for all projections on the virtual circumference at the current point are combined, and the combined vector is The self-position recognition method for a mobile robot according to claim 1, wherein the mobile robot moves in the direction indicated by a distance proportional to the magnitude of the combined vector.