JP5222971B2 - Walking robot apparatus and control program therefor - Google Patents

Description

  The present invention relates to a walking robot apparatus and a control program therefor.

  A walking robot that moves autonomously moves while avoiding obstacles and steps autonomously. Therefore, it is necessary for the walking robot to always calculate an accurate self-location using an environment map describing the moving environment. Since a conventional walking robot is required to move accurately in a moving environment, the environment map must be created with high accuracy.

JP 2007-322138 A

By the way, home robots intended for use in the home are not necessarily required to have high performance and high accuracy, but have the minimum necessary functions and satisfy the necessary and sufficient performance and accuracy. Sometimes it is good.
However, in the conventional environmental map creation and self-position estimation, it was necessary to install expensive sensors and measuring devices on the walking robot and perform heavy processing operations. There were limitations.

  The present invention has been proposed in view of such a conventional situation, and a walking robot apparatus capable of easily realizing environmental map creation and self-position estimation while ensuring the required accuracy, and its An object is to provide a control program.

In order to solve the above problems, the present invention generates an environment map based on an image captured using a single mounted imaging device, estimates the self-position using the generated environment map, and within a predetermined area. A biped walking robot device that creates an environment map is a representative point extraction that extracts a representative point indicating the object to be photographed for each object to be photographed from two map creation environment images. A position coordinate acquisition unit that obtains the position coordinates of the object to be photographed from the position of each representative point of the environment image for map creation photographed at two locations and the position coordinates of the photographed walking robot device, and the photographed An environment map creation unit that creates an environment map that associates the environment image for map creation, the position of the representative point, and the position coordinates of the object to be photographed, and the function of estimating the self-position is at any position coordinates A shooting object corresponding unit that associates a shooting object in the captured environment image for position estimation with a shooting object in the environment image for map creation included in the environment map, and a shooting object in the environment image for position estimation The position coordinate corresponding unit that associates the position coordinates of the object to be photographed in the corresponding mapping environment image with the representative point of the object, and the point representative of the object to be photographed in the environment image for position estimation corresponding to the position coordinate are used. have a self position calculating unit for calculating its own position of the walking robot by stomach resection, representative point indicating the imaging object, the imaging target obtained from a point where a plurality of features on the shooting target It is the point which shows a thing on behalf .

  According to the present invention, it is possible to provide a walking robot apparatus capable of realizing cost reduction, downsizing, and multi-functionality, and a control program thereof, while ensuring the required accuracy for creating an environmental map and self-position estimation. it can.

The figure which shows the external appearance structure of a robot apparatus. The figure which shows the structure of the control block which controls creation of the environmental map of a robot apparatus, and self-position estimation, and the flow of the signal between control blocks. The flowchart which shows the procedure which creates an environmental map. The figure for demonstrating the method of creating an environmental map. The schematic diagram which represented the feature point function by the three-dimensional orthogonal coordinate which makes a picked-up image an XY plane. The figure explaining a landmark candidate area | region. The figure explaining the processing method when a feature point overlaps. The figure which shows the method of calculating the gravity center position coordinate of a landmark. The flowchart which shows the procedure which estimates a self-position. The figure for demonstrating the method of estimating a self-position. The figure for demonstrating the method of calculating | requiring a present position.

[First embodiment]
The walking robot apparatus (hereinafter referred to as a robot apparatus) of the present embodiment mainly serves as a personal home concierge that provides information and services while allowing people to enjoy daily life. For this reason, the robot apparatus is required to freely move in the living space by “moving intelligence” as well as natural communication by “communication intelligence”. Therefore, it must be able to move autonomously to the destination in an environment where there are a plurality of furniture in the living space.

FIG. 1 is a diagram illustrating an external configuration of the robot apparatus 1.
This robot apparatus 1 is configured as a humanoid type having functions of vision, hearing, conversation and bipedal walking. In order to realize such a function at a low cost, a multiplex system is adopted that realizes a high-performance technology by combining a plurality of general-purpose technologies. For example, a design concept is adopted in which general-purpose parts are used for hardware such as a mother board and an actuator, and functions are complemented by software.

  The robot apparatus 1 is equipped with a camera 10 and images the environment around the robot apparatus 1. The robot apparatus 1 executes creation of an environment map and self-position estimation based on surrounding environment images captured by the camera 10. Furthermore, as shown in FIG. 1, the robot apparatus 1 is equipped with a plurality of pressure sensors, gyros, acceleration sensors, and the like, and realizes a stable bipedal walking operation by controlling the movement of each joint axis. Description is omitted. In the following description, it is noted that the robot device 1 can take an arbitrary posture based on a command from a control device (not shown) and can perform a stable biped walking operation in an arbitrary direction. I want.

Next, terms used in the following description are defined.
* Landmark: An area on the captured image that is treated as a target on the environmental map.
* Object to be photographed: An object in the photographed image.
* Landmark object: A shooting object corresponding to a landmark.
* Shooting information: Information including the shot image, shooting angle, and the position of the robot that shot the shot.
* Feature information: Information including the position of a characteristic point (feature point) on the photographed image, the feature amount of the feature point, the mathematical parameter (feature point amount) of the feature point, and the like.
* Landmark information: Information including shooting information for each landmark, feature information, coordinates representing the area, and barycentric coordinates of the area.

  FIG. 2 is a diagram illustrating a configuration of a control block that controls creation of an environment map and self-position estimation of the robot apparatus 1 and a signal flow between the control blocks.

  The robot apparatus 1 includes, as control blocks, an environment image capturing unit 21, an environment image connecting unit 22, a feature information generating unit 23, a walking processing unit 24, a current position coordinate calculating unit 25, a shooting information generating unit 26, and a landmark candidate generating unit. 27, a corresponding point extraction unit 30, a landmark generation unit 31, and a landmark object gravity center position coordinate calculation unit 32 are provided.

  The robot apparatus 1 is provided with a storage device 40. The storage device 40 stores an environmental map. As will be described in detail later, the environment map describes position information on the presence of landmarks such as furniture in the living space, feature information of the landmarks, and the like.

  In FIG. 2, a block for realizing the environment map creation function is disclosed. A frame represented as a position A represents a signal flow between control blocks at a position before movement, and a frame represented as a position B. The inside shows the signal flow between the control blocks at the position after movement. In addition, the control block provided with the same function attaches | subjects the same referential mark.

In the frame represented as the position A, the environment image capturing unit 21 acquires a surrounding environment image captured using the camera 10. The environment image connection unit 22 performs connection processing of a plurality of environment images to form one image. The feature information generation unit 23 extracts feature information from the connected images. The walking processing unit 24 integrates the stride, the number of steps, and the turning angle. The current position coordinate calculation unit 25 calculates a coordinate value representing the current position from the accumulated stride, the number of steps, and the turning angle.
The shooting information generation unit 26 generates shooting information that associates a shot image, a shooting position, a shooting direction, and the like. The landmark candidate generation unit 27 extracts landmark candidates from the shooting information and the feature information, and outputs the landmark candidate information.

  In the frame represented as the position B, the surrounding environment is photographed in the same manner as described above, and photographing information and feature information are generated. However, landmark candidate information is not generated at position B. The landmark candidate information is generated in a state before movement.

  The corresponding point extraction unit 30 compares information acquired from coordinates captured at different positions before and after movement, and associates the information with each other. The landmark generation unit 31 extracts regions (landmarks) having similar characteristics from each other's information. The landmark object center-of-gravity position coordinate calculation unit 32 calculates the center-of-gravity position coordinates of the photographing object (landmark object) corresponding to the landmark. These landmark information and landmark object center-of-gravity position coordinates are stored in the storage unit 40 as an environmental map.

FIG. 3 is a flowchart showing a procedure for creating an environment map. FIG. 4 is a diagram for explaining a method of creating an environment map. The environment map creation operation will be described with reference to FIGS.
When the user instructs the robot apparatus 1 to create an environment map, the robot apparatus 1 starts an environment map creation operation. Here, the coordinates in the walking space are expressed in a world coordinate system with a predetermined position as a reference (origin). In this coordinate system, the position of the robot apparatus 1 is represented by (x, y, θ) using X coordinate, Y coordinate, and angle information indicating the orientation of the robot apparatus 1 as parameters. The robot apparatus 1 moves autonomously. The robot apparatus 1 calculates the coordinates (x, y, θ) of the current position by integrating the number of steps, the stride, and the turning angle. Accordingly, the following procedure will be described on the assumption that the robot apparatus 1 has grasped the coordinates (x, y, θ) of the current position at the time when the environment map creation is instructed.

  In step S01 in FIG. 3, the robot apparatus 1 captures a characteristic target object as a landmark candidate by photographing around the current position. The process of step S01 is performed as follows.

(100) The current position coordinate calculation unit 25 acquires the current position A (X1, Y1, θ1) of the robot apparatus 1.
(101) The environment image photographing unit 21 photographs the surroundings of the current position A with the camera 10 on the head of the robot apparatus 1. The number of shots is not particularly limited and is determined from the range in which the camera rotates, the angle of view, and the like. Although it is desirable to photograph the entire periphery, it is sufficient to photograph four images in total, two on the left and right sides of the traveling direction. In FIG. 4, three images, one on the left side (θα) and two on the right side (θβ, θγ), are taken using the moving direction as a reference line.
(102) The photographing information generation unit 26 stores the photographed image, the coordinates of the photographing position, and the photographing direction in the storage unit 40 in association with the photographing information.

  (103) The environment image coupling unit 22 stitches together images having continuous shooting ranges in the horizontal direction to form one image.

(104) The feature information generation unit 23 searches for feature points from the captured image. A feature point is a point that has features that are invariant to various uncertainties such as changes in the environment such as lighting and differences in the appearance of objects, and is suitable for extracting features of an object. is there. Note that what is generally known as a feature quantity is a feature quantity based on a shape or texture (texture of the surface of an object), or a feature point extracted from an image and a feature vector at that point as a feature quantity. There are extracted SIFT feature values and SURF feature values.
The feature information generation unit 23 calculates a feature amount and a feature point amount for each searched feature point, and stores them in the storage unit 40 as feature information. In the feature quantity extraction by SURF, a feature vector having 64-dimensional or 128-dimensional parameters is obtained as a feature quantity. Further, as feature point quantities, the coordinates (X, Y), Laplacian (sign of Laplacian at that point), Size (feature size), Dir (feature direction), Hessian (value of Hessian: The strength of the feature can be estimated.)

  (105) The landmark candidate generation unit 27 extracts an area of the target object that is a landmark candidate based on the shooting information and feature information stored in the storage unit 40. The process of extracting the landmark candidate area is performed as follows.

(105-1) A feature point function Z (i) expressed by the equation (1) is defined for each feature point i.

Note that the coordinates of the feature point i in the coordinate system set in the photographed image are (X _i , Y _i , 0), the feature size at the feature point i is S (i), and the feature Hessian at the feature point i is H (i).

  FIG. 5 is a schematic diagram showing the feature point function in three-dimensional orthogonal coordinates with the captured image as the XY plane. The feature point function is a three-dimensional shape projected on a circle with a diameter having a feature point as the center on the XY plane. The value of the feature point function becomes the maximum value (= Hessian value) at the feature point, decreases according to the cosine function as the distance from the feature point, and becomes the minimum value (= 0) at the boundary point of the circular region. Note that the curve that decreases from the feature point toward the boundary point of the circular region is not limited to the cosine function. Any function that decreases monotonously is acceptable.

(105-2) The values of all feature point functions are added, and an area where the added value is equal to or greater than a predetermined threshold (T) is projected onto the XY plane in the captured image. This area is referred to as a “landmark candidate area”.
FIG. 6 is a diagram illustrating landmark candidate areas. FIG. 6A is a diagram illustrating a range in which the values of feature point functions on one straight line on the XY plane are added and become equal to or greater than the threshold value. In FIG. 6A, the range from L1 to L2 is obtained. In addition, feature points belonging to this range are also obtained. FIG. 6B is a diagram showing, on the XY plane, a range in which a value obtained by adding feature point function values is equal to or greater than a threshold value.

  (105-3) A rectangle F circumscribing the landmark candidate region and parallel to the X axis and the Y axis is obtained. An area surrounded by the rectangle F is referred to as a “landmark candidate”. In addition, a photographing object corresponding to the area surrounded by the rectangle F is referred to as a “landmark candidate object”. Next, the barycentric coordinates of the rectangle F, that is, the intersection coordinates of the diagonal lines are obtained.

  (105-4) The feature amount of at least one feature point included in the region surrounded by the rectangle and the feature point amount, the vertex coordinates of the rectangle, the barycentric coordinates, and the shooting information are associated with each other in the storage unit 40 as landmark candidate information. Store.

  In step S02 of FIG. 3, the robot apparatus 1 moves to a different position (hereinafter referred to as “position B”) in the environment map creation area. The current position coordinate calculation unit 25 calculates and acquires the coordinates (x2, y2, θ2) of the current position B by integrating (odometry) the number of steps, the stride, and the turning angle.

In step S03, the periphery of the current position B is photographed by the camera 10 on the head of the robot apparatus 1. The number of shots is not particularly limited and is determined from the range in which the camera rotates, the angle of view, and the like. For example, two images are taken on the left and right sides around the traveling direction, for a total of four images. In FIG. 4, three images are taken, one on the left side (θδ) and two on the right side (θε, θζ) with respect to the moving direction.
Then, the robot apparatus 1 extracts feature amounts and feature point amounts from the captured image. Since this process is the same as the process (104) in step 01, detailed description thereof is omitted.

  In step S04, the corresponding point extraction unit 30 associates the same object with the captured images before and after the movement. The process of step S04 is performed as follows. In the following description, an image captured by the robot apparatus 1 at the position A is referred to as an image A, and an image captured at the position B is referred to as an image B.

  (200) One feature point is arbitrarily selected from the area surrounded by the rectangle of the image A. Further, one feature point is arbitrarily selected from the image B. When selecting the feature points of the image B, those having the same feature point amount Laplacian of each feature point are selected. This is because features with different laplacians do not match.

(201) For the two selected feature points, Euclidean distance is calculated by Equation (2) based on all the parameters of the feature amount.

(202) When the Euclidean distance is equal to or smaller than the threshold value, the combination of the feature points and the calculated Euclidean distance are associated and stored in the storage unit 40.
(203) The processes (201) and (202) are executed for all feature points in the image B, and a combination of feature points that gives the minimum Euclidean distance is obtained.

  (204) The processes (201) to (203) are executed for all the feature points inside the rectangle of the image A, and a combination of feature points giving the minimum Euclidean distance is obtained. At this time, it is avoided that the feature points on the image B are associated with each other.

FIG. 7 is a diagram illustrating a processing method when feature points overlap.
It is assumed that the feature point a1 on the image A and the feature point b1 on the image B are associated with each other with Euclidean distance = 100. In the association process, when the feature point a2 on the image A and the feature point b1 on the image B are associated with each other at the Euclidean distance = 120, the feature point b1 is associated with the feature point b1 in an overlapping manner. At this time, the combination with the smaller Euclidean distance is adopted, and the combination with the larger Euclidean distance is rejected. As a result, the feature point a2 on the image A is associated with the feature point b2 on the image B, which is the combination with the second smallest Euclidean distance.

In step S05 of FIG. 3, the landmark generation unit 31 extracts a landmark. The process of step S05 is performed as follows.
(300) An XY orthogonal coordinate system is set on the image B, and coordinates of feature points associated with the image B are determined.

(301) Arbitrary four sets of feature points associated with each other on image A and image B are selected, and a temporary projective transformation matrix for matching image A and image B is obtained using the four sets.
(302) The coordinates of the feature points on the image A are converted into the coordinates on the image B using the temporary projective transformation matrix, and the distance from the coordinates of the corresponding feature points is obtained. Then, the number of pairs whose distance is equal to or smaller than a predetermined value is used as the score of the temporary projective transformation matrix.

(303) For all combinations for selecting four pairs of feature points associated with each other on the image A and the image B, the processing (301) to (302) is executed to obtain the score of the temporary projective transformation matrix.
(304) The provisional transformation matrix having the highest score is set as a true transformation matrix for matching the coordinate system of the image A and the coordinate system of the image B.

(305) Project the rectangle of image A onto image B using the true transformation matrix. On the image B, the rectangle is projected as a quadrilateral.
A region surrounded by the currently selected rectangle on the image A and a region surrounded by the corresponding quadrilateral on the image B are referred to as landmarks.
The object to be imaged corresponding to the area surrounded by the rectangle and the object to be imaged corresponding to the area surrounded by the quadrilateral are the same object to be imaged from positions A and B, respectively. Is called a landmark object.

  (306) Feature amounts and feature point amounts of feature points included in the region enclosed by the currently selected rectangle, vertex coordinates and barycentric position coordinates in the coordinate system of the rectangular image A, shooting information A, and the rectangle Storage unit as landmark information by associating feature amounts and feature point amounts of feature points included in a region surrounded by the corresponding quadrilateral, vertex coordinates and barycentric coordinates in the coordinate system of the image B of the quadrilateral, and shooting information B 40.

In step S06 in FIG. 3, the landmark object gravity center position coordinate calculation unit 32 calculates the gravity center position coordinates of the landmark. The process of step S06 is performed as follows.
FIG. 8 is a diagram illustrating a method of calculating the center-of-gravity position coordinates of the landmark.

  (400) Based on the shooting information A and the shooting information B, the length of a line segment connecting the position A and the position B (hereinafter referred to as “line segment AB”) is calculated.

  (401) When the currently selected landmark object is LMα and the center of gravity position of LMα set by photographing from the position A is position α ′, the position A and the position α are based on the landmark information. An angle θα formed by a line segment connecting '(hereinafter referred to as “straight line Aα”) and the line segment AB is calculated.

  (402) Assuming that the position of the center of gravity of LMα set by photographing from the position B is the position α ″, based on the landmark information, a line segment connecting the position B and the position α ″ (hereinafter referred to as “straight line Bα”) And the angle θδ formed by the line segment AB is calculated.

  (403) Based on the triangulation method, the position coordinates of the intersection of the straight line Aα ′ and the straight line Bα are calculated as temporary coordinates of the center of gravity position of the currently selected landmark object.

  (404) Temporary coordinates of the center of gravity position of the currently selected landmark object are also calculated in environment map creation at positions other than position A and position B. Is calculated as a new temporary coordinate of the center of gravity position of the currently selected landmark object. Whether or not the landmark objects are the same is determined by whether or not the difference (distance) in the center of gravity of the landmark objects is within a predetermined value.

  In step S07 of FIG. 3, the landmark object gravity center position coordinate calculation unit 32 stores the calculated temporary coordinates of the gravity center position of the landmark object in the storage unit 40.

  In step S08, when a further movement destination (robot position) is defined (No in step S08), in step S09, the processing from step S01 is repeatedly executed in the movement destination. Then, after the processing is executed at the prescribed position specified, the final provisional coordinates of the center of gravity position of the landmark object are used as the true coordinates of the center of gravity position of the landmark object. Save it in the environmental map in association with the mark information.

  In the above-described embodiment, the environment map is created with two robot positions. However, the environment map may be created with three robot positions by moving to an L shape (or inverted L shape). The environment map may be created with four robot positions by moving to a U-shaped (or inverted U-shaped). Furthermore, an environment map may be created with a plurality of robot positions combining any number of these operations. The movement path may be set according to the environment to which the robot apparatus 1 is applied. In step S09, the robot apparatus 1 may change the direction at the position B and change the position B to the new position A to create a map of the entire surroundings without moving, in step S09. As a result, the same information as that moved to the L shape or the U shape can be obtained. The movement in step S09 is described as a concept including such an operation.

In the above-described embodiment, a rectangle is formed at the robot position A, and a quadrilateral is formed at the robot position B. That is, the landmark is extracted by comparing the image captured at the robot position A with the image captured at the robot position B as a reference. For this reason, the landmark of the object that is not photographed at the robot position A and is photographed for the first time at the robot position B is not extracted.
Therefore, for example, when moving from robot position A → B → C → D, after extracting a landmark by comparison with robot position B using robot position A as a reference, robot position C May be extracted, and the landmark may be extracted by comparison with the robot position D using the robot position C as a reference.

Next, a method for estimating the self-location using the environment map created as described above will be described.
FIG. 9 is a flowchart showing a procedure for estimating the self-position. FIG. 10 is a diagram for explaining a method of estimating the self position. The operation of estimating the self position will be described with reference to FIGS.
The robot apparatus 1 estimates a position (self position) where the robot apparatus 1 is present when it is installed at an arbitrary position in the movement space and an operation start is instructed by the user. Furthermore, the robot apparatus 1 estimates its own position when it moves a predetermined distance. Therefore, when the robot estimates its own position, there are two cases: the case where the self position can be grasped by odometry and the case where the self position is completely unknown. The self-position estimation methods shown in FIGS. 9 and 10 disclose self-position estimation methods in these two cases.

  In step T01, the robot apparatus 1 sets a position recognition flag and starts a self-position estimation operation. The position recognition flag is turned on when the self position can be grasped by odometry, and the position recognition flag is turned off when the self position is completely unknown.

  In step T02, the robot apparatus 1 captures the surroundings and extracts the same landmarks as those on the environment map. The process in step ST02 is performed as follows.

  (500) The environment image photographing unit 21 photographs the surroundings of the current position (position C) with the camera 10 on the head of the robot apparatus 1. There is no limit on the number of shots. A plurality of photographed images are connected in the horizontal direction so as not to overlap the photographing ranges, and one image (image C) is generated.

  (501) A feature point is searched from the image C, and a feature amount and a feature point amount (feature information C) are calculated for each feature point. Since this process is the same as the process (104) of step 01, detailed description thereof is omitted.

  (502) With reference to the feature information C, the corresponding point extraction unit 30 associates the feature points (corresponding points) corresponding to the feature points on the landmark stored in the environment map among the calculated feature points. Since this process is the same as the processes (200) to (204) in step S04 in FIG. 3, detailed description thereof is omitted.

  (503) Note that if there is no corresponding point, the current position is inappropriate, so the position is changed. For example, when the feature point exists at the end of the image, the direction is changed to the direction of the end. If no feature point exists, the direction is changed backward. Then, processes (500) to (502) are executed.

(504) Based on the extracted corresponding points, an area having a feature similar to the landmark stored in the environment map is extracted from the image C. Since this process is the same as the processes (300) to (306) in step S05, a detailed description thereof will be omitted.
The extracted area is a shape obtained by projective transformation of the landmark stored in the environment map to the coordinate system of the image C. That is, the shooting target corresponding to the extracted area is the same as the landmark target set when creating the environment map, and the shooting target corresponding to the extracted area was shot from the position C. It is a landmark object. Actually, the position of the photographing object corresponding to the extracted area may be slightly different from the set position of the landmark object due to errors such as corresponding point extraction or projective transformation.

  In step T03 of FIG. 9, the position coordinates of the center of gravity of the object to be imaged corresponding to the area are obtained from the environment map. In this way, the coordinates of landmarks existing in the image C are obtained from the environment map.

In step T04 of FIG. 9, the current position C of the robot apparatus 1 is obtained by application of a meeting method often used in flat plate surveying. The process of step T04 is performed as follows.
FIG. 11 is a diagram for explaining a method for obtaining the current position.

  (600) Select two arbitrarily selected landmarks. As shown in FIG. 11, the landmark objects corresponding to the two selected landmarks are denoted by LMδ and LMε, and the center positions thereof are denoted by position δ and position ε, respectively. Based on the landmark information stored in the environmental map, ∠δCε (= 保存 θα) is calculated.

  (601) A center coordinate Oα and a radius R1 of a circle passing through the position δ, the position ε, and the position C are obtained. Since ∠θα is the circumference angle of the circle, the center angle is obtained as 2 × ∠θα. Therefore, the center coordinates of the circle can be obtained as vertex coordinates of an isosceles triangle having an apex angle of 2 × ∠θα that is equidistant (= R1) from the position δ and the position ε.

  (602) The center coordinates and radius of the circle are obtained for all combinations (nC2 ways) of selecting two landmarks among the extracted n landmarks.

  (603) If two circles are selected, the position C can be expressed by the intersection coordinates. Therefore, the intersection coordinates of the position C are obtained for all combinations for selecting two circles. These arithmetic averages are calculated as the coordinates of the current position C.

In step T05 of FIG. 9, it is determined whether or not the position recognition flag is ON.
When the position recognition flag is OFF (No at Step T05), that is, when the self position is completely unknown, the current position C obtained by applying the meeting method is set as the self position of the robot apparatus 1 at Step T06.
Then, the self-position estimation process ends.

When the position recognition flag is ON (No at Step T05), that is, when the self position can be grasped by odometry, the current position by odometry is calculated at Step T07.
In step T08, a new position coordinate is calculated by fusing the position coordinate calculated by odometry and the position coordinate calculated by the intersection method, and this coordinate is set as the true coordinate of the current position C.

  Here, the fusion is a method of calculating a value closer to the truth by processing both the values of the odometry and the association method so that the error due to the odometry and the error due to the association method cancel each other. The fusion can be performed with reference to, for example, POEM (Position Estimation module) or “Journal of the Robotics Society of Japan Vol.15 No.8, PP.1180-1187,1997”.

  As described above, according to the embodiment described above, various effects can be obtained.

The robot apparatus according to the present embodiment is not equipped with a distance measuring means such as a laser range finder that is normally equipped in a conventional robot. There is also one camera that captures the surrounding environment. Therefore, the cost of the robot apparatus can be reduced.
Also, in such an inexpensive hardware configuration, in order to ensure the required accuracy, the environmental map is composed of a set of landmark information such as feature quantities representing landmarks and barycentric position coordinates representing landmarks. Yes. As a result, the processing time and processing data for image processing and the like are reduced, and it is possible to eliminate arithmetic processing with a large load.

  In addition, the position of the landmark is defined by the coordinate position of one point as the center of gravity of the quadrilateral obtained from the feature amount distribution of the landmark. As a result, the self-position estimation process is simplified and can be processed efficiently.

  Thus, the robot apparatus according to the present embodiment is configured as a walking robot apparatus that can be realized at low cost while ensuring the accuracy required for home use.

  Note that the functions described in the above-described embodiments are not limited to being configured using hardware, and can be realized by causing a computer to read a program describing each function using software. Each function may be configured by appropriately selecting either software or hardware.

Note that the present invention is not limited to the above-described embodiment as it is, and can be embodied by modifying the constituent elements without departing from the scope of the invention in the implementation stage.
Various inventions can be formed by appropriately combining a plurality of constituent elements disclosed in the embodiment. For example, some components may be deleted from all the components shown in the embodiment. Furthermore, you may combine suitably the component covering different embodiment.

  DESCRIPTION OF SYMBOLS 1 ... Robot apparatus, 10 ... Camera, 21 ... Environmental image imaging | photography part, 22 ... Environmental image connection part, 23 ... Feature information generation part, 24 ... Walk processing part, 25 ... Current position coordinate calculation part, 26 ... Shooting information generation part 27 ... Landmark candidate generation unit, 30 ... Corresponding point extraction unit, 31 ... Landmark generation unit, 32 ... Landmark object center of gravity position coordinate calculation unit, 40 ... Storage unit.

Claims (5)

A walking robot device that generates an environment map based on an image captured using a single image pickup device, estimates a self-position using the generated environment map, and walks two legs in a predetermined area. ,
The ability to create an environmental map
A representative point extracting unit that extracts a representative point indicating the photographing object for each photographing object from the map creation environment images photographed at two locations;
A position coordinate acquisition unit for obtaining the position coordinates of the object to be photographed from the positions of the representative points of the map creation environment images photographed at two locations and the position coordinates of the photographed walking robot device;
An environment map creation unit that creates an environment map that associates the captured map creation environment image, the position of a representative point, and the position coordinates of the object to be photographed;
The function to estimate the self-position is
An imaging object corresponding unit that associates an imaging object in the position estimation environment image captured at an arbitrary position coordinate with an imaging object in the map creation environment image included in the environment map;
A position coordinate corresponding unit that associates the position coordinates of the imaging object in the corresponding mapping image with the representative point of the imaging object in the environment image for position estimation;
Position coordinates have a self position calculating unit for calculating its own position of the walking robot by resection method using representative points of the object to be shot in the location estimation environment image corresponding,
The walking robot device according to claim 1, wherein the representative point indicating the photographing target is a point representative of the photographing target obtained from a plurality of characteristic points on the photographing target . The function to estimate the self-position is
In the movement of the walking robot device, a current position coordinate calculation unit for calculating the self position by integrating the number of steps, the step length, and the turning angle;
The apparatus according to claim 1, further comprising means for calculating a new self-position by fusing the self-position calculated by the self-position calculation unit and the self-position calculated by the current position coordinate calculation unit . Walking robot device.   The apparatus further comprises a connecting unit that connects a plurality of images captured by rotating a head provided with the imaging device to generate the environment image for mapping and the environment image for position estimation. The walking robot device described in 1.   The function of creating an environmental map further includes a photographing unit for photographing an environmental image for map creation every time the vehicle stops and moves repeatedly to create an environmental map. Item 4. The walking robot device according to item 3. An environment map is generated based on an image photographed using one mounted image pickup device, the self-position is estimated using the generated environment map, and a walking robot apparatus that walks two legs in a predetermined area is controlled. A program,
When creating an environmental map,
Extracting representative points indicating the object to be photographed for each object to be photographed from the map creation environment images photographed at two locations;
Obtaining the position coordinates of the object to be photographed from the positions of the representative points of the map creation environment images photographed at two locations and the position coordinates of the photographed walking robot device;
Creating an environment map in which the captured map creation environment image, the position of the representative point, and the position coordinates of the object to be photographed are associated with each other.
When estimating your position,
Associating a photographing object in a position estimation environment image photographed at an arbitrary position coordinate with a photographing object in a map creation environment image included in the environment map;
Associating the position coordinates of the imaging object in the corresponding mapping environment image with the representative point of the imaging object in the position estimation environment image;
A step of calculating a self-position of the walking robot device by a meeting method using a representative point of the photographing object in the position estimation environment image corresponding to the position coordinates , and causing the control computer mounted on the walking robot to execute
The representative point indicating the object to be photographed is a program that represents the object to be photographed representatively obtained from a plurality of characteristic points on the object to be photographed .

Images