JP2018092682A - Autonomous mobile apparatus, autonomous mobile method, and program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an autonomous mobile apparatus capable of swiftly recovering from a state that own position cannot be estimated.SOLUTION: An autonomous mobile apparatus 100 has a movement processing part 11 that executes a series of processing to move autonomously. The position estimation section 13 estimates own position. An estimation determination section 14 determines whether the estimation result of own position by the position estimation section 13 is stable at a point. When the estimation determination section 14 determines that the estimation result is not stable at one point, a movement processing control unit 15 controls the movement processing part 11 to move to a position whether the estimation result of the own position made by the position estimation section 13 gets stable.SELECTED DRAWING: Figure 2

Description

  The present invention relates to an autonomous mobile device, an autonomous mobile method, and a program.

  Autonomous mobile devices that move autonomously according to their use have become widespread. For example, an autonomous mobile device that moves autonomously for indoor cleaning is known. In general, such an autonomous mobile device needs to create a map of the real space and estimate its own position in the real space.

As a technique for creating a map of the real space, for example, a SLAM (Multiple Localization And Mapping) method is known. The basic principle of SLAM technology using a monocular camera is described in Non-Patent Document 1, and by tracking the same feature points from a plurality of frames of moving images taken by the camera, the three-dimensional A process of alternately estimating a position (camera position) and a three-dimensional position of feature points (which are collected to form map information) is performed.
In addition, there is an autonomous mobile device that creates an environmental map in real space and performs autonomous movement (see, for example, Patent Documents 1 and 2).

Andrew J. Davison, “Real-Time Simulaneous Localization and Mapping with a Single Camera”, Proceedings of the 9th IEEE International Conference on Computer, Computer Viper. 1403-1410 Richard Hartley, Andrew Zisserman, “Multiple View Geometry in Computer Vision”, Second Edition, Cambridge. University Press, March 2004, chapter 9

Japanese Patent Laying-Open No. 2015-146091 JP 2014-186694 A

In the self-position estimation using such an image, if the shooting direction of the camera changes significantly, or if there are few feature points in the captured image, common feature points with the past images cannot be obtained, and There was a problem that the estimation of the aircraft position failed.
Further, in Patent Document 1, a candidate point of the self-position is estimated based on the amount of movement, a plurality of measurement points are defined on the map, the distance from the measurement point to the wall is set as the evaluation distance based on the distance map, In one evaluation step, the degree of coincidence is evaluated based on the evaluation distance, and in the second evaluation step, a measurement direction vector that is a unit vector in a direction facing the measurement point from the external sensor and a wall direction vector of the measurement point Disclosed is a mobile robot self-position estimation method that evaluates an inner product as an evaluation function, and estimates a candidate point having a large match in the first evaluation step and a large evaluation function in the second evaluation step as a self-position. Yes.
Further, in Patent Document 2, when activated, it is determined whether or not autonomous movement can be continued from the current own position based on the stop factor stored in the stop factor storage means, When it is determined that autonomous movement is possible, it starts autonomous movement based on the self-position stored in the self-position storage means, and when it is determined that autonomous movement is impossible, the stop state is set. An autonomous mobile automatic guided vehicle characterized by holding is disclosed.
However, although the technique described in Patent Document 1 can increase the accuracy of self-position estimation, there is a problem that it cannot be restored if the self-position estimation fails.
Further, the technique described in Patent Document 2 has a problem that it cannot be restored when it is determined that autonomous movement is impossible.

  Accordingly, the present invention has been made to solve the above-described problem, and an object thereof is to provide an autonomous mobile device and the like that can quickly return from a state where it is impossible to estimate its own position.

In order to achieve the above object, the autonomous mobile device of the present invention provides:
A movement processing unit for processing to move autonomously based on information of a plurality of images taken by the imaging unit;
A position estimation unit that estimates the position of the device using information of a plurality of images captured by the imaging unit;
An estimation determination unit that determines whether or not the estimation result of the own device position in the position estimation unit is determined as one;
When the estimation determination unit determines that the estimation result is not fixed to one, the movement processing unit is controlled so that the position estimation unit moves to a position where the estimation result of the own position is determined to be one. A movement processing control unit,
A position change history registration unit that registers the amount of change in the position change history storage unit each time the position of the own machine is changed by the movement processing unit;
When the estimation determination unit determines that the estimation result is determined as one, a position history registration unit that registers position information at that time together with time information in a position history storage unit;
With
In the case where the estimation determination unit determines that the estimation result is not fixed to one, the movement processing control unit goes back to the time when the device is at the position registered in the position history storage unit,
The movement processing unit is controlled to move in a direction opposite to a position change amount registered in the position change history storage unit.

  According to the present invention, it is possible to quickly return from a state where it is impossible to estimate its own position.

It is a figure which shows the external appearance of the autonomous mobile device which concerns on embodiment. It is a figure which shows the structure of the autonomous mobile apparatus which concerns on embodiment. It is a figure which shows the flowchart of the whole autonomous movement control process which concerns on embodiment. It is a figure which shows the flowchart of the initialization process of the own position estimation thread | sled in the autonomous movement control process which concerns on embodiment. It is a figure which shows the flowchart of the main process after the initialization of the own position estimation thread | sled in the autonomous movement control process which concerns on embodiment. It is a figure which shows the flowchart of the process of the map creation thread in the autonomous movement control process which concerns on embodiment. It is a figure which shows the flowchart of the process of the loop closing thread | sled in the autonomous movement control process which concerns on embodiment. It is a figure which shows the flowchart of the process of the movement process control in the autonomous movement control process which concerns on 1st embodiment. It is a figure which shows the flowchart of the process of the movement process control in the autonomous movement control process which concerns on 2nd embodiment. It is a figure which shows the structure of the autonomous mobile device which concerns on 3rd embodiment. It is a figure which shows the flowchart of the process of the movement process control in the autonomous movement control process which concerns on 3rd embodiment.

  Hereinafter, an autonomous mobile device according to an embodiment of the present invention will be described with reference to FIG. The autonomous mobile device 100 moves autonomously according to the application. Examples of this use include security monitoring, indoor cleaning, pets, and toys.

  The autonomous mobile device 100 includes an imaging unit 41 and a drive unit 42 in appearance.

  The imaging unit 41 includes a monocular imaging device (camera). The imaging unit 41 acquires an image (frame) at 30 fps, for example. The autonomous mobile device 100 performs autonomous movement while recognizing its own position and surrounding environment in real time based on images sequentially acquired by the imaging unit 41.

The drive unit 42 is an independent two-wheel drive type, and is a moving means including wheels and a motor. The autonomous mobile device 100 translates back and forth (translation) by driving the two wheels in the same direction, and rotates (changes direction) on the spot by driving the two wheels in the reverse direction. A swivel movement (translation + rotation (direction change) movement) can be performed by the changed drive. Each wheel is also equipped with a rotary encoder, which measures the rotational speed of the wheel with the rotary encoder and uses the geometric relationship such as the wheel diameter and the distance between the wheels to translate and move. The amount can be calculated. For example, if the wheel diameter is D and the rotation speed is R (measured by a rotary encoder), the translational movement amount at the ground contact portion of the wheel is π · D · R. Further, if the wheel diameter is D, the distance between the wheels is I, the rotation speed of the right wheel is R _R , and the rotation speed of the left wheel is _RL , the rotation amount of the direction change is 360 (when the right rotation is positive). ° × D × (R _L −R _R ) / (2 × I). By sequentially adding the translational movement amount and the rotation amount, the drive unit 42 functions as a so-called odometry, and measures its own position (position and orientation with reference to the position and orientation at the start of movement). Can do.

  The accuracy of the position and orientation of the own aircraft obtained from odometry often becomes low due to wheel wear or slip. In particular, due to the accumulation of errors, the accuracy deteriorates with time. However, the accuracy of the rotational component (direction information) of odometry can be improved by using angular velocity sensor information described later. In addition, by using an azimuth sensor (not shown) that detects geomagnetism and identifies the azimuth, information on the absolute direction using geomagnetism can be obtained regardless of the obtained value from the odometry.

  In addition, you may make it provide a crawler instead of a wheel, and you may make it move by providing a plurality (for example, two) leg | foot and walking with a leg | foot. In these cases as well, the position and orientation of the subject aircraft can be measured based on the movements of the two crawlers and the movements of the feet as in the case of the wheels.

  As shown in FIG. 2, the autonomous mobile device 100 includes a control unit 10, a storage unit 20, a sensor unit 30, an input unit 43, a communication unit 44, and a power supply 45 in addition to the imaging unit 41 and the drive unit 42.

  The control unit 10 is configured by a CPU (Central Processing Unit) or the like, and by executing a program stored in the storage unit 20, each unit (movement processing unit 11, map creation unit 12, position estimation unit 13, estimation unit) described later is executed. The functions of the determination unit 14 and the movement processing control unit 15) are realized.

  The storage unit 20 includes a ROM (Read Only Memory), a RAM (Random Access Memory), and the like, and includes an image storage unit 21, a map storage unit 22, a position change history storage unit 23, and a position history storage unit 24. The ROM stores a program executed by the CPU of the control unit 10 (for example, a program related to an SLAM method calculation and an autonomous movement control process described later) and data necessary for executing the program in advance. The RAM stores data that is created or changed during program execution.

  The image storage unit 21 stores an image captured by the imaging unit 41. However, in order to save the storage capacity, it is not necessary to store all captured images, and it is also possible to store the feature amount of the image instead of the image itself. For important images (key frames to be described later), information on the position of the own device (the position and orientation of the own device) when the image is taken is stored together with information on the image.

  The map storage unit 22 stores a map (information of feature points and three-dimensional positions of obstacles) created by the map creation unit 12 based on information from the SLAM method and the obstacle sensor 33 described later.

  The position change history storage unit 23 stores a history of how the position of the own device has been changed by the movement processing unit 11.

  The position history storage unit 24 stores a history of the own device position estimated by the position estimation unit 13.

  As the sensor unit 30, an acceleration sensor 31, an angular velocity sensor 32, and an obstacle sensor 33 are provided. The acceleration sensor 31 is a sensor that measures acceleration in the XYZ directions (three axes). The speed in each direction of XYZ can be obtained by time integration of these values, and the change amount (movement distance) of the position in each direction of XYZ can be obtained by further time integration of these speeds (however, (Since the sensor value is integrated twice, the accumulated error is considerably larger).

  The angular velocity sensor 32 is a sensor that measures an angular velocity (an angular movement amount per unit time). It is known that the accuracy is improved by obtaining the direction using the angular velocity sensor 32 rather than obtaining the direction of the own machine based on the rotation speed of the wheel. The obstacle sensor 33 is a sensor that detects an obstacle while traveling, and is, for example, an infrared sensor or an ultrasonic sensor. In addition, you may make it detect an obstruction using the imaging part 41, without mounting the independent obstruction sensor 33. FIG. Moreover, you may provide the bumper sensor (not shown) which detects having collided with the other object. Moreover, you may provide the azimuth | direction sensor (not shown) which detects geomagnetism and pinpoints an azimuth | direction. By using the azimuth sensor, it is possible to acquire information on the absolute direction using geomagnetism regardless of the acquired value from the odometry.

  As the input unit 43, an operation button for operating the autonomous mobile device 100 is provided. The operation buttons include, for example, a power button, a mode switching button (switching between the cleaning mode, the pet mode, and the like), an initialization button (re-creating the map), and the like. As the input unit 43, a microphone (not shown) that inputs sound and a voice recognition unit that recognizes a voice of an operation instruction to the autonomous mobile device 100 may be provided.

  The communication unit 44 is a module for communicating with an external device, and is a wireless module including an antenna when performing wireless communication with the external device. For example, the communication unit 44 is a wireless module for performing short-range wireless communication based on Bluetooth (registered trademark). The communication unit 44 exchanges data between the autonomous mobile device 100 and the outside.

  The power source 45 is a power source for operating the autonomous mobile device 100, and is generally a built-in rechargeable battery, but may be a solar cell or a system that is wirelessly powered from the floor. Also good. When the power supply 45 is a rechargeable battery, the autonomous mobile device 100 is charged by being docked at a charging station (home base).

  Next, functions of the control unit 10 will be described. The control unit 10 includes a movement processing unit 11, a map creation unit 12, a position estimation unit 13, an estimation determination unit 14, and a movement processing control unit 15. . The control unit 10 supports a multi-thread function, and can proceed with a plurality of threads (different processing flows) in parallel.

  The movement processing unit 11 performs processing for moving the own unit by moving the driving unit 42.

  The map creating unit 12 uses the information of the image stored in the image storage unit 21 and the information on the position and orientation of the own device at the time of capturing the image, and the three-dimensional position of the feature point estimated using the SLAM method ( (Map point) and the three-dimensional position of the obstacle obtained based on the position and orientation information of the aircraft when the obstacle sensor 33 detects the obstacle in the map storage unit 22 as map information. Remember.

  The position estimation unit 13 estimates the position and orientation of the own device as visual odometry based on the SLAM method described later.

  The estimation determination unit 14 determines whether or not the position estimation unit 13 can estimate its own position based on currently acquired information. More specifically, if the information necessary for estimating the position of the own device has been acquired and the estimation result of the position of the own device based on the information is determined as one, it is determined that the information can be estimated. On the other hand, information necessary for estimating the position of the aircraft cannot be acquired, or even if it can be acquired, the estimation result of the position of the aircraft based on the information is indefinite (a plurality of estimation candidates are generated, etc.) ), It is determined that estimation is impossible.

  The movement processing control unit 15 controls the movement processing unit 11 to autonomously move to the position where the own device position can be estimated when the estimation determining unit 14 determines that the estimation is not possible.

  FIG. 3 is a main flow of the autonomous mobile device 100. The control part 10 starts each thread | sled of an own machine position estimation thread | sled (step S101), a map creation thread | sled (step S102), and a loop closing thread | sled (step S103). Generation of map information and visual odometry (information of the own position estimated using the map and the image) is started based on the SLAM method by the operation of the own position estimation thread and the map creation thread. Then, the control unit 10 clears the position change history counter DC (a variable indicating the number of position change histories registered in the position change history storage unit 23) to 0 (step S104). Thereafter, the control unit 10 determines whether or not the operation is finished (step S105). If the operation is finished (step S105; Yes), the operation is finished, and if the operation is not finished (step S105; No), the map creating unit 12 displays the map. The information is created / updated (step S106).

  Next, the control unit 10 determines whether the value of the estimated state variable ES (a variable indicating whether or not position estimation is possible) is “position estimation impossible” (step S107). If position estimation is impossible (step S107; Yes), after performing "movement process control" which is a process for autonomously moving to a position capable of position estimation (step S110), the process returns to step S105. This “movement processing control” realizes the characteristic function of this embodiment, and details will be described later.

  If the value of the estimated state variable ES is not “position estimation impossible” (step S107; No), the position change history counter DC is incremented (step S108), and the movement processing unit 11 instructs the drive unit 42 to move autonomously. In addition to instructing the operation, the position change amount accompanying the operation is used together with the time stamp (the current time or the like that can be compared with the time stamp of the position history storage unit 24) and the DCth of the position change history storage unit 23. The item is registered (step S109), and the process returns to step S105. In the position change amount registration process in step S109, the movement processing unit 11 corresponds to a position change history registration unit.

  Here, the amount of change in the position is from the position before the movement processing unit 11 gives an instruction to the driving unit 42 in step S109 to the position after the driving unit 42 operates according to the instruction and the autonomous mobile device 100 moves. This is the amount of change (for one movement operation in step S109). This amount of change may be acquired using odometry information, or may be acquired using the values of the acceleration sensor 31 and the angular velocity sensor 32. Normally, the odometry information can be used to obtain a more accurate position change amount, but the odometry information is not useful when it is moved by a human hand (the values of the acceleration sensor 31 and the angular velocity sensor 32 are still in this case). Useful). Therefore, it is desirable to use sensor information to compensate for errors in odometry information, such as when a wheel slips or is moved by a human hand, by using both odometry information and sensor information. .

  As a typical example, when the autonomous mobile device 100 is initially placed in the charging station and the power supply 45 is turned on, the autonomous mobile device 100 relies on the obstacle sensor 33 to move through the rooms in the house. The position of an obstacle such as a wall can be specified by and map information including the position of the obstacle can be created. When a map is created to some extent, it is possible to know an area that has no map information but is considered to be movable, and it is possible to encourage the creation of a wider map by moving autonomously to that area. It becomes like this. And if the information of the map of the almost movable whole area is created, the efficient movement operation | movement using the information of a map will be attained. For example, it is possible to return to the charging station through the shortest route from any position in the room or to efficiently clean the room.

  The self-device position estimation thread activated in step S101 of the main flow (FIG. 3) of the autonomous mobile device 100 will be described with reference to FIGS. In the own position estimation thread, the position estimation unit 13 first performs an initialization process based on FIG. 4, and then estimates the own position based on FIG. 5 (by visual odometry using the image acquired by the imaging unit 41). This is a process of continuing (estimating its own position).

  First, the initialization process of the own position estimation thread will be described with reference to FIG. The position estimation unit 13 sets “in initialization” to the estimated state variable ES, and the position history registration counter HC (a variable indicating the number of position histories registered in the position history storage unit 24) and the position history search counter SC (position The variable used when acquiring the position history from the history storage unit 24 is cleared (step S201). Next, −1 is set to the frame counter N (a variable indicating the number of images acquired by the imaging unit 41) (step S202). Next, the position estimation unit 13 determines whether the operation is finished (step S203). If the operation ends (step S203; Yes), the operation ends. If the operation does not end (step S203; No), the imaging unit 41 acquires an image (step S204). The image can be acquired at 30 fps, for example (the acquired image is also called a frame).

  Next, 2D feature points are acquired from the acquired image (step S205). A 2D feature point is a characteristic part in an image, such as an edge part in the image, and can be obtained by using an algorithm such as SIFT (Scale-Invariant Future Transform) or SURF (Speed-Up Robust Features). . Note that other algorithms may be used to acquire 2D feature points.

  If the number of acquired 2D feature points is small, calculation by the Two-view Structure from Motion method, which will be described later, cannot be performed. Therefore, in step S206, the position estimation unit 13 acquires the number of 2D feature points and a reference value (for example, 10). If it is less than the reference value (step S206; No), the process returns to step S203, and the acquisition of the image and the acquisition of the 2D feature point are repeated until the number of 2D feature points equal to or greater than the reference value is obtained. At this point, the map information has not yet been created. However, in the typical example described above, for example, since the obstacle sensor 33 has been relied on to start moving all rooms in the house, this initialization process is performed. If the image acquisition and the 2D feature point acquisition are repeated, the image acquisition is repeated while moving, so that various images can be acquired, and it can be expected that an image having a large number of 2D feature points will be acquired.

  If the number of 2D feature points acquired is equal to or greater than the reference value (step S206; Yes), the position estimation unit 13 increments the frame counter N (step S207). Then, it is determined whether or not the frame counter N is 0 (step S208). If the frame counter N is 0 (step S208; Yes), it means that only one image has been acquired yet, so the process returns to step S203 to acquire the second image. Although not described in the flowchart of FIG. 4, the position of the own machine when the first image is acquired and the position of the own machine when the second image are acquired are somewhat distant. The accuracy of the posture estimated in the subsequent processing is improved. Therefore, when returning from step S208 to step S203, a process of waiting until the translation distance by odometry becomes a predetermined distance (for example, 1 m) or more by an operation instruction in the movement processing unit 11 in step S109 of the main flow (FIG. 3). May be added.

  If the frame counter N is not 0 (step S208; No), it can be seen that two images have been acquired, so that the position estimation unit 13 matches the 2D feature points between these two images (the same point in the actual environment). Are present in each image and can be dealt with) (step S209). In step S209, the position estimation unit 13 corresponds to a corresponding feature point acquisition unit. If the correspondence number of the feature points acquired here is less than 5, it is impossible to estimate the posture between two images, which will be described later. Therefore, the estimation determination unit 14 determines whether the correspondence number of the feature points is less than 5 (step) S210). If it is less than 5 (step S210; Yes), the process returns to step S202 to reacquire the initial image. If the number of corresponding feature points is 5 or more (step S210; No), using the Two-view Structure from Motion method, the posture between the two images (the difference between the positions where each image was acquired (translation vector) t) and the difference in orientation (rotation matrix R)) can be estimated (step S211).

  Specifically, this estimation is obtained by obtaining a basic matrix E from the corresponding feature points and decomposing the basic matrix E into a translation vector t and a rotation matrix R. The details are described in Non-Patent Document 2. I will omit it here. Note that the value of each element of the translation vector t obtained here (having three elements of X, Y, and Z with the position where the first image is acquired as the origin, assuming that it moves in a three-dimensional space) is Since it is different from the value in the real environment (the Two-view Structure from Motion method cannot obtain the value in the real environment itself, but obtains the value in a space similar to the real environment). These are regarded as values on the SLAM space, and will be described below using coordinates (SLAM coordinates) on the SLAM space.

  Note that even if the number of corresponding feature points is 5 or more, there may be cases where the estimated value of the posture between the two images is not fixed to one (if the condition becomes indefinite and the least squares method is used). In some cases, there are cases where the estimated value is not limited to one, but becomes more than one. When the position estimation unit 13 estimates the posture between two images (the difference in position (translation vector t) and the difference in orientation (rotation matrix R) from which each image was acquired), and the estimated value is determined as one. The estimation determination unit 14 determines that estimation is possible, and if the estimation value is not fixed to one, the estimation determination unit 14 determines that estimation is impossible (step S212).

  If the estimation determination unit 14 determines that the estimation is impossible (step S212; No), the process returns to step S202 and starts again from the acquisition of the initial image. If the estimation determination unit 14 determines that the estimation is possible (step S212; Yes), it means that the posture (translation vector t and rotation matrix R) between the two images has been obtained, and the process proceeds to step S213.

  When the posture (translation vector t and rotation matrix R) between the two images is obtained, the values are based on the first image (the position where the first image was acquired is the origin of SLAM coordinates, the translation vector is the 0 vector, the rotation matrix) Is the unit matrix I.) The orientation of the second image (the position (translation vector t) and orientation (rotation matrix R) of the own machine when the second image is acquired)) . Here, when the postures of the two images (the position (translation vector t) and direction (rotation matrix R) of the own device at the time of photographing the image (frame) are also referred to as frame postures) are obtained. Based on the following concept, the map creation unit 12 obtains the 3D position in the SLAM coordinates of 2D feature points (corresponding feature points) that can be matched between two images (step S213).

If the coordinates (frame coordinates: known) of the 2D feature point in the image are (u, v) and the 3D position (unknown) in the SLAM coordinate of the 2D feature point is (X, Y, Z), these are the same. These relationships when expressed in the next coordinates are expressed by the following formula (1) using the perspective projection matrix P. Here, the symbol “˜” represents “equal except for a non-zero constant multiple” (that is, equal or is a constant (non-zero) multiple), and the “′” symbol represents “transpose”. .
(U v 1) ′ to P (XY Y 1) ′ (1)

In the above equation (1), P is a 3 × 4 matrix, and from the 3 × 3 matrix A indicating the internal parameters of the camera and the external parameters R and t indicating the posture (frame posture) of the image, the following equation It is represented by (2). Here, (R | t) represents a matrix in which the translation column vector t is arranged to the right of the rotation matrix R.
P = A (R | t) (2)

  In the above equation (2), R and t are obtained as the frame postures as described above. Further, since the internal parameter A of the camera is determined by the focal length and the image sensor size, it is a constant if the image capturing unit 41 is determined.

One of the 2D feature points corresponding to the two images is the frame coordinates (u ₁ , v ₁ ) of the first image and the frame coordinates (u ₂ , v ₂ ) of the second image. ), The following equations (3) and (4) can be obtained. Here, I is a unit matrix, 0 is a zero vector, and (L | r) is a matrix in which a column vector r is arranged to the right of the matrix L.
(U ₁ v ₁ 1) ′ to A (I | 0) (XY Z 1) ′ (3)
(U ₂ v ₂ 1) ′ to A (R | t) (XY Z 1) ′ (4)

In the above formulas (3) and (4), because there are formulas for u ₁ , v ₁ , u ₂ , and v _2, there are four formulas, but there are three unknowns, X, Y, and Z. , X, Y, Z can be obtained, and this is the 3D position of the 2D feature point in the SLAM coordinates. Since the number of equations is larger than the number of unknowns, for example, X, Y, Z obtained from u ₁ , v ₁ , u ₂ and X, Y, Z obtained from u ₁ , v ₁ , v ₂ are used. May be different. In such a case, a simultaneous linear equation with an excess condition is obtained, and there is generally no solution, but the map creating unit 12 obtains the most probable X, Y, and Z by using the least square method (step S213).

  When the 3D position (X, Y, Z) of the 2D feature point in the SLAM coordinates is obtained, it is used as a Map point, and the map creation unit 12 is also referred to as a Map point database (Map point DB (Database), which is stored in the map storage unit 22. Registered) (step S214). As elements to be registered in the Map point database, at least “X, Y, Z which is a 3D position of a 2D feature point in SLAM coordinates” and “a feature amount of the 2D feature point” (for example, a feature amount obtained by SIFT or the like) )is necessary. Further, if a “time stamp” (a value at the time of registration in the Map point database of a key frame counter NKF (a variable representing the current key frame number) described later) is added to the registered elements in the Map point database. This is convenient when editing the Map point database (such as returning to the past state).

  Then, the map creation unit 12 determines whether all 2D feature points (corresponding feature points) corresponding to the two images have been registered in the Map point database (step S215), and all the registrations are still possible. If not (Step S215; No), the process returns to Step S213, and if all of them are registered (Step S215; Yes), the process proceeds to Step S216.

  Next, the position estimation unit 13 initializes NKF (a variable representing a counter of a key frame (pointing to an image to be processed in a subsequent thread)) to 0 (step S216), and the second image is a key frame. Is registered in a frame database (also referred to as a frame DB (Database), which is stored in the image storage unit 21) (step S217).

  Elements to be registered in the frame database are “key frame number” (value of the key frame counter NKF at the time of registration), “posture” (position (translation vector t) in the SLAM coordinates of the own device at the time of image capture) Direction (rotation matrix R)), “all extracted 2D feature points”, “a point where 3D position is known as a Map point among all 2D feature points”, “feature of key frame itself”, In addition to these, the “posture in the actual environment measured by odometry” (the position and orientation of the own machine determined based on the movement distance by the drive unit 42 in the actual environment) may be registered.

  Among the above, “features of the key frame itself” is data for improving the processing for obtaining the image similarity between the key frames, and it is usually preferable to use a histogram of 2D feature points in the image. The image itself may be a “characteristic of the key frame itself”. In addition, the “posture in the real environment measured by odometry” can be expressed by a translation vector t and a rotation matrix R. However, since the autonomous mobile device 100 normally moves on a two-dimensional plane, two-dimensional data Alternatively, it may be expressed as two-dimensional coordinates (X, Y) and direction φ with reference to the position (origin) and direction at the start of movement.

  Next, in order to notify the map creation thread that the key frame has been generated, the position estimation unit 13 enters the key frame queue of the map creation thread (the queue has a first-in first-out data structure). The key frame counter NKF is set (step S218).

  Then, in order to obtain the scale correspondence between the SLAM coordinates and the real environment coordinates, the position estimation unit 13 translates the translation distance (obtained by the coordinates in the real environment) by odometry in the SLAM coordinates estimated by the above processing. A scale S is obtained by dividing by the distance d (step S219).

  Since the initialization process of the own machine position estimation thread is completed as described above, the process proceeds to own machine position estimation (main process) in FIG. 5 (step S220).

  Next, the main process of the own position estimation thread will be described with reference to FIG. First, the position estimation unit 13 sets “position estimation possible” to the estimated state variable ES (step S231). This is because in the initialization process, the position of the own aircraft has been estimated, and it can be said that the position can be estimated at this point. Next, the position estimation unit 13 determines whether the operation is finished (step S232). If the operation ends (step S232; Yes), the operation ends. If the operation does not end (step S232; No), an image is acquired by the imaging unit 41 (step S233), and the frame counter N is incremented (step S234).

  Next, the position estimation unit 13 acquires 2D feature points included in the captured image (step S235). Then, from the information of the previous key frame (for example, the image whose key frame number is NKF) registered in the frame database, the 3D position is known among the 2D feature points included in the image information ( A 2D feature point (which is a Map point registered in the Map point database) is acquired, and a 2D feature point (corresponding feature point) that can be correlated with the currently captured image is extracted (step S236).

  Then, the estimation determination unit 14 determines whether or not the number of corresponding feature points is less than a predetermined value (for example, 5) (step S237), and if it is less than the predetermined value (step S237; Yes), the own vehicle position is likely. Therefore, the process proceeds to step S247. Here, instead of proceeding immediately to step S247, the process returns to step S236 to search the key frame registered in the frame database for the number of corresponding feature points equal to or greater than a predetermined value. good. In this case, if no key feature registered in the frame database has a corresponding feature point greater than or equal to a predetermined value, the process proceeds to step S247.

When the corresponding feature point of a predetermined value or more can be extracted (step S237; No), the position estimation unit 13 acquires the 3D position (X _i , Y _i , Z _i ) of each corresponding feature point from the Map point database ( Step S238). The frame coordinates of the corresponding feature points included in the image that has just been taken are (u _i , v _i ), and the 3D positions of the corresponding feature points are (X _i , Y _i , Z _i ) (i is from 1) Values corresponding to the number of corresponding feature points) and values (ux _i , vx _i ) obtained by projecting the 3D positions (X _i , Y _i , Z _i ) of the corresponding feature points onto the frame coordinate system according to the following equation (5): And the frame coordinates (u _i , v _i ) should ideally match.
(Ux _i vx _i 1) ′ to A (R | t) (X _i Y _i Z _i 1) ′ (5)

Actually, since (X _i , Y _i , Z _i ) and (u _i , v _i ) also contain errors, (ux _i , vx _i ) and (u _i , v _i ) match. I rarely do it. And the unknowns are only R and t (three-dimensional in the three-dimensional space, and 3 + 3 = 6 is the number of unknowns), but there are twice the number of corresponding feature points (for one corresponding feature point) Since there are equations for u and v of the frame coordinates), it becomes a simultaneous linear equation with an excess condition, and is obtained by the least square method as described above. Specifically, the position estimation unit 13 obtains an attitude (translation vector t and rotation matrix R) that minimizes the cost function E1 of the following equation (6). This is the attitude of the own machine in the SLAM coordinates obtained by the SLAM method (the position and orientation of the own machine represented by the translation vector t and the rotation matrix R). In this way, the position estimation unit 13 estimates the attitude of the own device (step S239).

  In the estimation of the position of the own machine, there may be cases where the estimated value of the attitude of the own machine is not fixed to one (if the condition becomes indefinite due to insufficient conditions (for example, the number of corresponding feature points in two images is less than 5) ), Or when the estimated value is not determined as one but becomes plural even if it is determined to be one by the least square method (for example, when the minimum value is larger than a predetermined value in Equation (6)) Or a minimum value (the smallest minimum value) divided by the second smallest minimum value is greater than a predetermined value). When the position estimation unit 13 estimates the attitude (translation vector t and rotation matrix R) of the own aircraft, if the estimated value is determined to be one, the estimation determination unit 14 determines that the estimation is possible, and the estimated value is one. If not determined, the estimation determination unit 14 determines that estimation is impossible (step S240).

  When the estimation determination unit 14 determines that the estimation is impossible (step S240; No), the process proceeds to step S247. If the estimation determination unit 14 determines that the estimation is possible (step S240; Yes), it means that the attitude (translation vector t and rotation matrix R) of the own device has been obtained, and the process proceeds to step S241.

  Since the current attitude (translation vector t and rotation matrix R) in the SLAM coordinates is obtained, the control unit 10 increments the position history registration counter HC and sets the HC-th item in the position history storage unit 24. The current attitude (translation vector t and rotation matrix R) of the own machine is registered together with a time stamp (the current time etc. can be compared with the time stamp of the position change history storage unit 23), and the position history search counter SC The value of the position history registration counter HC is set to (Step S241). In step S241, the control unit 10 corresponds to a position history registration unit. Then, the position estimation unit 13 obtains VO (visual odometry) by multiplying the current attitude of the own device by the scale S (step S242). The VO can be used as the position and orientation of the aircraft in the real environment.

  Next, the position estimation unit 13 moves more than a reference translational distance (for example, 1 m) from the position of the own device when the immediately preceding key frame registered in the frame DB (an image with a key frame number of NKF) is captured. (Step S243; Yes) After incrementing the key frame counter NKF (step S244), the current frame is registered as a key frame in the frame DB (step S245). . If it has moved less than the reference translation distance (step S243; No), the process returns to step S232.

  Here, the movement distance of the aircraft compared to the reference translation distance is the translation distance from the previous key frame to the current frame (the absolute value of the difference vector between the translation vectors of both frames (the square root of the sum of the squares of the elements)). It may be obtained from odometry, or may be obtained from VO (visual odometry) described above. As described above, the contents to be registered in the frame DB are “key frame number”, “posture”, “all extracted 2D feature points”, “a point where 3D position is known as a Map point among all 2D feature points” "," Characteristics of the key frame itself ".

  Then, the position estimation unit 13 sets the key frame counter NKF in the key frame queue of the map creation thread in order to inform the map creation thread that a new key frame has occurred (step S246). Then, the process returns to step S232. The key frame counter NKF, the scale S, the map point DB, the frame DB, the position history storage unit 24, the position change history storage unit 23, the position history registration counter HC, and the position history search counter SC are the position change history counter DC, The value is stored in the storage unit 20 so that the value can be referred to across threads.

  Next, step S247 and subsequent steps will be described. This is a process when the position of the aircraft cannot be estimated. First, the position estimation unit 13 sets “position estimation impossible” to the estimated state variable ES (step S247). Next, the position estimation unit 13 determines whether the operation is finished (step S248). If the operation ends (step S248; Yes), the operation ends. If the operation does not end (step S248; No), the imaging unit 41 acquires an image (step S249).

  Then, the control unit 10 searches the frame DB for key frames whose similarity to the acquired image is equal to or higher than a reference image similarity (for example, 0.9) (step S250). Here, when the features of the images (key frames) are represented by feature vectors, this similarity is obtained by normalizing the absolute values of the feature vectors of the two images (the square root of the sum of squares of the elements) to 1. Can be the similarity between the two images. Also, the reciprocal of the distance (square root of the sum of squares of the differences between the elements) of the feature vectors of the two images (the absolute value normalized to 1) may be used as the similarity.

  The control unit 10 determines whether a key frame whose similarity with the acquired image is equal to or higher than the reference image similarity is found from the frame DB (step S251). If not found (step S251; No), the process returns to step S248. If it is found (step S251; Yes), the 2D feature point of the acquired image and the 2D feature point of the found similar image are respectively acquired (step S252), and the correspondence of the 2D feature point is acquired between these two images. (Step S253).

  Then, the attitude of the aircraft is estimated by the same method as in step S239 (step S254). At this time, the estimation determination unit 14 determines whether or not estimation is possible in the same manner as in step S240 (step S255). If estimation is impossible (step S255; No), the process returns to step S248, and if estimation is possible (step S255; Yes) Since the machine position can be estimated, the process proceeds to step S231.

  As described above, after the initialization of the own position estimation thread, either “position estimation possible” or “position estimation impossible” is set in the estimation state variable ES. Therefore, in the main flow shown in FIG. If the ES value is “position estimation is possible”, the autonomous mobile device 100 performs a desired operation. If the ES value is “position estimation is impossible”, the autonomous mobile device 100 obtains a position where position estimation is possible. Control can be performed.

  Next, the map creation thread activated in step S102 of the main flow (FIG. 3) of the autonomous mobile device 100 will be described with reference to FIG. In this thread, the map creation unit 12 creates the map information (Map point DB) by calculating the 3D position of the corresponding feature point in the key frame.

  First, the map creation unit 12 determines whether the operation is finished (step S301). If the operation ends (step S301; Yes), the operation ends. If the operation does not end (step S301; No), it is determined whether the key frame queue is empty (step S302). If the key frame queue is empty (step S302; Yes), the process returns to step S301. If the key frame queue is not empty (step S302; No), data is extracted from the key frame queue and is processed by the MKF (key frame processed by the map creation thread). (A variable representing a number) (step S303). The map creation unit 12 determines whether MKF is greater than 0 (step S304). If MKF is 0 (step S304; No), the map creation unit 12 returns to step S301 and waits for data to enter the key frame queue. When MKF is 1 or more (step S304; Yes), the process proceeds to the following process.

  The map creation unit 12 refers to the frame DB, determines the 2D feature point of the previous key frame (key frame number MKF-1) and the 2D feature point of the current key frame (key frame number MKF) 2D feature points (corresponding feature points) that can be handled in step S305 are extracted. Since the posture (translation vector t and rotation matrix R) of each key frame is also registered in the frame DB, the 3D position of the corresponding feature point is processed in the same manner as in the process at the initialization of the own position estimation thread. Can be calculated. The map creation unit 12 registers the corresponding feature point for which the 3D position can be calculated as a Map point in the Map point DB (step S306). The map creation unit 12 registers the 3D position with respect to the 2D feature point for which the current 3D position can be calculated for the frame DB (step S307).

  If the corresponding feature point extracted by the map creation unit 12 has already been registered in the Map point DB, the 3D position calculation is skipped and processing for the next corresponding feature point (one not registered in the Map point DB) is performed. Alternatively, the 3D position may be calculated again, and the 3D position registered in the Map point DB or the 3D position for the corresponding feature point in the frame DB may be updated.

  Next, the map creating unit 12 determines whether or not the key frame queue is empty (step S308). If it is empty (step S308; Yes), bundle adjustment processing is performed on the orientation of all key frames and the 3D positions of all map points to improve accuracy (step S309), and the process proceeds to step S310. If the key frame queue is not empty (step S308; No), the process proceeds to step S310. Next, the map creating unit 12 sets MKF in the key frame queue of the loop closing thread (step S310), and returns to step S301.

  The bundle adjustment process is a nonlinear optimization method that simultaneously estimates the camera posture (key frame posture) and the 3D position of the Map point, and an error that occurs when the Map point is projected onto the key frame. Optimization is performed so as to minimize it.

  By performing the bundle adjustment process, it is possible to improve the accuracy of the 3D position of the key frame posture and the Map point. However, even if this process is not performed, the accuracy cannot be improved and a special problem does not occur. Therefore, even when there is no other process (for example, the key frame queue is empty), this process does not need to be performed every time.

  Further, when bundle adjustment processing is performed, a Map point may be found in which an error when projected onto a key frame is larger than a predetermined value. Such a Map point with a large error has an adverse effect on SLAM estimation. Therefore, a flag for identifying the Map point that requires a large error and is deleted from the Map point DB and the frame DB is set. You can put it. Note that the bundle adjustment process is treated as an option in the present embodiment, and the details of the process are omitted here.

  Next, the loop closing thread activated in step S103 of the main flow (FIG. 3) of the autonomous mobile device 100 will be described with reference to FIG. In this thread, the control unit 10 continues to check whether or not the loop closing process can be performed, and if possible, performs the loop closing process. Note that the loop closing process uses the difference between the posture value and the current posture value when you were in the same location before when you recognized that you have returned to the same location that you have been before. This refers to correcting the 3D position of the key frame and the related Map point in the trajectory from the previous time to the present.

  First, the control unit 10 determines whether the operation is finished (step S401). If the operation ends (step S401; Yes), the process ends. If the operation is not finished (step S401; No), it is determined whether or not the key frame queue is empty (step S402). If the key frame queue is empty (step S402; Yes), the process returns to step S401. If the key frame queue is not empty (step S402; No), data is extracted from the key frame queue and processed by the LKF (loop closing thread). It is set to a variable representing a key frame number (step S403). Next, the control unit 10 determines whether LKF is greater than 1 (step S404). If LKF is 0 or 1 (step S404; No), the process returns to step S401 and waits for data to enter the key frame queue. When LKF is 2 or more (step S404; Yes), the following processing is performed.

  The control unit 10 refers to the frame DB, and searches the frame DB for key frames in which the similarity between the current key frame (key frame with the key frame number LKF) and the “characteristic of the key frame itself” is equal to or higher than the reference image similarity. (Step S405).

  The control unit 10 determines whether or not a key frame having a similarity of “characteristic of the key frame itself” equal to or higher than the reference image similarity is found (step S406), and if not found (step S406; No), step S401. Returning to step S406 (Yes in step S406), the posture of the key frame in the locus from the found key frame to the current key frame and the 3D position of the Map point included in the key frame in the locus are corrected ( Step S407). For example, the control unit 10 corrects the posture of the current key frame as the same posture as the found key frame. Then, using the difference between the attitude of the found key frame and the attitude of the current key frame, linear correction is applied to the attitude of each key frame in the trajectory from the found key frame to the current key frame. Further, the 3D position of the Map point included in each key frame is also corrected according to the correction amount of the posture of each key frame. Then, the process returns to step S401.

  Next, the movement process control which is the process of step S110 of the main flow (FIG. 3) of the autonomous mobile device 100 will be described with reference to FIG. First, the control unit 10 determines whether or not the value of the position history search counter SC is 0 (step S501). If SC is 0, it means that the process of returning to all the positions stored in the position history storage unit 24 has been tried but the position estimation is not possible. In this case (step S501; Yes) In order to perform autonomous operation in a state where the position has not been estimated, the same operations as steps S108 and S109 of the main flow (FIG. 3) are performed (steps S502 and S503), and step S105 of the main flow (FIG. 3) is performed. Return.

  If the SC is not 0 (step S501; No), the SC-th information (information on the attitude (position) of the own device and the time at that time) stored in the position history storage unit 24 is acquired (step S504). Among the acquired information, the information on the posture (position) of the own device is stored in HP (a variable for storing information on the position of the own device acquired from the position history storage unit 24), and the time is acquired from the HT (position history storage unit 24). Stored in each variable). Then, it is determined whether or not the value of the position change history counter DC is 0 (step S505). If DC is 0, it means that all the changes in the position stored in the position change history storage unit 23 have been tried, but the position cannot be estimated. In this case (step S505; Yes), the position is estimated. In order to perform an autonomous operation in a state where the operation has not been performed, operations similar to steps S108 and S109 of the main flow (FIG. 3) are performed (steps S502 and S503), and the process returns to step S105 of the main flow (FIG. 3).

If DC is not 0 (step S505; No), DC-th information (position change amount and time at that time) stored in the position change history storage unit 23 is acquired (step S506).
. Of the acquired information, the position change amount is DP (a variable for storing the position change amount acquired from the position change history storage unit 23), and the time is DT (a variable for storing the time acquired from the position change history storage unit 23). Respectively. Then, it is determined whether the time HT acquired from the position history storage unit 24 is a past time (smaller value) than the time DT acquired from the position change history storage unit 23 (step S507).

  If the HT is older than the DT (step S507; Yes), the own device has not yet returned to the past HP position acquired from the position history storage unit 24, and the process proceeds to step S508. If HT is the same time as DT or if DT is in the past (step S507; No), the process proceeds to step S510 to perform the last movement process in the movement process control.

  In step S508, based on the position change amount DP registered in the position change history storage unit 23, the movement processing control unit 15 causes the movement processing unit 11 to perform a movement process that cancels the change amount (step S508). For example, if the DP is “advance 10 cm forward”, the moving process here is a process “return 10 cm backward”. Then, the position change history counter DC is decremented (step S509), and the process returns to step S505.

  Similarly to step S508, based on the position change amount DP registered in the position change history storage unit 23, the movement processing control unit 15 causes the movement processing control unit 15 to perform a movement process that cancels the change amount in step S510 (step S510). S510). By this movement processing, the autonomous mobile device 100 (if there is no error in the information stored in the position change history storage unit 23 and the movement distance of the movement processing unit 11) has been able to estimate the position (HP It is desirable to perform the last movement process at a reduced speed so that the own position estimation in the own position estimation thread is correctly performed. In particular, if the values of HT and DT are equal at the time of the determination in step S507, the movement should return to the position indicated by HP, so between step S510 and step S511 (with the own position estimation thread) A process of stopping without moving for a predetermined time may be added (in order to wait for the self position estimation).

  Next, the position change history counter DC is decremented (step S511), the position history search counter SC is decremented (step S512), and the process returns to step S105 of the main flow. The position history search counter SC is decremented if the current position estimation thread is not enabled by the current position estimation thread even in the current movement process control, and in the next movement process control, the position is further returned to the past position. This is to try to estimate the position of the aircraft. However, if the position estimation is successful until the next time, the value of the position history registration counter HC is set in the position history search counter SC in step S241 in FIG. By performing such movement processing control, it is possible to return to the position where the position of the own apparatus has been estimated in the past and enable the position of the own apparatus to be estimated.

  Next, as a movement process control different from the above, a second embodiment for performing the movement process control shown in the flowchart of FIG. 9 will be described. The second embodiment is the same as the first embodiment except for the flowchart of movement processing control.

  First, the control unit 10 determines whether or not the value of the position history search counter SC is 0 (step S601). If SC is 0, it means that the process of returning to all the positions stored in the position history storage unit 24 has been attempted, but the position has not been estimated. In this case (step S601; Yes). In order to perform autonomous operation in a state where the position has not been estimated, operations similar to steps S108 and S109 of the main flow (FIG. 3) are performed (steps S602 and S603), and step S105 of the main flow (FIG. 3) is performed. Return.

  If SC is not 0 (step S601; No), the SC-th information stored in the position history storage unit 24 (information HP of the attitude (translation vector t and rotation matrix R) of the own device and the time HT at that time) is displayed. Obtain (step S604). Then, the value of the position change history counter DC is set in the position change history work counter WC (a variable used when acquiring the position change history from the position change history storage unit 23), and the position change history stack (stores the position change amount). Stack) to be initialized (step S605). The stack has a first-in last-out data structure.

  Then, the control unit 10 determines whether or not the position change history work counter WC is 0 (step S606). If WC is 0, it means that all the changes in the position stored in the position change history storage unit 23 have been tried but the position cannot be estimated. In this case (step S606; Yes), the position is estimated. In order to perform an autonomous operation in a state in which it has not been performed, operations similar to steps S108 and S109 of the main flow (FIG. 3) are performed (steps S602 and S603), and the process returns to step S105 of the main flow (FIG. 3).

  If WC is not 0 (step S606; No), the control unit 10 acquires WC-th information (position change amount DP and time DT at that time) stored in the position change history storage unit 23 (step S607). Then, it is determined whether the time HT acquired from the position history storage unit 24 is a past time (smaller value) than the time DT acquired from the position change history storage unit 23 (step S608).

  If the HT is older than the DT (step S608; Yes), the own apparatus has not yet returned to the past HP position acquired from the position history storage unit 24, and the process proceeds to step S609. If HT is the same time as DT or if DT is in the past (step S608; No), the process proceeds to step S611 to estimate the position of the own device from the position change history.

  In step S609, the control unit 10 pushes the value of the position change amount DP onto the position change history stack. Then, the position change history work counter WC is decremented (step S610), and the process returns to step S606. As a result, the position change so far is accumulated in the position change history stack.

  In step S611, the control unit 10 sets the past HP posture (translation vector t and rotation matrix R) acquired from the position history storage unit 24 as an initial value MP (variable used for obtaining an estimated value of the own device position). The data (DP) is extracted from the position change history stack and added to MP until the position change history stack becomes empty. Thereby, the estimated value of the current own device position obtained by integrating the position change history DP is obtained from the position HP when the previous own device position can be estimated. In this process, the control unit 10 corresponds to an integrated position estimation unit.

  Next, the movement processing control unit 15 controls the movement processing unit 11 so as to move from the estimated value MP of the current own device position to the past own device position HP (step S612). Specifically, translation is performed by subtracting the MP translation vector from the HP translation vector, and the rotation represented by the rotation matrix obtained by multiplying the HP rotation matrix by the inverse matrix of the MP rotation matrix. Control is performed so that the rotation is performed by the amount. By this movement processing control, the autonomous mobile device 100 (if there is no error in the information stored in the position change history storage unit 23 and the movement distance of the movement processing unit 11) has previously been able to estimate the position (HP The position is not moved for a predetermined time between steps S612 and S613 (waiting for the own position estimation in the own position estimation thread). Processing to stop may be added.

  Next, the position history search counter SC is decremented (step S613), and the process returns to step S105 of the main flow. The position history search counter SC is decremented if the current position estimation thread is not enabled by the current position estimation thread even in the current movement process control, and in the next movement process control, the position is further returned to the past position. This is to try to estimate the position of the aircraft. However, if the position estimation is successful until the next time, the value of the position history registration counter HC is set in the position history search counter SC in step S241 in FIG. By performing such movement processing control, it is possible to return to the position where the position of the own apparatus has been estimated in the past and enable the position of the own apparatus to be estimated.

  Compared with the first embodiment, the second embodiment does not perform movement processing for each position change stored in the position change history storage unit 23 one by one, but performs movement processing collectively. There is an advantage that errors of the drive unit 42 during control are difficult to accumulate.

  In the two embodiments described so far, both have performed processing for returning to the position where the own position has been estimated in the past, but these are information stored in the position change history storage unit 23 and the movement processing unit. There is a problem that if the moving distance by 11 is not accurate, it cannot return to the past correct position.

  Therefore, a third embodiment that does not require the position change history storage unit 23 and the position history storage unit 24 will be described. When the configuration of the third embodiment is compared with the first and second embodiments shown in FIG. 10, the point that the rotation photographing unit 16 exists in the control unit 10 and the position change history in the storage unit 20. The storage unit 23 and the position history storage unit 24 do not exist, and the feature point storage unit 25 is different.

  For each flow of FIGS. 3 to 7, the processing of the third embodiment is simply performed by omitting processing relating to the position change history counter DC, the position change history storage unit 23, the position history registration counter HC, and the position history search counter SC. Same as flow. Since the process flow of the movement process control is greatly different from that of the first and second embodiments, it will be described below with reference to FIG.

First, the rotation photographing unit 16 sets 1 to the rotation photographing flag RF (variable that indicates whether the processing of the current rotation shooting) (step S701), cleared to zero the cumulative rotation angle phi _S (step S702). Next, an image is acquired by the imaging unit 41 (step S703). Next, 2D feature points are acquired from the acquired image (step S704). Then, to register the number of the acquired 2D feature points and the accumulated rotation angle phi _S to the feature point storage portion 25 (step S705).

Next, the rotational imaging unit 16 determines whether or not the rotational imaging flag RF is 2 (step S706). If RF is 2 (step S706; Yes), it means that imaging for one rotation has been completed, and the process proceeds to step S712. If RF is not 2 (step S706; No), it is determined whether the cumulative rotation angle φ _S + rotation angle φ ₁ is less than 360 ° (step S707). The rotation angle phi ₁ is the rotation angle of the rotary operation dose by rotating imaging unit 16, for example, 30 °. It is desirable to set φ _{1 to} an angle equal to or less than half of the viewing angle of the imaging unit 41 so that as many feature points as possible can be photographed from the surrounding environment.

If the cumulative rotation angle φ _S + rotation angle φ ₁ is less than 360 ° (step S707; Yes), the rotation photographing unit 16 instructs the drive unit 42 to perform the rotation operation with the rotation angle φ ₁ ( step S 708), then return is added to the rotational angle phi ₁ to the cumulative rotation angle phi _S (step S709) to step S703. The rotation operation here is preferably a rotation operation that does not move the position if the two wheels of the drive unit 42 can be rotated without moving the position by rotating them at the same speed in the opposite direction. When the two wheels can only be rotated in the same direction, it is desirable to perform a rotational operation with as small a rotation radius as possible.

When the cumulative rotation angle φ _S + rotation angle φ ₁ is equal to or greater than 360 ° (step S707; No), the rotation imaging unit 16 causes the drive unit 42 to rotate at a rotation angle (360 ° −φ _S ). (Step S710), the rotational shooting flag RF is set to 2 (step S711), and the process returns to step S703.

In step S712, the movement processing control unit 15 searches the feature point storage unit 25 for the accumulated rotation angle φ _S when the number of 2D feature points registered in the feature point storage unit 25 is the largest, and rotates the rotation angle φ _S. The movement processing unit 11 is controlled to operate (step S713). Thereafter, the movement processing control unit 15 controls the movement processing unit 11 so as to move slowly forward by a predetermined distance (for example, three times the reference translational distance) (step S714). By slowly moving, it is possible to increase the timing at which the own position can be estimated by the own position estimation thread.

Thereafter, the rotational imaging unit 16 clears the rotational imaging flag RF to 0 (step S715), and returns to step S105 of the main flow. By performing such an operation, since the movement is performed in the direction in which the number of 2D feature points is large, the possibility that the position of the own device is successfully estimated increases. In this embodiment, attention is paid only to the number of 2D feature points. However, focusing on not only the number but also quality, an evaluation function is defined by the number and quality of 2D feature points. In step S712, the 2D feature points are defined. A process for searching for the rotation angle φ _{S at} which the evaluation function becomes the highest may be performed.

  Further, the movement processing control of each of the first, second, and third embodiments may be switched as appropriate. For example, the movement processing control of the first embodiment is performed first, and if the position estimation of the device itself is not possible, the movement processing control of the third embodiment is performed first. The movement processing of the second embodiment is first performed. If control does not make it possible to estimate the position of the aircraft, the embodiment that performs the movement processing control of the third embodiment, the movement processing control of the first embodiment is performed first and the position of the aircraft itself can still be estimated. If not, the movement processing control of the second embodiment is performed, and if the position estimation of the device itself is not possible, the embodiment of performing the movement processing control of the third embodiment is conceivable.

  In each of the above-described embodiments, the reference translation distance is set to “1 m”, for example. However, the optimum reference translation distance varies depending on the size and speed of the autonomous mobile device itself, the environment in which it moves autonomously, the movement range, and the like. A supplementary description will be given of the method for setting the reference translation distance.

  As a method for setting the reference translation distance, a value obtained by averaging the distances from the 3D position of all Map points on the key frame (2D feature points whose 3D position is known) to the own position (= all the points observed in the key frame) It may be set by a ratio of the average distance of the depth to the Map point) and the translation distance (for example, a distance of 5% of the average distance of the depth to the Map point is set as the reference translation distance). The “all Map points” used here can be extracted by the control unit 10 referring to the Map point DB. In this process, the control unit 10 corresponds to a position estimable feature point extraction unit. Further, according to the size of the translation distance in the actual environment, for example, 10 cm may be set as the reference translation distance if moving on the table, 1 m if moving around the room, 10 m if moving around the open space, etc. Then, a value corresponding to the wheel diameter of the drive unit 42 (for example, 10 times the wheel diameter) may be set as the reference translational distance.

  When the reference translation distance is set as a distance in the actual environment, the translation distance in the SLAM coordinates is multiplied by the scale S and converted into the distance in the actual environment (the reverse is also true, When setting the distance in the SLAM coordinates as the reference translation distance, the distance in the real environment is divided by the scale S and converted into the distance in the SLAM coordinates for comparison.

  Each function of the autonomous mobile device 100 of the present invention can also be implemented by a computer such as a normal PC (Personal Computer). Specifically, in the above embodiment, the autonomous movement control processing program performed by the autonomous mobile device 100 has been described as being stored in advance in the ROM of the storage unit 20. However, the program is stored and distributed on a computer-readable recording medium such as a flexible disk, CD-ROM (Compact Disc Read Only Memory), DVD (Digital Versatile Disc), and MO (Magneto-Optical Disc). A computer capable of realizing each of the functions described above may be configured by reading and installing the program on a computer.

  As mentioned above, although preferable embodiment of this invention was described, this invention is not limited to the specific embodiment which concerns, This invention includes the invention described in the claim, and its equivalent range It is. Hereinafter, the invention described in the scope of claims of the present application will be appended.

(Appendix 1)
A movement processing unit for processing to move autonomously based on information of a plurality of images taken by the imaging unit;
A position estimation unit that estimates the position of the device using information of a plurality of images captured by the imaging unit;
An estimation determination unit that determines whether or not the estimation result of the own device position in the position estimation unit is determined as one;
When the estimation determination unit determines that the estimation result is not fixed to one, the movement processing unit is controlled so that the position estimation unit moves to a position where the estimation result of the own position is determined to be one. A movement processing control unit,
An autonomous mobile device comprising:

(Appendix 2)
The autonomous mobile device further includes a corresponding feature point acquisition unit that extracts corresponding feature points from a plurality of images,
When the number of corresponding feature points extracted by the corresponding feature point acquisition unit from a plurality of images used by the position estimation unit is less than a predetermined number, the estimation determination unit is not limited to one estimation result. judge,
The autonomous mobile device according to attachment 1.

(Appendix 3)
The autonomous mobile device further includes a corresponding feature point acquisition unit that extracts corresponding feature points from a plurality of images,
The position estimation unit estimates the own device position by minimizing a cost function defined based on the coordinates in the image of the feature points extracted by the corresponding feature point acquisition unit,
The estimation determination unit determines that the estimation result is not fixed to one when the minimum value of the cost function is larger than a predetermined value.
The autonomous mobile device according to attachment 1.

(Appendix 4)
The autonomous mobile device further includes a corresponding feature point acquisition unit that extracts corresponding feature points from a plurality of images,
The position estimation unit estimates the own device position by minimizing a cost function defined based on the coordinates in the image of the feature points extracted by the corresponding feature point acquisition unit,
The estimation determination unit determines the estimation result as one when the cost function has a plurality of minimum values and a value obtained by dividing the minimum minimum value by the second smallest minimum value is larger than a predetermined value. It is determined that
The autonomous mobile device according to attachment 1.

(Appendix 5)
The autonomous mobile device further includes a position change history registration unit that registers the amount of change in the position change history storage unit each time the position of the own machine is changed by the movement processing unit;
When the estimation determination unit determines that the estimation result is determined as one, a position history registration unit that registers position information at that time together with time information in a position history storage unit;
With
When the estimation determination unit determines that the estimation result is not fixed to one, the movement processing control unit traces back to the time when the device is at the position registered in the position history storage unit, Controlling the movement processing unit to move in the direction opposite to the amount of change in position registered in the position change history storage unit;
The autonomous mobile device according to any one of appendices 1 to 4.

(Appendix 6)
The autonomous mobile device further includes a position change history registration unit that registers the amount of change in the position change history storage unit each time the position of the own machine is changed by the movement processing unit;
When the estimation determination unit determines that the estimation result is determined as one, a position history registration unit that registers position information at that time together with time information in a position history storage unit;
When the estimation determination unit determines that the estimation result is not fixed to one, the position change history storage unit stores the position registered in the position history storage unit from the time corresponding to the position to the current time. An integrated position estimation unit that estimates the current position of the own aircraft by integrating the amount of change in the registered position;
With
When the estimation determination unit determines that the estimation result is not fixed to one, the movement processing control unit changes the position registered in the position history storage unit from the position estimated by the integrated position estimation unit. Controlling the movement processing unit to move toward
The autonomous mobile device according to any one of appendices 1 to 4.

(Appendix 7)
The autonomous mobile device further includes a rotational imaging unit that captures an image while performing a rotational operation,
When the estimation determination unit determines that the estimation result is not fixed to one, the movement processing control unit includes information used by the position estimation unit for estimation in an image captured by the rotational imaging unit. Controlling the movement processing unit to move in the direction when the most images were taken;
The autonomous mobile device according to any one of appendices 1 to 4.

(Appendix 8)
The rotation angle of one rotation operation performed by the rotational imaging unit is an angle that is half or less of the viewing angle of the imaging unit.
The autonomous mobile device according to appendix 7.

(Appendix 9)
The autonomous mobile device further includes an image storage unit that stores information of an image captured by the imaging unit each time the reference translation distance is moved,
The position estimating unit estimates the position of the device using information of a plurality of images stored in the image storage unit;
The autonomous mobile device according to any one of appendices 1 to 8.

(Appendix 10)
A value obtained by multiplying the diameter of the wheel of the drive unit by a predetermined value is set as the reference translational distance.
The autonomous mobile device according to attachment 9.

(Appendix 11)
The autonomous mobile device further includes a position estimable feature point extraction unit that extracts a feature point whose position can be estimated from a plurality of images.
A value obtained by dividing the average value of the distances between all the feature points extracted by the position estimable feature point extraction unit and the own machine position by a predetermined value is set as the reference translation distance.
The autonomous mobile device according to attachment 9.

(Appendix 12)
An autonomous movement method for autonomously moving based on information of a plurality of images taken by an imaging unit,
When the estimation result of the position of the own device based on information of a plurality of images taken by the imaging unit is not fixed to one, move to search for a position where the estimation result is fixed to one,
Autonomous movement method.

(Appendix 13)
A movement processing step for processing to move autonomously based on information of a plurality of images captured by the imaging unit;
A position estimation step of estimating the position of the own device based on information of a plurality of images captured by the imaging unit;
An estimation determination step for determining whether or not the estimation result of the own device position in the position estimation step is determined as one;
If it is determined in the estimation determination step that the estimation result is not fixed to one, the movement processing step is performed so that the estimation result of the own position in the position estimation step is moved to a single position. A movement processing control step for controlling the processing of
An autonomous movement method comprising:

(Appendix 14)
On the computer,
A movement processing step for processing to move autonomously based on information of a plurality of images captured by the imaging unit;
A position estimation step of estimating the position of the own device based on information of a plurality of images captured by the imaging unit;
An estimation determination step for determining whether or not the estimation result of the own device position in the position estimation step is determined as one;
If it is determined in the estimation determination step that the estimation result is not fixed to one, the movement processing step is performed so that the estimation result of the own position in the position estimation step is moved to a single position. A movement process control step for controlling the process of
A program for running

DESCRIPTION OF SYMBOLS 100 ... Autonomous mobile device, 10 ... Control part, 11 ... Movement process part, 12 ... Map preparation part, 13 ... Position estimation part, 14 ... Estimation determination part, 15 ... Movement process control part, 16 ... Rotation imaging part, 20 ... Storage unit, 21 ... Image storage unit, 22 ... Map storage unit, 23 ... Position change history storage unit, 24 ... Position history storage unit, 25 ... Feature point storage unit, 30 ... Sensor unit, 31 ... Acceleration sensor, 32 ... Angular velocity Sensor, 33 ... Obstacle sensor, 41 ... Imaging unit, 42 ... Drive unit, 43 ... Input unit, 44 ... Communication unit, 45 ... Power source

In order to achieve the above object, the autonomous mobile device of the present invention provides:
A movement processing unit for processing to move autonomously based on information of a plurality of images taken by the imaging unit;
A position estimation unit that estimates the position of the device using information of a plurality of images captured by the imaging unit;
An estimation determination unit that determines whether or not the estimation result of the own device position in the position estimation unit is determined as one;
When the estimation determination unit determines that the estimation result is not fixed to one, the movement processing unit is controlled so that the position estimation unit moves to a position where the estimation result of the own position is determined to be one. A movement processing control unit,
A position change history registration unit that registers the amount of change in the position change history storage unit each time the position of the own machine is changed by the movement processing unit ;
With
The movement processing control unit, when the estimated determination section determines that the estimation result is not determined to one, move in the opposite direction to the previous SL position change history change amount of the position registered in the storage unit The movement processing unit is controlled as described above.

Claims (14)

A movement processing unit for processing to move autonomously based on information of a plurality of images taken by the imaging unit;
A position estimation unit that estimates the position of the device using information of a plurality of images captured by the imaging unit;
An estimation determination unit that determines whether or not the estimation result of the own device position in the position estimation unit is determined as one;
When the estimation determination unit determines that the estimation result is not fixed to one, the movement processing unit is controlled so that the position estimation unit moves to a position where the estimation result of the own position is determined to be one. A movement processing control unit,
A position change history registration unit that registers the amount of change in the position change history storage unit each time the position of the own machine is changed by the movement processing unit;
When the estimation determination unit determines that the estimation result is determined as one, a position history registration unit that registers position information at that time together with time information in a position history storage unit;
With
In the case where the estimation determination unit determines that the estimation result is not fixed to one, the movement processing control unit goes back to the time when the device is at the position registered in the position history storage unit,
Controlling the movement processing unit to move in a direction opposite to the amount of change in position registered in the position change history storage unit;
Autonomous mobile device. A movement processing unit for processing to move autonomously based on information of a plurality of images taken by the imaging unit;
A position estimation unit that estimates the position of the device using information of a plurality of images captured by the imaging unit;
An estimation determination unit that determines whether or not the estimation result of the own device position in the position estimation unit is determined as one;
When the estimation determination unit determines that the estimation result is not fixed to one, the movement processing unit is controlled so that the position estimation unit moves to a position where the estimation result of the own position is determined to be one. A movement processing control unit,
A position change history registration unit that registers the amount of change in the position change history storage unit each time the position of the own machine is changed by the movement processing unit;
When the estimation determination unit determines that the estimation result is determined as one, a position history registration unit that registers position information at that time together with time information in a position history storage unit;
When the estimation determination unit determines that the estimation result is not fixed to one, the position change history storage unit stores the position registered in the position history storage unit from the time corresponding to the position to the current time. An integrated position estimation unit that estimates the current position of the own aircraft by integrating the amount of change in the registered position;
With
When the estimation determination unit determines that the estimation result is not fixed to one, the movement processing control unit changes the position registered in the position history storage unit from the position estimated by the integrated position estimation unit. Controlling the movement processing unit to move toward
Autonomous mobile device. The autonomous mobile device further includes a corresponding feature point acquisition unit that extracts corresponding feature points from a plurality of images,
When the number of corresponding feature points extracted by the corresponding feature point acquisition unit from a plurality of images used by the position estimation unit is less than a predetermined number, the estimation determination unit is not limited to one estimation result. judge,
The autonomous mobile device according to claim 1 or 2. The autonomous mobile device further includes a corresponding feature point acquisition unit that extracts corresponding feature points from a plurality of images,
The position estimation unit estimates the own device position by minimizing a cost function defined based on the coordinates in the image of the feature points extracted by the corresponding feature point acquisition unit,
The estimation determination unit determines that the estimation result is not fixed to one when the minimum value of the cost function is larger than a predetermined value.
The autonomous mobile device according to claim 1 or 2. The autonomous mobile device further includes a corresponding feature point acquisition unit that extracts corresponding feature points from a plurality of images,
The position estimation unit estimates the own device position by minimizing a cost function defined based on the coordinates in the image of the feature points extracted by the corresponding feature point acquisition unit,
The estimation determination unit determines the estimation result as one when the cost function has a plurality of minimum values and a value obtained by dividing the minimum minimum value by the second smallest minimum value is larger than a predetermined value. It is determined that
The autonomous mobile device according to claim 1 or 2. A rotational imaging unit that captures an image while performing a rotational operation with the cumulative rotation angle from the initial value as φ _S and the rotation angle for _one rotation as φ ₁ ;
A feature point storage unit that acquires feature points from an image captured by the rotation imaging unit and stores the number of the feature points and the cumulative rotation angle (φ _S ) together;
With
The movement processing control unit, when the estimation determination unit determines that the estimation result is not fixed to one, the cumulative rotation when the number of feature points stored in the feature point storage unit is the largest. Corner (
(φ _S ) is retrieved, and the rotational imaging unit is caused to perform a rotational operation based on the retrieved cumulative rotational angle (φ _S ).
The autonomous mobile device according to claim 1 or 2. The movement processing control unit, the rotating imaging unit, wherein when the accumulated rotation angle phi _{S +} rotation angle phi ₁ of the dose is less than 360 ° causes the rotational movement of the rotation angle phi ₁ of the dose When the cumulative rotation angle φ _S + the one rotation angle φ ₁ is 360 ° or more, the rotation angle (360 ° −
The autonomous mobile device according to claim 6, wherein a rotational movement of φ _S ) is performed. The rotation angle of one rotation operation performed by the rotational imaging unit is an angle that is half or less of the viewing angle of the imaging unit.
The autonomous mobile device according to claim 6. The autonomous mobile device further includes an image storage unit that stores information of an image captured by the imaging unit each time the reference translation distance is moved,
The position estimating unit estimates the position of the device using information of a plurality of images stored in the image storage unit;
The autonomous mobile device according to any one of claims 1 to 8. A value obtained by multiplying the diameter of the wheel of the drive unit by a predetermined value is set as the reference translational distance.
The autonomous mobile device according to claim 9. The autonomous mobile device further includes a position estimable feature point extraction unit that extracts a feature point whose position can be estimated from a plurality of images.
A value obtained by dividing the average value of the distances between all the feature points extracted by the position estimable feature point extraction unit and the own machine position by a predetermined value is set as the reference translation distance.
The autonomous mobile device according to claim 9. A movement processing step for processing to move autonomously based on information of a plurality of captured images;
A position estimation step for estimating the position of the device based on information of the plurality of captured images;
An estimation determination step for determining whether or not the estimation result of the own device position in the position estimation step is determined as one;
If it is determined in the estimation determination step that the estimation result is not fixed to one, the movement processing step is performed so that the estimation result of the own position in the position estimation step is moved to a single position. A movement processing control step for controlling the processing of
A position change history registration step of registering the amount of change in the position change history storage unit each time the position of the own machine is changed in the movement processing step;
When it is determined that the estimation result is determined as one in the estimation determination step, the position history registration step of registering the position information at that time together with time information in the position history storage unit;
With
In the movement processing control step, when it is determined that the estimation result is not fixed to one in the estimation determination step, while going back to the time when the own device was at the position registered in the position history storage unit, Controlling the processing in the movement processing step so as to move in the direction opposite to the amount of change in position registered in the position change history storage unit;
Autonomous movement method. A movement processing step for processing to move autonomously based on information of a plurality of captured images;
A position estimation step for estimating the position of the device based on information of the plurality of captured images;
An estimation determination step for determining whether or not the estimation result of the own device position in the position estimation step is determined as one;
If it is determined in the estimation determination step that the estimation result is not fixed to one, the movement processing step is performed so that the estimation result of the own position in the position estimation step is moved to a single position. A movement processing control step for controlling the processing of
A position change history registration step of registering the amount of change in the position change history storage unit each time the position of the own machine is changed in the movement processing step;
When it is determined that the estimation result is determined as one in the estimation determination step, the position history registration step of registering the position information at that time together with time information in the position history storage unit;
When it is determined in the estimation determination step that the estimation result is not fixed to one, the position change history storage unit from the time corresponding to the position to the current time is registered at the position registered in the position history storage unit. An integrated position estimation step for estimating the current position of the own aircraft by integrating the amount of change in the position registered in
With
In the movement process control step, when it is determined in the estimation determination step that the estimation result is not fixed to one, it is registered in the position change history storage unit from the position estimated in the integrated position estimation step. Controlling the processing in the movement processing step so as to move toward the position where
Autonomous movement method. A program for causing a computer to execute the autonomous movement method according to claim 12 or 13.

Images