JP2013030162A - Moving object motion control device and sloshing control device using the same - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a moving object motion control device capable of precisely controlling a moving object without using a method for extracting feature quantity on the basis of image information reflecting a state of the moving object.SOLUTION: In a sloshing control device including a water bath 10 for housing water, a driving part 12 for reciprocating the water bath 10 in one direction, a camera 14 for visually recognizing the water bath 10, and a control part 16 for controlling the water bath 10 by the driving part 12 on the basis of image information acquired by the camera to settle water in the water bath 10 from a swaying state to a static state, the control part 16 interprets image information of water in the water bath 10 sequentially acquired from the camera 14 according to a predetermined period as one point in a Hilbert space except a numerical vector space, and controls a motor 12 so as to settle the water in the water bath 10 from the swaying state to the static state by a stabilization control method.

Description

  The present invention relates to a control device that drives and controls a moving body based on image information, and a sloshing control device that uses this control method.

  A method of preventing vibration of fluids such as molten metal used for casting, oil contained in tanks, and water (sloshing prevention) by installing a damping device or monitoring and controlling the surface of oil or water Is considered. The vibration control device is intended to prevent the oil and water contained in the container from shaking by suppressing the vibration of the container itself, but it is used by moving the container itself as in the case of transporting molten metal. In this case, by controlling the movement of the container, it is possible to convey the oil and water with as little swing as possible. In other words, such control actively uses the movement of the moving body to suppress oil and water oscillation.

  In order to control the movement of the moving body, which changes every moment, the movement state of the moving body, that is, the current state of the moving body is detected, and the moving body is gradually guided to the target operation state (stabilization). Control). For example, in the sloshing prevention method, attention is paid to the position of the center of gravity of the fluid contained in the container, and control is performed so that the center of gravity that fluctuates with the oscillation of the fluid in the container converges to the position of the center of gravity in the stationary state of the fluid. There is a method (Patent Document 1, etc.). Such a control method is a control method in which a specific feature amount (in this example, the center of gravity position) is extracted from the operating state of the moving body, and the feature amount is stabilized to a specific target value.

  There are various sensing methods for detecting the operating state of a moving object that is a controlled object. One method is to monitor the moving body with a camera and detect the state of the moving body from the obtained image information of the moving body. Instead of monitoring (viewing) the moving body, it is also possible to detect the operation state (moving position, etc.) of the moving body based on image information from a camera provided on the moving body (automobile, robot, etc.). The conventional method for controlling the moving body by analyzing the operation state of the moving body based on the image information detects the feature quantity such as the above-mentioned center of gravity position from the image information, and moves the moving body based on the feature quantity. It is a method to control.

Japanese Patent Laid-Open No. 9-10924 JP 2004-264925 A JP-A-6-89103

  The above-described method for extracting a feature amount from image information indicating an operation state of a moving body and controlling the moving body based on the feature amount requires that the feature amount be artificially defined by an object obtained from the image information. There is a problem that it cannot be applied universally as a method of controlling a moving body based on image information. In other words, even though it can be applied to image information that is easy to define feature values, it is difficult to apply to image information that is difficult to define feature values, and the definition of feature values is not made accurately and effective control cannot be performed. There was a problem.

  The present invention relates to an operation of a moving body that enables accurate control of the moving body without using a technique of extracting a feature amount based on image information obtained as data reflecting the state of the moving body. It is an object of the present invention to provide a control device and a sloshing control device as an application example thereof.

The moving body motion control apparatus according to the present invention includes a moving body that is a control object, a driving unit that drives the moving body, a visual recognition unit that visually recognizes a detection target that varies with the movement of the moving body, A control unit that controls the movement of the moving body via the driving unit based on image information by the visual recognition unit, and the control unit captures individual image information sequentially taken from the visual recognition unit according to a predetermined period, It is interpreted as one point in the Hilbert space excluding the number vector space, and the detection target is guided to the target state by the stabilization control method.
In the Hilbert space, various spaces such as a number vector space, a polynomial space, and an execution sequence space are defined. The control method using the number vector space corresponds to a method of extracting and controlling so-called feature values. The present invention is a method for controlling image information by interpreting it as one point in a Hilbert space such as a polynomial space or an execution sequence space, excluding a number vector space, and a method for controlling the image information without extracting feature values from the image information. It is.
The stabilization control method is a control method for setting the moving body in a desired state, and an LQG control method, an ^H∞ control method, and the like can be used.

Further, the sloshing control device according to the present invention is a water tank in which the mobile body contains water, a drive unit in which the drive unit reciprocates the water tank in one direction, and the visual recognition unit is a camera that visually recognizes the water tank, The control unit is configured as a sloshing control device that controls the moving body by the driving unit based on image information acquired by the camera and sets the water in the water tank from a rocking state to a stationary state, and the control unit Interprets the image information of the water in the aquarium sequentially taken from the camera according to a predetermined period as one point in the Hilbert space excluding the number vector space, and swings the water in the aquarium by the stabilization control method. The drive unit is controlled to settle from a state to a stationary state.
In the sloshing control device, the control unit extracts one point of the polynomial space from the image information of the water in the aquarium, and monitors and controls the one point to oscillate the water in the aquarium. In the method of settling to a stationary state, or in the control unit, by extracting one point of the matrix space from the image information of the water in the aquarium and monitoring and controlling that one point, the water in the aquarium is It is possible to effectively control by setting from the swinging state to the stationary state.

  The movement control device for a moving object according to the present invention controls the moving object by interpreting and analyzing image information as one point on the Hilbert space, and thus without performing an operation of extracting feature amounts from the image information. There is an advantage that the moving body can be controlled by performing general-purpose analysis even when various pieces of image information are captured.

It is a block diagram which shows the structure of the sloshing control apparatus used for the test. It is an example of the image of the state which the wave produced in the water tank. It is a graph which shows the measurement result of a water surface, and an approximate curve about the image of FIG. It is the graph which compared the output of the identification model with the actual output. It is the graph which expanded and showed the range for 50 to 150 seconds in FIG. It is the graph which compared the output of the identification model at the time of making a gravity center into control object, and an actual output. It is the graph which expanded and showed the range for 50 to 150 seconds in FIG. It is a graph which shows the time change of command voltage. It is a graph which shows the time change of the component of polynomial space. It is a graph which shows the time change of a gravity center. It is an image of the state which blocked the front of the aquarium. It is a graph which shows the time change of the rotation angle of a motor. FIG. 6 is an image diagram in which image information is grayscaled and reduced to a (6, 6)) type matrix. It is the graph which compared the output of the identification model with the actual output. It is a graph which shows the norm of the matrix M (t) and estimation matrix ^ M (t) in the case of the control method according to the present invention. It is a graph which shows the norm of the water surface coordinate y (t) in the case of being based on the conventional control method, and the estimated coordinate ^ y (t). It is a graph which shows the input voltage of the motor in the case of the control method (Proposal) based on this invention, and the conventional control method (Conventional). It is a graph which shows the movement state of the water tank in the case of the control method which concerns on this invention, and the conventional control method. It is a graph which shows the input voltage of the motor in the case of LQG control and P control. It is a graph which shows the movement state of the water tank in the case of LQG control and P control.

(Configuration of sloshing control device)
As an application example of the moving body motion control apparatus according to the present invention, an example in which a sloshing control apparatus is configured will be described.
FIG. 1 is a block diagram showing the configuration of the sloshing control device. The sloshing control device of the present embodiment includes a water tank 10 containing water, a motor 12 that drives the water tank 10 to move forward and backward in one direction, a camera 14 that is disposed in front of the water tank 10 and photographs the entire front surface of the water tank 10, and a camera. 14 includes a control unit 16 that analyzes an image obtained from the image 14, and controls the drive of the motor 12 based on the analysis result. In the sloshing control device of the present embodiment, the water tank 10 is a moving object to be controlled, the motor 12 is a driving unit for the moving object, and the camera 14 is a viewing part. In this case, the detection target that fluctuates with the movement of the moving body is the state of water (wave state) in the water tank 10.

  The water tank 10 is formed in a rectangular parallelepiped container made of a transparent glass plate, and is supported on a support plate made of a horizontally supported flat plate so that the reciprocating direction and the longitudinal direction of the support plate are parallel to each other. The reciprocating direction of the support plate is defined in one direction by a guide rail fixed to the bottom surface of the support plate. The support plate is connected to a rack gear connected to the output shaft of the motor 12, and the support plate (water tank 10) is driven back and forth in one direction as the motor 12 rotates forward and backward. A DC servo motor was used as the motor 12, and the moving position of the water tank 10 was detected based on the encoder pulse input to the control unit 16.

  In this embodiment, in order to make it easy to identify the position (surface shape) of the surface of the water stored in the aquarium 10, about half of the total amount of the aquarium 10 is colored blue, and the remainder is liquid paraffin ( The water tank 10 was filled with water and liquid paraffin, and the water tank 10 was sealed. Since liquid paraffin is lighter than water, when the water tank 10 is stationary, the water is stored in a state where the water surface is horizontal below the water tank 10, and the liquid paraffin is placed on the water. The reason why the liquid paraffin is filled is to make it easy to recognize the shape of the wave so that the wave front does not become entangled when the water tank 10 is swung (reciprocated) to generate a wave.

(Control method: Control based on the concept of polynomial space)
The control unit 16 that drives and controls the motor 12 controls the reciprocating operation of the water tank 10 (support plate) to make the water still (the water surface is horizontal). In this test, the water tank 10 was first reciprocated for a certain time to generate a wave, and then the control was started so that the water in the water tank 10 was stopped.
In this embodiment, the image of the wave in the aquarium 10 acquired by the camera 14 is analyzed, and the movement of the aquarium 10 is controlled by the stabilization control method to stop the water. Since the state of the wave in the aquarium 10 changes every moment, an image of the aquarium 10 is acquired from the camera 14 at a short period (15 fps in the embodiment), and is analyzed for each individual image, and based on the respective analysis results. Then, the movement of the water tank 10 was controlled to make the wave stationary.

FIG. 2 is an example of an image in a state where the water tank 10 is reciprocated in one direction (a direction parallel to the longitudinal direction of the water tank) to generate a wave in the water tank 10. The dark area on the lower side of the aquarium 10 is water.
FIG. 3 shows an approximate curve (broken line) obtained by analyzing this image. The approximate curve is approximated by a linear curve for convenience.
The shape (wave shape) of the water surface in the aquarium 10 is specified as the water surface by converting the image acquired by the camera 14 into RGB and selecting a set of pixels equal to or greater than a specific threshold. When the obtained water surface shape (wave shape) is viewed as a function, the water surface shape can be expressed as a linear combination of bases in a polynomial space defined as a Hilbert space.
The surface of the water in the aquarium 10 actually fluctuates three-dimensionally. However, when a wave is generated by reciprocating the aquarium 10 parallel to the longitudinal direction as in this embodiment, the wave fluctuates two-dimensionally. Even if it sees, a big error does not arise on control.

The orthonormal basis e _m (x) of the polynomial space is expressed by a Legendre polynomial of the following equation (1).
The element f (x) of the polynomial space can be expressed by a linear combination of the basis as in the following equation (2) using the orthonormal basis of the equation (1). That is, the wave shape obtained as an image in the present embodiment can be expressed as a linear combination of the bases.
Since the wave shape changes with time, the wave shape changing with time is expressed by a polynomial f (x, t) (0 ≦ t ≦ ∞) of the following equation (3). That is, the change of the wave from appearing as a time change component a _m (t), it is possible to control the wave by monitoring controls the components a _m (t).

(System identification)
The control method by the control unit 16 in the present embodiment monitors the time change of the component a _m (t) when the observed wave shape is expressed as a linear combination of the basis in the polynomial space, and estimates the subsequent change. Asymptotically stabilized to a target state (state where the water surface is stationary). The order of the observed quantity a _m (t) to be controlled is not limited, but in this embodiment, the observed quantity is a ₂ (k), that is, the component of the observed object, as the first step for verification. The system was identified by using the ARX method, which is an equation error model of the prediction error method. In this model, the input / output response is defined by the discrete-time linear time-invariant system (4).

The following chirp signal u (t) was applied to identify the system. This identification operation means that the water tank 10 is swung (reciprocated) at various frequencies to specify how the water tank 10 (system) responds to an external force. In this embodiment, the order 34 is specified. Until identified.
u (t) ＝ Acos (2πf _c t + παt ² + θ)
A: Amplitude 0.55 (V), f _c : center frequency 0.198 (Hz), α: chirp rate 2.3761 × 10 ^-3 (Hz / s),
θ: Initial phase-(1/2) π (rad)

The system identification conditions are as follows.
Water: 2.52 (l), Water temperature: 22.4 (° C), Liquid paraffin: 5.05 (l), Paraffin temperature: 22.1 (° C), Aquarium size: Width 490 x Depth 180 x Height 300 (mm), Camera mode: 5M Mode, camera height: 130 (mm), camera-water tank distance: 430 (mm)

The identified matrices A, B, C, and D of Equation (4) are shown below.

  FIG. 4 is a graph comparing the output of the identification model with the actual output. FIG. 5 is an enlarged view of the range of 50 to 150 seconds in FIG. It can be seen from the graph that the actual output and the output of the identification model are in good agreement. That is, it was confirmed that the above-described identification model of this system is appropriate.

For comparison, a control test was also performed for the case where the position of the center of gravity, which is a conventional method, is a control target. The center of gravity (x _G , y _G ) is given by the following equation (9).
In Equation (9), W is the sum of all pixel values, n and m are the number of dots in the image, and X _j and Y _j are coordinate values. In FIG. 3, the position of the center of gravity obtained by the equation (9) is indicated by dots.

Even when the position of the center of gravity is to be controlled, it is necessary to identify in advance. The state variable vector having x _G as an element is controlled by the above ARX method, and the discrete-time linear time-invariant system of Equation (4) is Defined and identified. The chirp signal and identification conditions are the same as described above.
FIG. 6 shows the actual output and the output of the identification model. FIG. 7 is an enlarged view of the range of 50 to 150 seconds shown in FIG. The actual output and the identified output substantially match, indicating that the identification model is valid.

(Control test)
In the control test of the present embodiment, the LQG control method was used as a technique for stabilizing the target water surface shape (horizontal plane) with respect to the linear time-invariant system represented by Equation (4). There are various methods for controlling a control target to have a specific behavior, and the method is not limited to the LQG control method.

In the LQG control method, the following equation (10) is set as the evaluation function J, and the design is made such that the state feedback F that minimizes the evaluation function J is added.
In Equation (10), Q _f ≧ 0 and R _f > 0 are optimum feedback weights.
The feedback gain F is the Riccati equation
Is given by the following equation (12).
In this embodiment, the feedback gain is designed as Rf = 5 and Qf = I × 0.001. The feedback gain in this case is shown below.

The conditions of the system used in the control test are the same except that the water temperature is 20.8 (° C.) and the paraffin temperature is 21.0 (° C.) in the above-described identification conditions.
In the test, between 0 and 15.00 (s), the water tank 10 was swung with the input voltage to the motor (DC motor) 12 set to 0.55 sin (2πt × 0.285) (V), and the control was performed after 15.00 (s). Then, the behavior of the water tank 10 thereafter was observed.

  FIG. 8 shows the time variation of the command voltage for the control method according to the present embodiment, the control method for controlling the center of gravity, and the case where no control is performed (no control). It can be seen that the voltage is periodically applied until 15 seconds elapses, and according to the control method of the present embodiment, after 15 seconds elapses, the command voltage gradually decreases and converges. In the control method that controls the center of gravity, the command voltage tends to be suppressed from the initial applied voltage after 15 seconds, but the degree of convergence is inferior to the control method of this embodiment. When control is not performed, the command voltage after 15 seconds is 0 (V).

FIG. 9 shows the temporal change of the component a _{2 in} the polynomial space, in other words, the wave motion (how to shake) in the aquarium 10 for the above three methods. It can be seen that in both the case of the control method according to the present embodiment and the case of the control method in which the center of gravity is controlled, the waves are settled (settling) in a shorter time than in the case of no control.
Figure 10, for 3 methods described above, shows the time variation of the center of gravity x _G. Also in this case, it can be seen that both the control method of the present embodiment and the control method for controlling the center of gravity are shorter in settling time than in the case of no control, and control is performed to prevent sloshing.

FIG. 11 is an example of testing whether or not the anti-sloshing function works when a part of the front surface of the water tank 10 is shielded with paper. The left half of the front surface of the aquarium 10 was hidden with paper, and the control test was performed so that only the movement of the waves in the right half of the aquarium 10 could be detected from the camera 12.
FIG. 12 shows the rotation angle of the motor at that time. When the control method of this embodiment is used, the test results are shown in FIG. 1. After the water tank 10 is swung for 15 seconds and then the control is started, the wave gradually settles and converges. It became out of control.

  This test result is not based on the concept of extracting and controlling the feature amount in the control method of the present embodiment, and therefore the motion of the wave in the entire aquarium 10 is predicted from partial image information. This indicates that control can be performed, and accurate control is possible even when control is not possible with the method of extracting and controlling feature quantities such as the position of the center of gravity.

  In the above embodiment, the control of the anti-sloshing device is taken up as an example of the control without using the concept of the conventional number vector as a method for controlling the behavior without extracting the feature amount from the image information. is there. In the embodiment, in order to facilitate the control, the movement direction of the aquarium is limited to one direction, the wave in the aquarium is set to change in two dimensions, and a polynomial space that is one of the Hilbert spaces is assumed. The possibility of control was verified by The verification result shows that the control method can be effectively used without extracting the feature quantity from the image information.

(Control method: Control based on the concept of matrix space)
In the above-described embodiment, as a control method of the sloshing control device, the shape of the water surface in the water tank is controlled as a linear combination of the basis of the polynomial space. Hereinafter, a method of controlling the linear control theory by a method of reconstructing the matrix on the matrix space will be described.

The matrix space is a Hilbert space, and the orthonormal basis in this space is shown below.
By using the inner product of the matrix X, which is an element in the matrix space, and the orthonormal basis e _{k, l} , the matrix X can be expressed as a linear combination as in the following equation.
In the case of a time-varying matrix X (t), the matrix components are denoted by a _{k, l} (t) and are written as follows:
Since the matrix space is a Hilbert space like the number vector space, the following linear control theory holds.
That is, it is possible to control the image information obtained from the camera as one point in the matrix space and apply linear control theory.

Also in this embodiment, the above-described sloshing device is used. In the experiment, the sampling cycle of image information is set to 15 [fps], and a matrix ((480, 640) type) obtained by coordinate-converting an RGB matrix (image information: 480 × 640 pixels) obtained from the camera 14 into a grayscale space, (6, 6)) type matrix (M (t)) is reduced to the average value, and the component obtained by the inner product of the matrix M (t) and the orthonormal basis of the matrix space is the control variable (a _{k, l} (t)). FIG. 13 is an image diagram showing an example in which image information is grayscaled and reduced to a (6, 6)) type matrix.

In order to control using the concept of matrix space, an identification test of the device was performed.
A chirp signal (input voltage c (t)) is input to the motor 12, and the identification output is a component a _{k, l} obtained by the inner product of the matrix M (t) and the orthonormal basis e _{k, l} (t) in the matrix space. (t). The system was a SIMO system with input 1 and output 36, and was identified by the subspace method. In this embodiment, the n4sid method is used as the subspace method. The chirp signals used are as follows.
c (t) = Acos (2πf _c t + παt ² + θ) [V]
A: Amplitude 0.6 (V), f _c : center frequency 1.28 × 10 ^-2 (Hz), α: chirp rate 9.0 × 10 ^-6 (Hz / s),
θ: Initial phase-(1/2) π (rad)

For each 6 × 6 identification output a _{k, l} (t), the matching rate between the output of the model and the actual output was calculated. The component in the fourth row corresponding to the boundary surface of the water surface of the aquarium 10 The rates were 64.97 (%), 74.75 (%), 62.36 (%), 36.1 (%), 67.98 (%), and 65.35 (%), confirming that the modeling was good.
FIG. 14 shows actual output and model output of output a _4,1 (t) as examples of identification output. FIG. 14B is an enlarged view of the range of 100 to 120 seconds shown in FIG.

(Comparative test)
For the above sloshing control device, a comparative experiment was performed on the control method of the present embodiment and the conventional control method by paying attention to the center of gravity.
In the control method according to the present embodiment, LQG control is performed on the matrix M (t) based on the identification result described above. The regulator weight R _I = 5 × 10 ⁴ I, Q _I = 100, the Kalman filter weight R _g = 1 × 10 ⁸ I, and Q _g = 1 × 10 ⁻⁶ I.
Even when the center of gravity of the conventional method is controlled, an operation for identifying the center of gravity was performed, and LQG control was performed based on the identification result. The regulator weight R _I = 1, Q _I = 1, the Kalman filter weight R _g = 1 × 10 ³ , and Qg = 1.

  In the comparative test, first, a sine wave voltage is input to the motor 12 for 0 to 15 seconds, the water tank 10 is reciprocated in one direction (linearly), and then control is started to shift the water surface to a stationary state. Observed. In the experiment, at the start of control, the hand was moved in front of the camera 14 so as to cross the field of view of the camera 14, and occlusion was performed to block the field of view of the camera 14 for a short time, and the behavior by each control was examined.

FIG. 15 shows the norms of the matrix M (t) and the estimation matrix ^ M (t) in the case of the control method of the present embodiment. The norm of the matrix M (t) fluctuates greatly when 15 seconds have elapsed since the start of reciprocation. This is due to the influence of occlusion. While the norm of the matrix M (t) fluctuates greatly, the norm of the estimation matrix ^ M (t) is hardly affected by occlusion and is settled within a short time after stopping the forced oscillation.
Here, the norm is the difference between the matrix M (t) and the target image, or the difference matrix between the estimation matrix ^ M (t) and the target image, and the square root is obtained by adding the squares of all components. It is. Therefore, the fact that the norm of the estimation matrix ^ M (t) is hardly influenced by occlusion indicates that the occlusion is removed as noise and does not disturb the estimation matrix ^ M (t). At the same time, it can be said that the matrix M (t) or the estimation matrix ^ M (t) settled to the target image when the norm becomes zero.

  FIG. 16 shows the norm of the water surface coordinate y (t) and the estimated coordinate ^ y (t) in the case of control by the conventional method. The LQG control was performed so as to shift from a state where the water surface was swung to a state where the water surface was stationary. However, as shown in FIG. 16, the estimated coordinates ^ y (t) greatly fluctuated due to the influence of occlusion. In other words, the conventional method shows that the estimated coordinates ^ y (t) are disturbed without removing occlusion as noise.

  FIG. 17 shows the input voltage (control voltage) of the motor 12 in the case of control according to this embodiment and the conventional method. The input voltage for 0 to 15 seconds is a preset sine wave voltage, and the control voltage is after 15 seconds. In the case of conventional control, the voltage after 15 seconds has greatly fluctuated due to the effect of occlusion, whereas in the case of the control according to this embodiment, the voltage is not affected by occlusion in a short time. Converges in the vicinity of 0 [V], and the fluctuation of the water surface is settled in a short time.

  FIG. 18 shows the movement position on the support plate of the water tank in the control state according to the present embodiment and the conventional method. When control is started from the state in which the water tank is first reciprocated (periodic reciprocating motion), the water tank moves delicately to keep the water surface stationary. In the control of this embodiment shown in FIG. 18, the water tank stopped moving within a short time after the start of control, and the water surface was in a stationary state. On the other hand, in the case of the conventional method, the water tank has moved greatly left and right (one direction) due to the influence of occlusion.

  The results of the above comparative experiment show that the control method using the concept of the matrix space can be used effectively for the stabilization of the water surface of the aquarium, and in particular, the control method that focuses on features such as the center of gravity has a slight effect of occlusion. However, while it is impossible to perform accurate control, it has been confirmed that the method of controlling without paying attention to the feature amount as in the present embodiment can be accurately controlled without being affected by occlusion.

In addition, in order to compare the control state by the difference in the control method, the experiment which made the water surface still was performed using LQG control and P control (proportional control). In either case, the water tank was first reciprocated for 15 seconds, then occlusion was performed to block the camera's field of view, and the subsequent course was observed.
FIG. 19 shows the motor input voltage in the LQG control and the P control. In the case of P control, the voltage after the start of control fluctuates greatly due to the action of occlusion, whereas in LQG control, it is settled in a short time.
FIG. 20 shows the position of the water tank on the support plate. In the case of the LQG control, the control is set at a fixed position at the start of the control, whereas in the P control, the water tank is not accurately controlled and diverges. This experimental result shows that LQG control is superior to P control, and that LQG control theory in matrix space is useful.

As described above, as described in each embodiment, the method for controlling the feature amount without extracting it based on the image information is not limited to the polynomial space, but a matrix that is a vector space in which the sum of the matrix and the scalar multiplication are defined. This also applies to space. The matrix space is one of the Hilbert spaces similarly to the polynomial space, and can be controlled as one point in the matrix space in the same manner as the method of controlling the image information as one point in the polynomial space. The method of controlling image information by interpreting it as one point in polynomial space or one point in matrix space is to find a specific feature amount such as the centroid position for each object to be controlled, and control the feature amount. Compared to conventional methods that do not become necessary, it is considered that the method is effective in that it can provide a method that can be controlled universally based on image information regardless of the object to be controlled.
Note that the concept of the matrix space is characterized by the fact that the control target is not limited at all and can be used universally for any control target. The feature quantity is not extracted at all, so the image processing / control system design stage is not trial and error. is there. The concept of the polynomial space is characterized in that effective control is possible when the object to be controlled is represented by a curve like the water surface of a water tank and the concept of the polynomial space is easy to apply.

DESCRIPTION OF SYMBOLS 10 Water tank 12 Motor 14 Camera 16 Control part

Claims (5)

A moving object that is a control object;
A drive unit for driving the moving body;
A visual recognition unit for visually recognizing a detection target that fluctuates with the movement of the moving body;
A control unit that controls movement of the moving body via the driving unit based on image information by the visual recognition unit;
The control unit interprets each piece of image information sequentially taken from the visual recognition unit according to a predetermined cycle as one point in the Hilbert space excluding a number vector space, and sets the detection target to a target state by a stabilization control method. An apparatus for controlling an operation of a moving body, characterized by being guided.   2. The moving body motion control apparatus according to claim 1, wherein an LQG control method is used as the stabilization control method. The movable body is a water tank containing water, the drive unit is a drive unit that reciprocates the water tank in one direction, the visual recognition unit is a camera that visually recognizes the water tank, and the control unit is an image acquired by the camera. A sloshing control device for controlling the moving body by the driving unit based on information to set the water in the water tank from a rocking state to a stationary state,
The control unit interprets the image information of the water in the aquarium sequentially taken from the camera according to a predetermined cycle as one point in the Hilbert space excluding the number vector space, and uses a stabilization control method to determine the water in the aquarium. A sloshing control device, characterized in that the drive unit is controlled so as to set the motor from a swinging state to a stationary state.   In the control unit, one point in the polynomial space is extracted from the image information of the water in the aquarium, and the water in the aquarium is set from the swinging state to the stationary state by monitoring and controlling the one point. The sloshing control device according to claim 3 characterized by things. In the control unit, one point in the matrix space is extracted from the image information of the water in the aquarium, and the water in the aquarium is set from the swinging state to the stationary state by a method of monitoring and controlling the one point. The sloshing control device according to claim 3 characterized by things.