JP2001091289A - Sensor unit - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a sensor unit capable of precisely sensing an attitude, without depending on the service environment. SOLUTION: This sensor unit is formed by storing in a portable casing a gyro unit 6 for detecting an attitude and a movement, an artificial retina IC 7 for detecting the relative attitude and the movement of an object in a view, acoustic sensors 8 for detecting the presence of a sound source present around the periphery, D-GPS9, and a controller 5 for controlling the operation of these sensors. The controller 5 accepts the results of detection by the sensors 3 to 9 in real time, records these results of detection in updatable manner in an information space formed in a specified memory area and, based on the recorded results of detection, induces the information indicating the events produced around the casing. Thus, the movement simulating human perceptual actions can be detected.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

[0001] The present invention relates to humans, robots,
High-precision detection results by detecting the posture, motion component, position, etc. of the subject such as vehicles and aircraft in a complex manner, and autonomously correcting the instrument error based on other detection data as necessary To a sensor unit that constantly obtains

[0002]

2. Description of the Related Art A technique for detecting a posture and a movement of a subject and linking the detection result and expression data of the video in real time to display a video corresponding to the posture of the subject on a display means has been proposed. is there. In such a technique, when detecting the posture or movement of the subject, conventionally, the following two methods are mainly employed. The first method uses a sensor unit in which a gyro sensor for detecting angular velocity and an accelerometer for detecting acceleration are respectively arranged on a two-dimensional or three-dimensional axis, and the posture and movement of a subject to which the sensor unit is attached. Is a method of detecting The second method is a method in which a three-dimensional weak magnetic field is generated from a magnetizing coil, and a change in the magnetic field is detected by using a magnetic force sensor to detect the posture or the like of the subject. In this method, sensing is performed in directions and places within a range of several meters affected by the magnetic field.

[0003] However, the conventional method has the following problems. The first method has an advantage that the gyro sensor can follow a fast movement, but a drift peculiar to the sensor easily occurs, and high-precision detection of the direction of the subject cannot be performed. In the gyro sensor, the detected angular velocity is integrated and converted into an angle, so that even a small error is accumulated in a power. Such errors are caused by changes in the temperature characteristics of the electric circuit section. This causes a change in the characteristic to be picked up by the temperature of the resistor or the amplifier, which appears as a drift in the detection data. In addition, there is one due to the characteristics of the gyro sensor. Vibratory gyro sensors of this type are mainly used, but a rigid body vibrates during use, so that it takes time to reach stable vibration, which appears as drift. Such a drift is inevitable as long as the gyro sensor is employed. Since the gyro sensor cannot autonomously correct the drift, the gyro sensor needs to correct the drift by using a reference device using an infrared sensor or the like that can detect that the subject is facing front. Since such a reference device is usually installed at a fixed location, if drift correction is performed sequentially, the range of use of the sensor unit will be limited, and it will not be possible to use it outdoors or with a portable device. is there.

[0004] The second method is susceptible to disturbance and requires various filtering processes. In addition to the use of a weak magnetic field, there is a problem that the time required for defining a detection result is long. That is, there is a problem that responsiveness is poor. As a result, the actual movement of the subject cannot be output in real time, but is output after a certain period of time, which is not suitable for applications requiring real-time processing. Responsiveness can be solved to some extent by strengthening the magnetic field and making it more resistant to noise.However, if the subject is a human, the magnetic field may be affected due to concerns about the adverse effect of the magnetic field on the human body, or the effect on the cardiac pacemaker in the medical field. Is required by the market as much as possible, and it can be said that solving this time delay is almost impossible. In the second method, the posture and the like of the subject are measured in a limited place because the magnetizing coil is installed. As in the case of the first method, sensing in a free space is impossible. It is.

SUMMARY OF THE INVENTION In view of the above background, an object of the present invention is to provide a sensor unit that enables high-precision sensing of a posture and the like without depending on a use environment.

[0006]

Means for Solving the Problems The present applicant solves the above-mentioned problems by constructing a sensor unit capable of simulating a human perceptual operation which does not exist conventionally.

-First sensor unit- The first sensor unit includes a first sensor for detecting its own posture and movement, a second sensor for detecting the relative posture of an object in its own field of view and its movement, At least two sensors among the third sensors that detect the presence of a sound source present in the surroundings, and a controller that controls the operation of these sensors are housed in a predetermined housing, and the controller is
The detection results of each sensor are received in real time, and these detection results are renewably recorded in an information space formed on a predetermined memory area. It is configured to derive information representing an event that occurs in.

In a preferred embodiment of the first sensor unit, the controller is configured to derive one common physical quantity common to each sensor by mutually complementing data output from each sensor. When the sensor unit is configured to include the first sensor and the second sensor, the second sensor is configured based on output data from the first sensor.
The output data of the sensor is predicted, and when the prediction result is positive, the error component by the first sensor is autonomously corrected based on the predicted data.

The first sensor includes a plurality of gyro sensors for measuring angular velocities around a multi-dimensional axis formed on the housing, and the second sensor extracts an image contour from an input image and outputs the image outline. When the controller has means for deriving a motion component of the contour by calculation, the controller outputs the first frequency band data out of the output data from the first sensor and the first frequency band out of the output data from the second sensor. It is configured to integrate data of a second frequency band different from the band.

[0010] From the viewpoint of simplifying the configuration, the means for projecting the target and the means for extracting the image contour of the target and deriving the motion component of the image contour by calculation are formed as one chip as the second sensor. The artificial retinal IC that has been used is used.

The third sensor is, for example, an acoustic sensor including one or a plurality of directional microphones for measuring the intensity of a sound signal. It detects the direction of the sound source.

[0012] At least one of a GPS sensor for detecting an absolute position and an azimuth sensor for detecting an absolute azimuth is further housed in the housing to form a sensor unit, and the controller detects information detected by these sensors. In consideration of the above, the current position of the housing may be derived.

-Second sensor unit- The second sensor unit is attached to the subject to be sensed, and includes a first sensor for detecting the posture and movement of the subject, and a second sensor unit viewed from the subject. A second sensor for detecting the relative attitude of the object and its movement, and a controller for controlling the operation of each sensor, the controller accepting the detection results of each sensor in real time, and Information that is renewably recorded in an information space formed on a memory area, complements each other with the recorded detection results, and indicates an event that occurs around the subject based on the complemented detection results. Is derived.

The second sensor unit may be configured to be attached to a subject having main control means for controlling its own operating posture or movement, and to transmit information derived by the controller to the main control means. The subject in this case is, for example, an autonomous walking robot, a flying object, a vehicle, or the like.

In a preferred embodiment of the second sensor unit, the controller constantly predicts a range of data to be detected next based on data output from each sensor in real time, and generates a prediction result. It is configured to generate different information according to the information. For example, when a situation in which the prediction result is negative occurs, a control signal that autonomously ignores the situation or a notification information to the main control unit is generated. Also, at least one of a third sensor for detecting the presence of a sound source around the subject, a fourth sensor for detecting the absolute position of the subject, and a fifth sensor for detecting the orientation of the subject with respect to a reference orientation. Further, the controller is configured to reflect data output from the third, fourth, or fifth sensor in the information representing the event.

[0016]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment in which the present invention is applied to a sensor unit that simulates human perception will be described below. FIG. 1A is an explanatory diagram showing human perception. Human perception is realized by the interaction of a semicircular canal 1 that controls parallel sensation, an eye 2 that controls vision, an ear 3 that controls hearing, and a brain 4 that integrally manages these organs.

The semicircular canal 1 senses its own relative posture, its movement and direction, and its principle is to sense the degree of inclination and direction of the front, rear, left and right of the head from the movement of the water component. . However, with the semicircular canal 1 alone, for example, if you rotate several times in a dark room, you lose your target direction, you can't keep standing on one leg with your eyes closed, you open your eyes, or you stand on both legs. Sometimes. In such a case, the absolute direction and inclination
It is complemented by visual information from the eyes 2 and sound source information from the ears 3.

The eye 2 senses its own posture, its own relative position with respect to surrounding objects, its own movement, its relative speed with its opponent, its distance to the opponent, and the like. It is often experienced that it is difficult to cope with the opponent's fast movement or that the sensing accuracy is low in a dark state. In such a case, the sensed information is complemented by information from the semicircular canal 1 and the ear 3.

The ear 3 senses the presence of the sound source and the distance / direction between the sound source and itself, but the information from the ear 3 alone has low sensing accuracy. This complements the sensing accuracy.

The brain 4 performs mutual complementation of information, confirmation of occurrence of an event, and other judgments based on information from the semicircular canals 1, eyes 2, and ears 3. For example, based on current or past visual information and sound information, the presence or arrangement of things around the user is spatially recognized. People can turn around when they tap on their shoulders because they know that there is nothing behind them. That is, the brain 4 sorts information that always exists spatially by taking into account not only information that is currently recognized but also information that has been acquired in advance. In this way, a space suitable for one's own movement is created, and people can walk with peace of mind.

In order to realize a function equivalent to such a human perception, in the present embodiment, a sensor unit having a configuration as shown in FIG. This sensor unit includes, in a portable housing, a controller 5 corresponding to the brain 4, a gyro unit 6 corresponding to the semicircular canal 1, an artificial retinal IC 7 corresponding to the eye 2, an acoustic sensor 8 corresponding to the ear, and D
-GPS9 (differential-globalpositioning syste
It is an abbreviation for m. ) And these power supplies (not shown).

The gyro unit 6 includes a plurality of gyro sensors and an accelerometer, and detects the posture of the housing and the movement thereof by using these instruments. The gyro sensor is arranged on X, Y, Z axes (pitch, roll, yaw) or X, Y axes formed in the housing, and measures an angular velocity around each axis. The accelerometer is arranged on the X, Y, Z axes or the X, Y axes, and measures the sum of the gravitational acceleration and the motion acceleration due to the inclination of the axis. In the present embodiment, the gyro sensor is used for measuring the posture including the turning angle of the housing,
The accelerometer is used for measuring the acceleration / stop acceleration of the housing, the lateral acceleration due to turning, and the like. Data obtained by this measurement (hereinafter, referred to as “measurement data” in this embodiment) is amplified by an amplifier (not shown) as necessary, and output to the controller 5 in a time series. . The gyro unit may include a compass. In this case, data representing the relative orientation of the housing with respect to the absolute orientation is added to the above measurement data.

In the gyro sensor, a temperature change may affect the measured data, and therefore, when the gyro sensor is used in a cold region or a warm region, it is desirable to add a temperature correcting means for correcting the measured data with a temperature coefficient. . The gyro unit (especially the gyro sensor) 6 can output measurement data in direct response to the movement of the housing, and outputs the measurement data based on the motion component around the axis. Is more accurate for
It is relatively difficult to ensure accuracy when the movement around the axis is small, that is, for movement at low frequencies. Also,
The gyro sensor has an essential problem that a drift component is accumulated in the measurement data as described above.

The artificial retinal IC 7 is a sensor that recognizes the presence of a target around the housing and detects the positional relationship of the recognized target and the movement component of the target or the target itself. As the retinal prosthesis IC7, for example, one disclosed in JP-A-10-340349 can be used. The disclosed retinal prosthesis IC can select a CCD (charge coupled device) function, an image contour extraction function, and the like.
A logic gate array generates a clock for inputting an image, and the result of processing with the content can be stored in an on-chip memory.

In the present embodiment, a target object in its own field of view is projected as a moving image using the artificial retinal IC 7, and the outline information of each image obtained thereby is extracted, and the difference between the extracted outline information is extracted. Then, a physical quantity representing a relative motion component of the target object with respect to itself, specifically, an angular velocity, a speed, a moving direction, a moving amount, a azimuth after the movement, and the like are calculated. For this purpose, first, a motion vector generated between itself and a target is detected. FIG. 2 is a diagram showing a procedure for detecting a motion vector in the retinal prosthesis IC7, and FIG. 3 is a conceptual diagram thereof. In the retinal prosthesis IC7, as shown in FIG. 2, first, an image for one screen is input from a moving image showing a target, and contour information of the image of the target is extracted (steps S101 and S102). FIG. 3A conceptually illustrates this processing. The extracted outline information for one screen is recorded in an in-chip memory (not shown) in a form as shown in FIG. 3B (step S103). After a certain period of time, the same target is projected at the same angle, and the same processing is performed. Here, assuming that the image projected immediately before is the previous image, and the image projected thereafter is the current image, as shown in FIG. 3C, based on the contour information of the previous image and the specific part of the contour information of the current image, A vector component having the quantity and direction as parameters can be extracted by a logical operation. For example, only the moved dot can be extracted by determining the exclusive OR of the contour information (dot) of the previous image and the contour information (dot) of the current image. Artificial retina IC7
Then, this vector component is detected as a motion vector for the image (steps S104 and S105).

In detecting the movement vector, the search speed of the moved dot is determined by predicting the movement speed of the target. In the search range, the search range is first scanned in the horizontal direction, and then the search range is scanned in the vertical direction. Then, the color information of the searched dot is referred to, and if the color information is the same as that of the dot of the specific portion (starting dot), the dot is set as the destination dot. The above is the procedure when the target is moving. However, even when the target itself does not move and the artificial retinal IC 7 is moving, the motion vector can be detected in substantially the same procedure.

The artificial retinal IC 7 is functionally CC
In this embodiment, since it is sufficient to detect the motion vector of the target as described above, the motion component based on the contour information of the image of the target can be easily extracted. In this case, an artificial retinal IC is used. Therefore, it is not denied that a sensor having the same function is used instead.

Although the artificial retinal IC 7 can be used alone, two artificial retinal ICs 7 are provided side by side at a predetermined interval like the human eye, and the movement of the same target and the like are controlled by the two artificial retinal ICs.
A configuration in which detection is performed simultaneously in step 7 is also possible. In this way, the distance from the sensor unit to the target can be calculated from the relative position of the contour information of the target recorded in the in-chip memory of each retinal prosthesis IC7. When mounted on a vehicle, it can be used as an inter-vehicle distance meter.

The retinal prosthesis IC7 calculates the motion component based on the image of the target actually projected, and therefore has high detection accuracy for low-frequency motion, but is binarized because of performing image processing. Since it involves processing and usually involves processing for recognizing a relative change in the position of the target, it is difficult to maintain high accuracy with respect to high-frequency movement of the target.

The acoustic sensor 8 is a sensor that measures sound signals from the same sound source from a plurality of directions, and can specify the direction of the sound source based on the arrival time difference of the measured sound signals. The specific configuration of the acoustic sensor 8 includes a directional microphone for detecting a sound signal from a sound source and a specific frequency, for example, 50 to 100 Hz, based on a sound signal received by the directional microphone.
And processing means for performing an arithmetic operation for specifying the sound source direction from the arrival time difference between the extracted sound signals. Like a human ear, two directional microphones may be arranged at regular intervals so that the sound source direction can be specified from the arrival time difference of sound signals input simultaneously to the two directional microphones. In the processing means,
The intensity change pattern of the sound in the low frequency region is cut by a band-pass filter, and the cut pattern and the direction of the sound source at that time are stored, and when a similar pattern occurs, the angle of the direction is referred to. It is also possible to increase the processing efficiency by setting the value to a value.

D-GPS 9 is a kind of GPS,
A system in which an S signal is received by a GPEX (Satellite Positioning Information Center) fixed reference station, an error is calculated, and correction data is transmitted by FM multiplex broadcasting. Output. It is well known that this D-GPS 9 can obtain a detection accuracy 10 times or more higher than that of a general GPS.

The controller 5 is formed by a CPU reading and executing a predetermined program recorded in, for example, a ROM or the like, and has a main function of controlling the operation of each of the sensors 6 to 9. A function of integrating the measurement data from each sensor 6 to 9 into an information space formed in an internal memory such as a RAM. The advantages of each sensor 6 to 9 are extracted based on the integrated measurement data. The function to correct with the sensor's measurement data, predicts the next measurement data based on the measurement data that is sensed and integrated every moment, and makes a reasonable (within a certain tolerance) It has a function of recognizing the movement of itself (sensor unit) and the movement of a target. In the information space of the internal memory, the measurement results of each of the sensors 6 to 9 over 360 degrees, such as -180 to 0 to +180, are recorded so as to be updated at any time together with time information. It enables integration of spatial and temporal data, and data prediction and other judgment processing using the integrated data.

Next, the operation of the sensor unit according to the present embodiment will be described. In this sensor unit, a semicircular canal operation simulating a human semicircular canal, a visual operation simulating vision, and an auditory operation simulating hearing are performed. <Semicircular canal operation> In the semicircular canal operation of the sensor unit, the posture (inclination, direction, and the like) of the subject such as the housing and the housing attachment and the movement thereof are recognized through the controller 5 mainly based on the measurement data of the gyro unit 6. This is achieved by: However, with the gyro unit 6 alone, it may be difficult to determine the correct posture of the subject when the subject to which the sensor unit is attached rotates several times, for example, after the power is turned off and then restarted. .
Assuming such a case, in the present embodiment, the data for simulating the semicircular canal operation is based on the artificial retinal IC7.
The visual information thus obtained, the information on the physical quantity calculated based on the measurement data from the artificial retinal IC 7, the information detected by the acoustic sensor 8, and the information on the physical quantity based on the information are complemented. Further, depending on the degree of movement (detection frequency), measurement data from the artificial retinal IC 7 is used instead of measurement data from the gyro unit 6.

<Visual operation> Visual operation in the sensor unit is mainly based on the measurement result of the retinal prosthesis IC7, the existence of the target, the relative movement of the subject itself or the target, the posture, position, and orientation of the subject. Is realized through the controller 5. However, artificial retinal IC
With only 7, it is difficult to cope with a fast movement of the subject or the target, or in a dark environment such as at night, the absolute position or posture of the target or the subject cannot be correctly recognized. Assuming such a case, in the present embodiment, in the case of a fast movement, the detection result by the artificial retinal IC 7 is complemented with information on the posture and orientation based on the measurement data of the gyro unit 6,
As for the position and orientation, information detected by the acoustic sensor 8 and information on physical quantities calculated based on the information, D-
Complement with the position information detected by GPS9.

<Hearing Operation> The hearing operation in the sensor unit is realized by recognizing the direction of the subject and the distance to the sound source through the controller 5 mainly based on the measurement result of the acoustic sensor 8. However, if only the information from the acoustic sensor 8 is used, if the intensity of the sound signal is too weak or the angle in two directions with respect to the sound source is small, it may not be possible to ensure sufficient accuracy. On the other hand, when detecting the direction of the sound source, the time required for searching for the sound source direction can be reduced by predicting the direction of the sound source based on the image actually projected by the artificial retinal IC 7. Therefore, in the present embodiment, the information detected by the artificial retinal IC 7 is used for complementing the auditory operation.

<Integration processing by controller> In the above-mentioned three semicircular canal operations, visual operations, and auditory operations, the information collected by the controller 5 of the sensor unit from time to time from each of the sensors 6 to 9 is spatially and temporally updated in a memory area. It is realized by freely and integrally recording the information, and complementing and predicting the information based on the information. In addition to supporting the above operations, the controller performs recognition of an event that has occurred around the subject and other various determination processes. Hereinafter, the contents of each process will be described in detail.

When the movement of the target is detected (calculated) through image processing in the artificial retinal IC 7, it is determined whether the movement of the image of the input target is the result of the movement of itself or the result of the movement of the target. There is a need to. The controller 5 uses the measurement data of the gyro unit 6 for the determination in this case. FIG. 4A is a video example in the case where the retinal prosthesis IC7 itself is advancing toward the target (illustrated by an ellipse) 41, and FIG. Is an example of the case in which it moves to the left from the center of the figure. FIG.
In (a), the target 41 increases as the artificial retinal IC 7 advances, but the angular velocity of the gyro unit 6 does not change. In this case, it can be recognized that the image indicates that the artificial retinal IC 7 is moving. On the other hand, in FIG. 4B, the target 41 moves to the left with the same size, but the angular velocity of the gyro unit 6 does not change. In such a case, it can be recognized that the target 41 is moving.

Also, when the retinal prosthesis IC 7 and the target move together, the actual motion component can be corrected to be accurate using the detection result of the gyro unit 6. For example, as shown in FIG. 5A, when it is known that the retinal prosthesis IC7 is fixed, if there is a motion component, it is a result of the movement of the target, and only the motion vector 51 of the set portion in the target is obtained. Can be used to calculate a physical quantity related to the movement of the target. However, as shown in FIG. 5B, when both the target and the artificial retinal IC 7 are moving (movement in the same direction in the illustrated case), in addition to the motion vector 52 for the target, the movement by the gyro unit is performed. (Vector amount) 53 needs to be considered. FIG. 5C shows a difference 54 between the two vectors, and this difference vector represents an actual motion vector. By correcting the motion vector in this way, it is possible to improve the accuracy of detecting the motion.

As described above, the gyro unit 6 has high accuracy for high-frequency movement, while the retinal prosthesis IC7 has high accuracy for low-frequency movement. Focusing on this point, the controller 5 compares the data of the first frequency band in the output data from the gyro sensor 6 with the data of the second frequency band different from the first frequency band in the output data from the artificial retinal IC 7. To obtain integrated measurement data covering both frequency ranges. Specifically, the high-pass filter 6a removes low frequency components from the detection signal of the gyro unit 6, and outputs a signal of about 60 Hz to 0.5 Hz. Conversely, the detection signal of the artificial retinal IC 7 is a low-pass filter. At 7a, a high frequency component is removed, and a signal of 0.5 Hz or less is output. Then, both signals are integrated. Further, for example, if sensing is performed at 60 Hz, the angular velocity can be calculated at 60 Hz with the measurement data of the gyro sensor 6, but the artificial retinal IC 7 calculates at a maximum of 20 Hz. This is artificial retinal IC7
This is because the angular velocity is calculated based on the difference between three points (triangle points) that are temporally separated. That is, the speed is calculated from the difference between the first point, which is one vertex of the triangle, and the second point, which is another vertex, and the first, second,
This is because the angular velocity is calculated based on the difference between the third points. In the case of the gyro unit 6, the angular velocity can be calculated at each point. The measurement data from the two types of sensors 6 and 7 having different frequency characteristics as described above is
Mixing is performed through filters that match the respective frequency characteristics. In the above example, one of the three points is measured once by the artificial retinal IC 7, and the gyro sensor 6
And mix the two. Thereby, the advantage of the gyro unit 6 and the advantage of the artificial retinal IC 7 can be combined to obtain a detection result based on measurement data in a highly accurate region.

The measurement data of the gyro unit 6 and the calculation result of the artificial retinal IC sensor 7 can be referred to each other to improve the detection accuracy of the posture and the like. FIG.
FIG. 4 is an explanatory diagram of a processing procedure of the controller 5 in this case. The controller 5 integrates (integrates) the measurement data (data representing the angular velocity) by the gyro unit 6 for the time required for the calculation by the artificial retinal IC 7 (step S2).
01), the angular velocity calculated by the retinal prosthesis IC7 is predicted (step S202). The angular velocity calculated by the artificial retinal IC 7 based on the measurement data of the gyro unit 6 is predicted for the following reason. The angular velocity obtained by the gyro unit 6 can be obtained by integrating instantaneous measurement data, whereas in the case of the artificial retinal IC 7, it can be obtained directly without requiring integration. However, the artificial retinal IC 7 requires a certain amount of time before the calculation because it involves the processing for binarizing the image as described above. If both operations are normal, the result is almost the same and there is no problem.
If there is a processing error or the like, an error occurs in the calculation result. However, this can only be understood by actually calculating. Therefore, first, the measurement data of the gyro unit 6 is relied upon, and after the calculation in the artificial retinal IC 7, the presence or absence of an abnormality is determined.

It is determined whether the angular velocity calculated by the artificial retinal IC 7 is out of a predetermined range, that is, exceeds a value predicted by the gyro sensor 6, and if it is, the angular velocity by the gyro unit 6 is adopted (step). S2
03, S204: No, S205), step S201
Return to the processing of. Such a situation mainly occurs when the motion is of a high frequency. On the other hand, if both calculation results are within the predetermined range, the angular velocity calculated by the artificial retinal IC 7 having relatively high accuracy is adopted (steps S203, S204: Yes, S206). Thereafter, a difference between the adopted angular velocity and the angular velocity measured by the gyro unit 6 is obtained. Since no drift component occurs in the artificial retinal IC7, when there is a difference between the two,
The difference is usually a drift component of the gyro sensor.
Therefore, the difference is feedback corrected (step S208).

By mutually confirming the measurement data of the gyro sensor 6 and the measurement data of the artificial retinal IC 7,
The drift component of the gyro sensor can be corrected. Here, the retinal prosthesis I is generated every unit time (for example, one second).
It is assumed that measurement data representing a motion component obtained by the C7 and the gyro unit 6 is input to the controller 5. In the controller 5, when no motion component is detected by the retinal prosthesis IC7 and the angular velocity of the gyro unit 6 is 0, there is no error in the gyro unit 6, so that both values are set to the initial values. . On the other hand, if no motion component is detected in the artificial retinal IC 7 but the angular velocity of the gyro unit 6 is a finite value, the angular velocity obtained by the gyro unit 6 is a drift component. To be corrected. in this way,
The disadvantages of the gyro unit 6 and the artificial retinal IC 7 can be covered by each other's advantages, and highly accurate posture detection and the like can be performed. Further, it becomes possible to configure a sensor unit capable of autonomous correction of a drift component in the related art.

By the above-described auditory operation, for example, the amount of deviation from the front of the subject can be detected. Therefore, the frontal direction of the subject detected (calculated) by the artificial retinal IC 7 or the like can be complemented by the auditory operation. As an application of the direction detection of the sound source, the moving amount of the sound source can be calculated by detecting the moving amount of the sound source per unit time by the acoustic sensor 8, and furthermore, the moving speed of the sound source can be calculated while the sound source is fixed. By detecting the component, the fact of the movement of the sensor unit, the amount of movement, the direction of movement, and the like can be detected. As described above, it is difficult for the retinal prosthesis IC7 to correctly detect the movement of the target at night or the like.
The moving component of the sound source is detected by the acoustic sensor 8 in synchronization with the measurement by the artificial retinal IC 7, and the detection result is used as the artificial retinal IC.
7, the retinal prosthesis IC7
Detection accuracy can be improved.

A common physical quantity can be detected based on data from the gyro unit 6, the artificial retinal IC 7, the acoustic sensor 8, and the D-GPS 9. For example, the movement speed and movement amount of the subject can be detected from the accelerometer included in the gyro unit 6, and the movement amount and movement speed of the subject can be detected by integrating the movement components of the retinal prosthesis IC7. Further, as described above, the moving speed and the moving amount of the sound source or the subject can also be detected by the acoustic sensor 8. The speed can also be calculated by the D-GPS9. Therefore, a usage form in which these common physical quantities are compared with those of other sensors, and each of them is confirmed to be within a certain allowable range, and then output to the outside as a detection result is also possible. In addition, a reference sensor is determined in advance for each physical quantity, and a detection result from the reference sensor is used preferentially, or a deviation from a sensor having a higher priority is detected to determine the operation of another sensor. A form in which a defect is detected is also possible. In this regard, the same applies to data representing the posture of the subject.

Further, as described above, based on the common physical quantity, it is possible to predict a defect in the measurement result due to the detection result by another sensor. For example, in D-GPS 9, a position or the like is detected by satellite information, but a radio wave generates a multipath (multiple paths) due to reflection of a building or the like, and a "jump" occurs in the position information. Specifically, suddenly 200
The position flies about m. However, if the acceleration data included in the gyro sensor 6 cannot move 200 m per second compared to the expected speed (if the prediction result is negative), it is assumed that it is a clear occurrence of multipath. , D-GPS9 is ignored.

As described above, according to the sensor unit of the present embodiment, sensing simulating human perception is performed, the advantages of the sensors 3 to 9 are integrated, and the disadvantages are compensated for by other sensors. As a result, high-precision sensing that has not been achieved in the past becomes possible. In addition, since the sensors 3 to 9 and the controller 5 are housed in a portable housing, sensing independent of the use environment can be performed.

[0047]

Next, the application scene of the sensor unit of the present embodiment will be specifically described. <Autonomous walking robot> The sensor unit of the present invention can be applied as a posture sensor of an autonomous walking type biped robot. FIG. 7 is a diagram showing a state where the sensor unit V of the present invention is arranged on the head of the robot W. The sensor unit V functions as the eyes, ears, and semicircular canals of the robot W. That is, similarly to the human perception, the position, the facing direction, the posture, and the movement of the robot W itself are detected. It also outputs the distance to the target and the relative movement.

The robot W is provided with servo motors at joints for performing a walking motion similar to that of a human, and each servo motor is controlled by the main controller U.

The main controller U detects the current robot W and the like with the sensor unit V based on the detection information from the sensor unit V, and performs the optimal control of the corresponding servomotor based on the detected information. An amount (a control amount for realizing a target posture and a movement mode) is calculated. For example, if the robot W is tilted by 5 degrees from the horizontal, a calculation is performed to level the robot at an optimum speed, and a control amount according to a predetermined modeling curve is calculated. The modeling curve is
For example, when the inclination of the robot W is returned to a horizontal state after 3 seconds, a curve for determining a control timing such as whether to make the angular velocity linear for that, or to increase the angular velocity at first and then slowly is set. Say. This modeling curve shows the torque of each servomotor and the robot W
Depends on the location of the control.

After calculating the optimal control amount, the modeling curve is divided at a predetermined control timing (for example, 10 Hz), and the driving power is output to the corresponding servo motor. The procedure of detecting the amount of movement of the robot W by the sensor unit V based on the output of the drive power and confirming the next control amount of the servo motor is repeated. In this way, autonomous walking is realized while stabilizing the center of gravity of the entire robot.

The controller of the sensor unit V predicts the measurement result at the next control timing and transmits the result to the control device of the robot W. The main controller U can also predict the expected result from the previously determined optimum control amount, so that the normal feedback control is performed if the difference between the two is within the allowable range. On the other hand, if the difference between the two exceeds the allowable range, the robot W and the sensor unit W
Is determined to be abnormal, control is stopped. Thus, by applying the sensor unit of the present invention, an autonomous walking type biped robot can be easily realized.

<Automatic guided vehicle> The sensor unit of the present invention is an automatic guided vehicle that moves autonomously with a main controller U similar to a bipedal robot, such as an object transporting machine in a severe environment or time, or a cleaning vehicle. It can also be used as a sensory sensor. The use form in this case, for example, the control form by the controller is the same as that of the biped robot.

<Flying Object> The sensor unit of the present invention is mounted on a flying object attitude detection sensor, for example, a helicopter operated by a radio control, and can be used as a sensor for detecting the altitude, attitude, and the like of the helicopter. . For example, a figure of a mark in a circle shown on the right side of FIGS. 8A to 8D is described at a landing point, and a change in the figure is detected to recognize the altitude and attitude of the helicopter. Can be. In this case, the motion vector representing the movement of the helicopter is due to the movement of the helicopter itself, and the figure at the landing point does not move, so that the detection result can be obtained only by the vector calculation.

When the helicopter is equipped with a device such as the main control unit U of a bipedal robot, it is possible to autonomously correct its own posture in response to an external instruction. Further, in the helicopter, various disturbances such as a change in own weight due to wind and fuel consumption occur, and these disturbances can be detected by the sensor unit of the present invention, and the results can be reflected in flight control. .

<Vehicle> The sensor unit of the present invention can be mounted on a vehicle and used as various sensors. For example, the prediction operation by the controller 5 can be used to recognize obstacles and dangerous objects during driving and to avoid danger such as jumping out of a child. That is, the predicted detection result is always compared with the actual detection result, and if the actual detection result at a certain time during driving does not match the predicted detection result (if negative), the vehicle It can be recognized that an obstacle exists in the vicinity of.
Also, the speed of the self detected by the D-GPS 9 is 4 km / km.
At time h, if the speed detected by the artificial retinal IC 7 is the speed immediately after the start, the detection result of the D-GPS 9 is recognized as the initial speed of the vehicle. If the angular velocity is recognized and the angular velocity is not detected by the gyro unit 6, it is recognized that an unexpected operation of the object, that is, an object that suddenly jumps out, is provided, and it is possible to prompt the stopping operation of the vehicle.

The sensor unit can also be used as an inter-vehicle distance meter. That is, the viewing angle to the preceding vehicle is obtained by the artificial retinal IC 7 and the distance to the preceding vehicle can be calculated from this viewing angle. The sensor unit can be used as a speedometer by displaying an image of a center line on which an intermittent pattern continues on a road surface. That is, the viewing angle from the artificial retinal IC 7 to the center line is obtained, and the change in the rhythmic pattern of the center line is detected as a motion vector by the artificial retinal IC 7 to calculate the speed of the vehicle equipped with the sensor unit. be able to.

[0057]

As is clear from the above description, according to the sensor unit of the present invention, the detection results of a plurality of sensors are complemented to utilize the advantages of the sensors, and the disadvantages are compensated for by the advantages of other sensors. As a result, the detection accuracy can be maintained at a high level. Further, when a gyro sensor is included, the drift component is autonomously corrected based on the detection result of another sensor, so that sensing regardless of the use environment becomes possible. Furthermore, since the detection result based not only on the data sensed from time to time but also on the spatially and temporally integrated data is obtained, the detection result close to the human perceptual operation can be obtained accurately. It can be widely applied to applications that were not possible with sensors or sensor units.

[Brief description of the drawings]

FIG. 1A is an explanatory diagram showing a human perception operation,
(B) is a block diagram of the sensor unit of the present invention.

FIG. 2 is a diagram showing a procedure of extracting a motion vector in an artificial retinal IC.

FIGS. 3A to 3C are diagrams showing an outline of extraction of a motion vector in an artificial retinal IC.

FIGS. 4A and 4B are explanatory diagrams for distinguishing subjects of movement in the artificial retinal IC. FIG. 4A is an image example when the artificial retinal IC itself is moving toward a target, and FIG. This is an example in which the retinal IC does not move and the target moves from the center of the figure to the left.

FIGS. 5A and 5B are explanatory diagrams showing an outline of correction of movement in the retinal prosthesis. FIG. 5A shows a case where only a target moves, and FIG. , (C) is a diagram showing the motion amount after correction.

FIG. 6 is a diagram showing a processing procedure of mutual complementation between the gyro unit and the artificial retinal IC sensor.

FIG. 7 is a conceptual diagram when the present invention is applied to an autonomous walking robot.

FIGS. 8A to 8D are diagrams showing an outline of processing when the present invention is applied to a helicopter helicopter.

[Explanation of symbols]

 5 Controller 6 Gyro unit 7 Artificial retina IC sensor 8 Acoustic sensor 9D-GPS

──────────────────────────────────────────────────続 き Continued on the front page (51) Int.Cl. ⁷ Identification symbol FI Theme coat ゛ (Reference) G06T 1/00 G06F 15/62 380 5L096 7/20 15/70 410 F term (Reference) 2F029 AA02 AA05 AA07 AA08 AB03 AB05 AB07 AC01 AC04 AC09 AC12 AC16 AC20 2F065 AA04 AA22 AA45 DD03 JJ03 JJ16 JJ23 JJ26 QQ04 QQ14 QQ24 QQ32 2F105 AA02 AA03 AA06 AA10 BB17 BB20 5B057 AA05 BA02 BA04 DA07 DB02 DC06 DC07 DC02 DC07 DC02

Claims (12)

[Claims] 1. A first sensor for detecting a posture and a movement of a self, a second sensor for detecting a relative posture and a movement of an object in a field of view of a self, and a third sensor for detecting the presence of a sound source existing around the self. At least two sensors,
A controller for controlling the operation of these sensors is housed in a predetermined casing, and the controller receives detection results of each sensor in real time, and forms these detection results on a predetermined memory area. A sensor unit configured to be renewably recorded in an information space, and to derive information representing an event occurring around the housing based on the recorded detection result. 2. The controller according to claim 1, wherein the controller is configured to derive one physical quantity common to each sensor by mutually complementing data output from each sensor. Sensor unit. 3. It has the first sensor and the second sensor,
The controller predicts output data of the second sensor based on output data from the first sensor, and when the prediction result is positive, autonomously calculates an error component by the first sensor based on the predicted data. The sensor unit according to claim 1, wherein the correction is performed dynamically. 4. It has the first sensor and the second sensor,
The first sensor includes a plurality of gyro sensors for measuring angular velocities around a multi-dimensional axis formed on the housing, and the second sensor extracts an image outline from an input image and moves the image outline. Means for deriving a component by an operation, wherein the controller is different from the first frequency band data of the output data from the first sensor and the first frequency band of the output data from the second sensor. The sensor unit according to claim 1, wherein the sensor unit is configured to integrate data with a second frequency band. 5. An artificial retinal IC in which the second sensor is a one-chip integrated means for projecting a target, and means for extracting an image contour of the target and calculating a motion component of the image contour by calculation. The sensor unit according to claim 1. 6. The third sensor is an acoustic sensor including one or a plurality of directional microphones for measuring the intensity of a sound signal, and measures the intensity of a sound signal from the same sound source from different angles to thereby determine its own The sensor unit according to claim 1, wherein the direction of the sound source is detected. 7. A housing further comprising at least one of a GPS sensor for detecting an absolute position and an azimuth sensor for detecting an absolute azimuth, wherein the controller considers information detected by these sensors. The sensor unit according to claim 1, wherein the sensor unit is configured to derive a current position of the housing. 8. The sensor unit according to claim 1, wherein the housing is portable, and each of the sensors and the controller is driven by a portable power supply. 9. A first sensor attached to a subject to be sensed and detecting a posture and a movement of the subject, and a second sensor detecting a relative posture and a movement of the object viewed from the subject. A controller that controls the operation of each sensor, the controller accepts the detection results of each sensor in real time, and updates these detection results in an information space formed on a predetermined memory area. A sensor unit configured to record, mutually complement the recorded detection results, and derive information representing an event occurring around the subject based on the detection results after the complementation. 10. The apparatus according to claim 9, wherein the subject has main control means for controlling its own operating posture or movement, and is configured to transmit information derived by the controller to the main control means. The sensor unit as described. 11. The controller according to claim 1, wherein the controller constantly predicts a range of data to be detected next based on data output from each sensor in real time, and generates different information according to a prediction result. The sensor unit according to claim 9. 12. A third sensor for detecting the presence of a sound source around the subject, and a fourth sensor for detecting an absolute position of the subject.
A sensor, at least one of a fifth sensor for detecting an orientation of the subject with respect to a reference orientation, wherein the controller represents data output from a third, fourth, or fifth sensor to indicate the event The sensor unit according to claim 9, wherein the sensor unit is configured to be reflected on information.