JP2001225289A - Robot device and method of controlling its attitude - Google Patents

Abstract

PROBLEM TO BE SOLVED: To permit autonomous attitude reset into a normal attitude state from an abnormal attitude state such as an overturned state. SOLUTION: The transition of the attitude state of a device body into the abnormal attitude state different from the normal attitude state is recognized by a CPU 102 on the basis of acceleration information obtained as the detected output of an acceleration sensor 41. Using track plan data previously prepared and stored in a memory 101 for attitude reset from an overturned state, various drivers 3D-7D are controlled by play-back to perform reset action into the normal attitude state.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a robot apparatus having a function of autonomously returning a posture from an abnormal posture state such as a falling state to a normal posture state and a posture control method thereof.

[0002]

2. Description of the Related Art Hitherto, various types of robot apparatuses having different types of mechanical systems, such as a tire type robot that runs by rotation of a tire and a biped or quadruped autonomous walking robot, have been proposed. This type of robot apparatus has a mechanism system in which an actuator having a predetermined degree of freedom and a sensor for detecting a predetermined physical quantity are arranged at predetermined positions, respectively, and a control unit using a microcomputer outputs and outputs various sensors. By individually controlling the driving of various actuators according to a control program, the actuator can self-run and perform a predetermined operation. Also,
This type of robot apparatus is assembled in a predetermined shape by connecting respective constituent units such as a body, a leg, and a head in a state having a predetermined correlation.

Some multi-legged walking robots having two or more legs are in the form of animals such as cats and dogs. The multi-legged walking robot of such a form has, for example, four legs, and each leg has a predetermined number of joints. Methods for controlling the joints of the feet of this type of robot include a method of recording and reproducing position information and speed information by teaching, and a method of generating and executing position information and speed information by calculation using a motion model. There is.

In the control of the conventional robot apparatus, both the method based on the teaching and the method based on the motion model are premised on the operation under the environment expected by the designer. In some cases, the situation may be contrary to the intention, and the abnormal posture may impair the function or structure of the device, cause a failure, or bring a life to the use environment.

[0005]

SUMMARY OF THE INVENTION Accordingly, an object of the present invention is to prevent the failure or accident of a robot device caused by using the robot device in an abnormal posture such as a fall, in view of the above-mentioned situation of the conventional robot device. It is in. It is another object of the present invention to provide a robot apparatus capable of autonomously returning to a normal posture state from an abnormal posture state such as a falling state, and a posture control method thereof.

[0006]

A robot device according to the present invention is a robot device having at least four legs. In order to solve the above-mentioned problem, the robot device recognizes the posture of the device body and outputs a recognition result. The fall determination means for determining that the posture has become a fall posture of either the posture or the left fall posture, and the fall determination device determines a rightward fall posture or a left fall posture according to the determination result. A control means for performing the fall-back operation by operating the legs, and returning to a normal posture from the fall posture via the prone posture state at least once.

Further, in the robot apparatus according to the present invention, based on the posture recognition step of recognizing the posture of the robot apparatus main body, and the recognition result of the posture recognition step, the robot apparatus main body is at least turned right or left. In the fall judging step of judging that one of the two postures has fallen, and in the fall judging step, if it is judged that the posture is a right-hand fall posture or a left-hand fall posture, a fall-return operation according to the discrimination result is performed by the leg. Done by operating the unit, at least once,
Returning from a falling posture to a normal posture via a prone posture state.

[0008]

Embodiments of the present invention will be described below in detail with reference to the drawings.

The present invention is applied to, for example, a multi-legged walking robot 1 having a configuration as shown in FIG.

The multi-legged robot 1 is an articulated robot, which is in the form of an animal having four legs. The articulated robot 1 has a main body 2, a right forefoot 3, a left forefoot 4, a right It has a hind foot 5, a left hind foot 6, a head 7, a body part 8, a tail 9, and the like.

The articulated robot 1 has joints 10, 11, 1 such as a right front leg 3, a left front leg 4, a right rear leg 5, a left rear leg 6, and the like.
2 and 13 are provided with a brake mechanism 30. By using the operation of the brake mechanism 30, the relative positional relationship of an arbitrary moving part (leg) of each of the right front foot 3, the left front foot 4, the right rear foot 5, and the left rear foot 6 is operated by a direct teaching method. Can teach the position.

The main body 2 is provided with brackets 20, 21, 22, 23 for a right forefoot 3, a left forefoot 4, a right hindfoot 5, and a left hindfoot 6. The head 7 is set at a front part on the main body 2, and the body part 8 is located behind the head 7.
The tail 9 protrudes upward from the body 8.

Each component installed on the main body 2 will be described sequentially.

First, the right forefoot 3 has a leg 3a, a leg 3b, a bracket 20, joints 10, 10a, a brake mechanism 30, servo motors 3c, 3d, 3e and the like.

The upper end of the leg 3a is connected to the bracket 20, and the leg 3a is connected to an arrow R1 about a center axis CL1.
It can rotate in the direction. The leg 3a and the leg 3b are connected by a joint 10. The servo motor 3c is built in the main body 2, and when the servo motor 3c operates,
The bracket 20 can rotate around the central axis CL2 in the direction of arrow R2. When the servomotor 3d operates, the leg 3a can rotate in the direction of the arrow R1 about the central axis CL1. When the servomotor 3e operates, the leg 3b can rotate in the direction of the arrow R3 about the center axis CL3 with respect to the leg 3a.

The left forefoot 4 is composed of legs 4a, 4b,
1. It has joints 11, 11a, a brake mechanism 30, and servomotors 4c, 4d, 4e.

The leg 4a is connected to the bracket 21 so as to be rotatable in the direction of arrow R4 about the center axis CL4. The leg 4 b is connected to the leg 4 by the joint 11.
a. The servo motor 4c is built in the main body 2, and when the servo motor 4c operates,
The bracket 21 rotates around the central axis CL5 in the direction of arrow R5. When the servo motor 4d operates, the legs 4a
Rotates about the center axis CL4 with respect to the bracket 21 in the direction of arrow R4. When the servo motor 4e operates, the leg 4b rotates in the direction of the arrow R6 about the central axis CL6.

Next, the right rear foot 5 has legs 5a, 5b, a bracket 22, joints 12, 12a, a brake mechanism 30, and servomotors 5c, 5d, 5e.

The upper end of the leg 5a is connected to the bracket 22. When the servo motor 5c operates, the bracket 2
2 can rotate in the direction of arrow R7 about the center axis CL7. When the servo motor 5d operates, the legs 5a
Can rotate about the central axis CL8 in the direction of the arrow R8. When the servo motor 5e operates, the leg 5b can rotate in the direction of arrow R9 about the center axis CL9.

The left hind foot 6 includes legs 6a and 6b,
3. It has joints 13, 13a, a brake mechanism 30, and servomotors 6c, 6d, 6e.

When the servo motor 6c operates, the bracket 23 can rotate in the direction of arrow R10 about the central axis CL10. When the servo motor 6d operates, the leg 6a can rotate in the direction of the arrow R11 about the central axis CL11. When the servo motor 6e operates, the leg 6b
It can rotate in the direction of arrow R12 about L12.

As described above, each of the right forefoot 3, the left forefoot 4, the right hindfoot 5, and the left hindfoot 6 are each composed of leg parts having three degrees of freedom, and are driven by the servomotor around a plurality of axes. be able to.

The head 7 includes servo motors 7a, 7b, 7c
When the servo motor 7a operates, it can swing about the central axis CL20 in the direction of the arrow R20. When the servo motor 7b operates, the head 7 moves to the center axis CL21.
Swinging in the direction of the arrow R21 around the center. When the servomotor 7c operates, the head 7 can swing about the central axis CL22 in the direction of arrow R22. That is, the head 7 has three degrees of freedom.

The body 8 has a servo motor 8a. When the servo motor 8a operates, the center axis CL2
The tail 9 swings in the direction of the arrow R23 around the center 3.

Further, as shown in FIG. 2, the articulated robot 1 has a built-in three-axis (x, y, z) acceleration sensor 41 in the main body 2 so that the main body 2 can move to the main body 2 in an arbitrary posture. The acceleration and the angular velocity can be detected. The head 7 has a CCD camera 43 and a microphone 44.
Are arranged. Further, contact sensors 45 are provided on the head, each leg tip, abdomen, throat, buttocks, and tail. As shown in FIG. 3, a detection output from each sensor is output from a CPU provided in the control unit 100 of the articulated robot 1.
(Central processing unit) 102 via a bus 103.

Here, FIG. 3 shows this articulated robot 1.
FIG. 3 shows an example of a connection relationship between the control unit 100 and the respective servomotors and position sensors for driving the respective joint axes of the right forefoot 3, left forefoot 4, right hindfoot 5, left hindfoot 6, head 7, and tail 9; ing.

The control unit 100 includes a memory 101 and a CPU
(Central processing unit) 102, and the bus 103 of the CPU 102 is connected to the above-described elements of the right forefoot 3, left forefoot 4, right hindfoot 5, left hindfoot 6, head 7, and tail 9 described above. ing.

The right forefoot 3 is a servomotor 3c, 3d, 3
e, and position sensors 3P1, 3P2, 3P3. The servo motors 3c, 3d, 3e are connected to a driver 3D, respectively, and the position sensors 3P1, 3P
2, 3P3 are also connected to the driver 3D, respectively.
Each driver 3D is connected to the bus 103.

Similarly, the servo motor 4 of the left forefoot 4
c, 4d, 4e, position sensors 4P1, 4P2, 4P3
Are connected to the driver 4D. Servo motors 5c, 5d, 5e of right hind leg 5 and position sensors 5P1, 5P
2, 5P3 are connected to the driver 5D, respectively. The servo motors 6c, 6d, 6e of the left rear foot 6 and the position sensors 6P1, 6P2, 6P3 are connected to a driver 6D.

Servo motors 7a, 7b, 7c of head 7
And the position sensors 7P1, 7P2, 7P3
D. The servo motor 9a of the tail 9 and the position sensor 9P1 are connected to a driver 9D.

Each position sensor 3P1, 3P2 of the right forefoot 3
3P3, each position sensor 4P1, 4P2, 4P of the left forefoot 4
3, each position sensor 5P1, 5P2, 5P3 of the right hind leg 5 and each position sensor 6P1, 6P2, 6P3 of the left hind leg 6,
Position information at each location is obtained. For example, as these position sensors, a rotation angle sensor such as a potentiometer for detecting a joint angle can be used. Position sensors 3P1-6P such as this rotation angle sensor
When the position information obtained by 3 is fed back to the CPU 102, the CPU 102 gives a command to each driver based on the fed back position information. As a result, the corresponding driver performs servo control on the corresponding motor, and the servo motor rotates to the command position given by the CPU 102.

FIGS. 4 to 7 show the multi-legged walking robot 1 shown in FIG. 1 in a more simplified manner. The torso portion 8 has a head 7, a right front leg 3, a left front leg 4, a right rear leg 5, and a left rear leg 6. Joints 10, 11, and 1 are provided on each of the feet 3 to 6, respectively.
2, 13, 30, 30, 30, 30 are provided, respectively.

The posture of the multi-legged walking robot 1 shown in FIG.
This is a basic posture in which the right forefoot 3, the left forefoot 4, the right hindfoot 5, and the left hindfoot 6 are straightened. FIG. 5 shows a state in which the joints 11 and 30 of the left forefoot 4 have been moved from the basic posture of FIG.

The right front leg 3, the left front leg 4, the right rear leg 5, and the left rear leg 6 of the multi-legged walking robot 1 shown in FIG. In the state of FIG. 5, since the motion is given to the joints 11 and 30 of the left forefoot 4, the left forefoot 4 is in a posture protruding forward.

When the operator operates the multi-legged walking robot 1
To determine the angle between the joint 11 corresponding to the left front elbow of the left front leg 4 and the joint 30 corresponding to the left front shoulder, edit the motion pattern of the multi-legged walking robot as follows. I do.

In the editing operation for giving motion to the joints 11 and 30 to the multi-legged walking robot 1 shown in FIGS. 4 and 5, the external editing instruction computer of the control unit 100 shown in FIG. Figure 400 on 400 software
The position of the center of gravity W0 of the multi-legged walking robot 1 shown in FIG.
It is possible to automatically set the angles of the joints of at least one of the other right forefoot 3, right hindfoot 5, and left hindfoot 6 so that the multi-legged walking robot 1 does not fall from the position of the center of gravity W0. I can do it. By giving this instruction from the external editing instruction computer 400 to the CPU 102 of the control unit, the CPU 102 can give an operation instruction to the servo motor of the corresponding foot.

In this case, the weight of each part of the multi-legged walking robot 1, that is, the weight of the torso portion 8 and the main body 2, the right front leg 3, the left front leg 4, the right rear leg 5, the left rear leg 6, and the head 7 The weight of each unit is set in advance by the external editing instruction computer 4.
00, and the position of the center of gravity W0 of the multi-legged walking robot 1 shown in FIG. 4 can be calculated based on the weight data.

Next, an example of a method for editing a motion pattern of a multi-legged walking robot will be described with reference to FIG.

First, in step S1, information such as the weight and shape of each component of the multi-legged walking robot 1 is stored in the memory 101 of the multi-legged walking robot 1 in advance. That is, information on the weight and shape of each element such as the main body 2, the torso portion 8, the head 7, the right front foot 3, the left front foot 4, the right rear foot 5, the left rear foot 6, the tail 9, and the like is stored in the memory. Then, the information is transferred from the memory 101 to the memory 402 of the external editing instruction computer 400. This is step S
1 to obtain information such as weight and shape.

Next, in step S2, editing of the posture of the multi-legged walking robot 1 is started. That is,
From the basic posture shown in FIG. 4, the posture is made such that the left front leg 4 protrudes forward as shown in FIG. At this time, the movement is taught to the joints 11 and 30. However, if it is left as it is, the center of gravity of the multi-legged walking robot 1 moves to the left front foot 4 side as shown in FIG. You will fall to the left front.

In order to prevent the multi-legged walking robot 1 from overturning, as shown in FIG. 5, when the left forefoot 4 is moved forward to bend the joints 11 and 30, In step S3, the external editing instruction computer 400 of the control unit 100 shown in FIG. 3 changes the center of gravity W0 of the multi-legged walking robot 1 to a new center of gravity W1 Is calculated, and the data is used as the value of the new calculated center of gravity. In order to shift the center of gravity W0 to the new center of gravity W1 in this manner, FIG.
The joints 1 of the right front foot 3, right rear foot 5, and left rear foot 6 as shown in FIG.
The movement is given to 0, 12, 13 and the joints 30, 30, 30. It is the external editing instruction computer 400 that gives this movement.

In this case, in order to ensure the balance of the multi-legged walking robot 1, the joints 10, 12, 13 and the joints 30, 30, 30, 30, 30, of the right front leg 3, the right rear leg 5, and the left rear leg 6, respectively. 3
It is preferable that the motion given to 0 be as in steps S4 and S5. That is, the projection point IM of the new center of gravity W1 of the multi-legged walking robot 1 onto the ground plane 300 is located within the triangular center-of-gravity position proper range AR. This proper range AR is the ground contact point CP1 of the right forefoot 3
And the ground point CP2 of the right hind foot 5 and the ground point CP of the left hind foot 6
3 is a triangular region formed by connecting the three.

Since the projection point IM of the center of gravity W1 is always within the appropriate range AR, the multi-legged walking robot 1 is prevented from falling down, and the right front leg 3, the right rear leg 5, and the left rear leg 6 are prevented from falling. Joints 10, 12, 13 and joints 30, 3
0, 30 movements can be given, and such a stable posture can be selected with the least movement.

As is clear from the comparison between FIG. 5 and FIG. 7, when the operator applies a posture in which the left forefoot 4 protrudes forward to the multi-legged walking robot 1, the center of gravity is automatically shifted from W0. Shifting to W1, the rear side of the multi-legged walking robot 1 is lowered as a whole. Thus, step S3
After calculating the position of the center of gravity in step S4, it is determined whether or not the multi-legged walking robot 1 falls in step S4, and if so, the external editing instruction computer 400 calculates the movement (table of angles) of other joints. After that, the position of the center of gravity is calculated again in step S3.

If it is clear in step S4 that the robot has not fallen, the flow proceeds to step S6, and the external editing instruction computer 400 ends editing of the operation pattern of the multi-legged walking robot 1. When you finish editing in this way,
The external editing instruction computer 400 formally inputs an operation pattern to the CPU 102 of the multi-legged walking robot 1 (step S7).

The multi-legged walking robot 1 has a main body 2
Acceleration information A in each axis (x, y, z) direction detected by a three-axis (x, y, z) acceleration sensor 41 built in
The control unit 100 detects a fall based on ccXt, AccYt, and AccZt. When a fall state is detected, the posture is returned to a normal posture state.

Here, the algorithm of the fall determination by the control unit 100 is shown in the flowchart of FIG.

That is, the control unit 100 detects a fall as follows based on the acceleration information AccXt, AccYt, AccZt in each axis (x, y, z) direction detected by the acceleration sensor 41. .

First, in the fall determination process, first, in step S11, the oldest acceleration information A in the data buffer is stored.
Discard ccXn, AccYn, AccZn and change the time tag of the data in the data buffer. In this multi-legged walking robot 1, the buffer amount of the data buffer is 5
0.

AccXk → AccXk + 1 (k = 0 to n−1) (Equation 1) AccYk → AccYk + 1 (k = 0 to n−1) (Equation 2) AccZk → AccZk + 1 (k = 0 to n−1) (Equation 3) )

In the next step S12, acceleration information AccXt, AccYt, AccZt in each axis (x, y, z) direction measured by the acceleration sensor 41 is stored in the data buffer. The rate of this data update is 10 ms in the case of this multi-legged walking robot 1.

AccXo → AccXt (Equation 4) AccYo → AccYt (Equation 5) AccZo → AccZt (Equation 6)

In the next step S13, time average accelerations AccX, AccY, AccZ in the directions of the axes (x, y, z) are calculated from the data in the data buffer.

AccX = ΣAccXk / n (k = 0-n) (Equation 7) AccY = ΣAccYk / n (k = 0-n) (Equation 8) AccZ = ΣAccZk / n (k = 0-n) (Equation 9) )

In the next step S14, the average acceleration Ac
declination θ between c and YZ plane, YZ of average acceleration Acc
The angle φ between the projection component on the plane and the Z axis is determined (FIG. 10).
(A) and FIG. 10 (B)).

Acc = (AccX ² + AccY ² + AccZ ² ) ^1/2 (Equation 10) θ = asin (AccY / ((AccY ² + AccZ ² ) ^1/2 )) (Equation 11) φ = asin (AccZ / Acc (Equation 12)

In the next step S15, it is determined whether or not the average acceleration (Euclidean distance) Acc is within an allowable error (ΔAcc) range. If the error is outside the error range, it is determined that a large force is received from the outside due to, for example, lifting, and the process exits from the fall determination process.

Acc> 1.0 + ΔAcc [G] or Acc <1.0−ΔAcc [G] → Processing exception (Equation 13)

In the next step S16, the declination θ between the average acceleration Acc and the YZ plane, the angle φ between the projection component of the average acceleration Acc on the YZ plane and the Z axis, and
A template angle θm between the average acceleration Acc and the YZ plane, which is template data in the current posture state, and a template angle φm between the projection component of the average acceleration Acc on the YZ plane and the Z axis. If the position is within the respective tolerances (Δθm, Δφm), the posture is assumed to be normal,
If it is out of the range, it is determined that the vehicle has fallen or has an abnormal posture. During walking, θ = −π / 2, φ = arbitrary.

Θ> θm + Δθm or θ <θm−Δθm (Expression 14) φ> φm + Δφm or φ <φm−Δφm (Expression 15)

Here, the phenomenon of overturning is a phenomenon of a very low frequency with respect to the sampling period of the angular velocity. Therefore, by taking an average of a certain time by using a data buffer as data of overturning detection, instantaneous noise is generated. Accidental discrimination can be reduced. This method has an advantage that the load on the data processing is smaller than that of the low-pass processing using a digital filter or the like.

If a fall is detected by the above-described fall determination process (step S17), the process proceeds to the fall return step S18, and the posture is changed to a normal posture as follows.

That is, in the posture transition processing, first, the declination θ between the average acceleration Acc and the YZ plane calculated at the time of detecting the fall, and the angle φ between the YZ projection component of the average acceleration Acc and the Z axis are represented by: The fall direction is determined. In the multi-legged walking robot 1, when the robot falls during walking, it falls only in the four directions shown in FIGS. 11A, 11B, 11C, and 11D due to the constraint condition due to the shape. , 0 <φ <(1/4) π or − (1/4) π <φ <0 (Equation 16), the forward falling state (Head Side Dow) shown in FIG.
n), and (1/4) π <φ <(3/4) π (Equation 17) is used to determine the right overturning state (Right Sid) shown in FIG.
e Down) is determined, and − (1/4) π>φ> − (3/4) π (Equation 18) is used to determine the left side overturning state (Left Side state) shown in FIG.
Down), and (3/4) π <φ or φ> − (3/4) π (Equation 19), the rear falling state (Tail Side) shown in FIG.
Down) is determined.

Then, the four falling states (HeadSide Down, Right Si
In order to perform the posture return from de Down, Left Side Down, and Tail Side Down), the posture is returned to the normal posture by the playback using the trajectory planning data. Note that there is a case where the fall state changes during the execution of the fall return operation. For example, if you fall forward (Hea
d Side Down) starts the operation for fall return, and if the situation changes to the side fall state, in such a case, the currently executing fall return operation is immediately terminated, and By executing the detected fall return operation, the return operation from the fall state can be quickly executed.

FIG. 13 shows a state in which the vehicle has fallen forward (Head).
The progress of the return operation from Side Down) to the normal posture is schematically shown.

The trajectory planning data for performing the posture return from the above-mentioned forward fall state is based on the relative positional relationship between the right front foot 3, the left front foot 4, the right rear foot 5, and the left rear foot 6 of the articulated robot 1. It can be generated by the operator teaching the position in advance by the direct teaching method described above and stored in the memory 101.

Here, in the description of the articulated robot 1, the control unit 10 is controlled based on acceleration information of a three-axis (x, y, z) acceleration sensor 41 built in the main body 2.
0 to determine the fall state, and the four types of fall state (Hea
d Side Down, Right Side Down, Left Side Down, Tail
The control unit 100 performs a return operation to return to a normal posture state from the (Side Down) position. However, the control unit 100 falls over based on detection outputs of an angular velocity sensor, an angular acceleration sensor, an inclination sensor, and the like built in the main body 2. The determination may be performed, and the returning operation to the normal posture state may be performed. Further, the control unit 100 may perform a fall determination based on image information obtained by the CCD camera 43 or a detection output from the contact sensor 45, and perform an operation of returning to a normal posture state. Further, the control unit 100 may perform a fall determination using composite detection outputs from various sensors and perform an operation of returning to a normal posture state.

In the case of a four-legged robot device, an abnormal posture can be detected by comparing the internal posture model with the output of a contact sensor installed on each of the legs and the main body.

That is, for example, in the standing posture as shown in FIG. 14, only the contact sensors 45A and 45B of the legs among the contact sensors 45A, 45B and 45C detect the contact state. When doing something with the hand, the contact sensor 45B at the tip of the rear leg and the contact sensor 45C installed at the buttocks detect the contact state.
Therefore, the posture executed by the robot device and the ideal state of the contact sensor 45 at that time are stored in the device main body, and the abnormal posture is detected by comparing the output of the contact sensor 45 during the posture execution. be able to.

Further, in a robot apparatus having an image input device, it is possible to detect an abnormal posture by recognizing a road surface and correlating the position with the current intended posture of the device.

That is, in the articulated robot 1 described above,
When the standing posture as shown in FIG. 16 is normal,
As an image output by the CCD camera 43, in a normal posture, an image in which the floor F is horizontal can be obtained as shown in FIG. 17A, whereas in an abnormal posture, an image shown in FIG. As shown in FIG. 17 (C) and FIG. 17 (D), an image in which the floor F is inverted and an image in which the floor F is inclined as shown in FIGS. The abnormal posture state can be detected by determining the state of the floor surface F of the image to be obtained.

To determine the state of the floor F in the above image,
For example, as shown in FIG. 18, the operation of detecting the edge in the Y direction in the coordinate system of the image is repeated, and the line segment is obtained from the coordinates of the plurality of detected positions to obtain the horizontal edge of the floor surface F. Similarly, from the detected position coordinates obtained as a result of the operation of detecting the edge in the X direction, the floor surface F
May be determined, and a line segment of the inclined floor F may be detected by combining them.

Further, in a tire type robot apparatus using a moving mechanism using wheels, since the operating environment is limited to a state in which the wheels are in contact with the road surface, the abnormal posture may be detected as follows.

That is, for example, as shown in FIG. 19, by detecting that the rotation state observed by the rotation detection device RD attached to the non-drive shaft is different from the rotation required for the rotation output device RO, An abnormal posture can be detected.

Alternatively, by installing a floor detecting device FD as shown in FIG. 20, it is possible to detect an abnormal posture when falling down. As the floor detecting device FD, a contact type sensor device such as a non-contact type sensor having a light emitting and a light receiving portion or a micro switch can be used.

Here, when the fall-back operation is performed by the playback method, the operation of returning from the fall state is limited to a specific state transition depending on the shape of the robot apparatus. In the case of a four-legged robot apparatus such as the articulated robot 1, the four types of falling states (Head Side Down, Right S
There are six types of states, including ide Down, Left Side Down, and Tail Side Down), a back side down state with the back on the floor (Back Side Down), and a prone state with the abdomen on the floor (Stomach Side Down). When returning from a fall state, make sure to
mach Side Down). Also, in the case of the back side down state (Back Side Down), as shown in FIG. 21, before reaching the prone state (Stomach Side Down), the four types of the above-mentioned top down state (Head Side Down, Right Side Down) must be performed.
n, Left Side Down, or Tail Side Down). By utilizing this property, when a fall is detected, the robot operation is controlled in such a manner that playback operation data is created by subdividing the return operation of the fall and playing back in accordance with a change in the fall state. With this configuration, when the fall state changes due to unpredictable disturbance, the fall return operation can be switched immediately. Further, in this way, each return operation can be divided and return operation data can be created, which facilitates creation of operation data.

When such a method is not adopted,
For example, when returning operation data is created and reproduced by setting the returning operation from the back side down state (Back Side Down) as one operation. Even when the posture is forcibly changed to a normal state by an external operation, it is not possible to transition to the next operation until the return operation is completed.

2. For example, when the return operation data is created in a state of passing through a left side down state (Left Side Down), other states (for example, a right side down state (Right Side Down state) may be caused by an external factor (for example, a protrusion on the floor surface). Down)), it is not possible to return by the return operation,
You have to do wasted work.

3. In these methods, if the operation being performed is stopped and the operation is restarted when the falling state changes, a large load is applied to the joint due to the occurrence of the discontinuous operation. You will have problems such as.

[0080]

As described above, according to the present invention, the attitude of the apparatus main body is recognized, the recognition result is output, and based on the recognition result, the apparatus main body is at least tilted to the right or left. Is determined to be in any of the falling postures, and when it is determined to be the rightward falling posture or the leftward falling posture, a fall return operation according to the determination result is performed by operating the legs, and at least once. By returning from a falling posture to a normal posture via a prone posture state, for example, autonomously returning to a normal posture state from a rightward falling posture or a leftward falling posture regarded as an abnormal posture state. Can be.

Therefore, according to the present invention, it is possible to provide a robot apparatus capable of autonomously returning from an abnormal posture state such as a falling state to a normal posture state and a posture control method thereof. As a result, since the robot apparatus has a function of autonomously returning from an abnormal posture state to a normal posture state, it is possible to prevent a failure or accident of the robot apparatus due to use in an abnormal posture state such as a falling state. In addition, the use environment can be prevented from being destroyed, and the user can be released from work for returning to the posture.

[Brief description of the drawings]

FIG. 1 is a perspective view schematically showing a structure of a multi-legged walking robot to which the present invention is applied.

FIG. 2 is a perspective view schematically showing an installation state of various sensors such as an acceleration sensor used for detecting a falling state of the multi-legged walking robot.

FIG. 3 is a block diagram schematically illustrating a configuration of a control system of the multi-legged walking robot.

FIG. 4 is a perspective view schematically showing a basic posture of the multi-legged walking robot.

FIG. 5 is a perspective view schematically showing a state in which a left front leg is raised from a basic posture of the multi-legged walking robot.

FIG. 6 is a perspective view schematically showing a state in which the posture of the multi-legged walking robot is collapsed.

FIG. 7 is a perspective view schematically showing a state in which the posture of the multi-legged walking robot does not collapse.

FIG. 8 is a flowchart showing an example of a method of editing a behavior pattern of the multi-legged walking robot.

FIG. 9 is a flowchart illustrating an example of an algorithm of fall determination by a control unit in the multi-legged walking robot.

FIG. 10 shows an average acceleration A obtained in the above-mentioned fall determination process.
declination θ between cc and YZ plane, Y- of average acceleration Acc
FIG. 4 is a diagram schematically illustrating a relationship between a projection component on a Z plane and an angle φ formed with a Z axis.

FIG. 11 is a diagram schematically showing a relationship between an overturning direction and an angle φ at the time of walking determined by a constraint condition based on the shape of the multi-legged walking robot.

FIG. 12 is a side view schematically showing various falling states of the multi-legged walking robot during walking.

FIG. 13 is a side view schematically showing a process of a returning operation from a fall state to a normal posture state of the multi-legged walking robot.

FIG. 14 is a diagram schematically showing a contact detection state by a contact sensor in a standing posture of the multi-legged walking robot.

FIG. 15 is a diagram schematically showing a contact detection state by a contact sensor in a sitting posture of the multi-legged walking robot.

FIG. 16 is a diagram schematically showing a state in which image information is captured by a CCD camera in the standing posture of the multi-legged walking robot.

FIG. 17 is a diagram schematically illustrating image information captured by a CCD camera in a normal posture and an abnormal posture.

FIG. 18 is a diagram for explaining a method of determining the state of the floor based on image information captured by the CCD camera.

FIG. 19 is a schematic perspective view of a tire type robot apparatus provided with a rotation detecting device as an abnormal posture detecting means.

FIG. 20 is a schematic perspective view of a tire-type robot device provided with a floor detecting device as an abnormal posture detecting means.

FIG. 21 is a diagram schematically illustrating a state transition of a return operation from a rear-side fall state.

[Explanation of symbols]

1 robot device, 2 main body, 3 right front leg, 4 left front leg, 5 right rear leg, 6 left rear leg, 7 head, 8 body,
9 tail, 41 acceleration sensor, 43 CCD camera,
44 microphone, 45 contact sensor, 100 control unit, 101 memory

Claims (3)

[Claims] 1. A robot device having at least four legs, a posture recognition unit for recognizing a posture of the device body and outputting a recognition result, and the device body based on the recognition result of the posture recognition unit. When at least one of a rightward falling posture and a leftward falling posture is determined by the fall determining means, and the rightward falling posture or the leftward falling posture is determined by the fall determining unit. Control means for performing a fall return operation according to the determination result by operating the leg portion, and returning to the normal posture from the fall posture via the prone posture state at least once. Robot device. 2. The method according to claim 1, wherein the recognizing means is an acceleration sensor, and the fall determining means determines the fall posture based on the magnitude and direction of the acceleration obtained by the acceleration sensor. Item 4. The robot device according to Item 1. 3. A posture control method for a legged mobile robot device, comprising: a posture recognition step of recognizing a posture of the robot device main body; A fall determination step of determining that the vehicle has fallen to one of a fall posture and a left fall position; and in the fall determination step, when it is determined that the posture is a right fall posture or a left fall posture, A leg-based mobile robot device comprising: performing a corresponding fall-return operation by operating the leg portion, and returning at least once from the fall posture to a normal posture via the prone posture state. Attitude control method.