JPH09272083A - Two-foot walking robot - Google Patents

Abstract

PROBLEM TO BE SOLVED: To automatically correct an attitude detection error generated in an inertia sensor used in attitude detection, by estimating an attitude detection error due to an attitude detection part to correct its attitude value from movement of a robot obtained from the attitude detection part and an articulate angle detection part. SOLUTION: In an attitude detection part 22, on the basis of a rate gyro 21, an attitude of the body is detected, in an articulate angle detection part 23, each articulate angle of a leg part is detected, in a leg angle detection part 24, a front/rear angle of the leg part is detected. In an attitude detection error estimating part 25, from movement of a robot obtained from the attitude detection part 22 and the articulate angle detection part 23, an attitude detection error by the attitude detection part 22 is estimated. Further, in a coefficient multiplying part 26, a value, multiplying the estimated attitude detection error in the attitude detection error estimating part 25 by a prescribed coefficient, is added to an attitude detection value of the attitude detection part 22, to be corrected, an output of the coefficient multiplying part 26 is supplied to the attitude detection part 22, by a sampler 27.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to control of a bipedal walking robot using an inertial sensor such as a rate gyro.

[0002]

2. Description of the Related Art FIG. 8 is a schematic view of the structure of a conventional bipedal walking robot. In this figure, 1 is the right leg of the robot,
2 is a motor RR with a built-in encoder, 6 is a motor LR with a built-in encoder, 7 is a left foot of the robot, 9 is a motor P with a built-in encoder, 10 is a right ankle angle sensor, and 11 is a left ankle angle sensor.

This bipedal walking robot achieves complete dynamic walking without the need for ankle torque. However, in the conventional bipedal walking robot, as shown by reference numerals 10 and 11 in FIG. 8, an ankle angle sensor (front and rear, left and right) is required for posture detection, and the angle and each motor 2, 6, 9 are used. The posture of the robot was detected by measuring the rotation angle of the robot (Reference: Proceedings of the Japan Society of Mechanical Engineers (C), Volume 433, No. 433 (Sho 57-9), pp. 1445 to 1455. Dynamic Walking of Walking Robot Author: Isao Shimoyama).

[0004]

As described above, the conventional 2
Since the legged walking robot detects the posture using an ankle angle sensor, there is a problem that if the angle of the ground is unknown or the ground is uneven, an error will occur in the detection of the posture and it will be impossible to walk. there were.

The present invention has been made to solve the above problems and uses an inertial sensor for posture detection.
However, when the inertial sensor is used, due to the nature of the inertial sensor, an error occurs and accumulates due to impact and vibration at the time of landing, and there is a problem that walking is impossible unless the error is corrected.

Therefore, an object of the present invention is to obtain a bipedal walking robot which uses an inertial sensor for posture detection and which can automatically correct a posture detection error caused by the inertial sensor.

[0007]

2 according to the invention of claim 1
The legged robot includes a posture detection unit that detects the posture of the body of the bipedal robot using an inertial sensor, a joint angle detection unit that detects the joint angle of each joint of the leg of the bipedal robot, and A posture detection unit and a posture detection error estimation unit that estimates a posture detection error by the posture detection unit from the movement of the robot obtained from the joint angle detection unit, and the posture is calculated using the estimated value of the posture detection error estimation unit. It is configured to correct the attitude value of the detector.

In the bipedal walking robot according to a second aspect of the present invention, the posture detection error estimator estimates the posture value detected by the posture detector and the joint angle of each joint detected by the joint angle detector. The posture of the body is estimated from the above, and the difference between the estimated posture value thus obtained and the posture value detected by the posture detection unit is obtained as a posture detection error.

In the bipedal walking robot according to a third aspect of the present invention, a value obtained by multiplying the posture detection error estimated by the posture detection error estimation unit by a predetermined coefficient is added to the posture detection value of the posture detection unit and corrected. A coefficient multiplication unit for performing the above is further provided.

A bipedal walking robot according to a fourth aspect of the present invention comprises a posture detecting section for detecting the posture of the body using an inertial sensor, and a walking speed commanding section for generating a walking speed command value for the legs. A walking speed calculation unit that detects the actual walking speed of the leg, and a posture value of the posture detection unit is corrected based on the walking speed command value and the actual walking speed detected by the walking speed calculation unit. And a correction value calculation unit for

In the bipedal walking robot according to the invention of claim 5, the correction value calculation unit subtracts the output value of the walking speed command unit from the output value of the walking speed calculation unit, and is obtained in this way. This value is multiplied by a predetermined constant and the time required for one step to obtain a correction value.

[0012]

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

Embodiment 1 FIG. FIG. 1 is a schematic diagram showing the overall configuration of a bipedal walking robot according to a first embodiment of the present invention. 1 is the right foot of the robot, 2 is the encoder built-in motor RR that moves the right foot 1 left and right, 3 is the encoder built-in motor RP that moves the encoder built-in motor RR2 back and forth, 4
Is a body of the robot, 5 is a motor LP with a built-in encoder that moves the left foot 7, 6 is a motor LR with a built-in encoder that moves the left foot 7 to the left and right, 7 is the left foot of the robot, and 8 is a three-axis inertial sensor fixed to the body of the robot It is a rate gyro. Next, the mutual relation of each part will be described. One end (upper end) of the right foot 1 of the robot has a first encoder built-in motor RR
It is fixed to 2 and the first encoder built-in motor RR2
The rotation shaft of is fixed to the rotation shaft of the second encoder built-in motor RP3. The second motor RP3 is fixed to the body 4. The third encoder built-in motor LP5 is fixed to the body 4, and the rotation shaft of the third encoder built-in motor LP5 is fixed to the rotation shaft of the fourth encoder built-in motor LR6. One end (upper end) of the left foot 7 is fixed to the fourth encoder built-in motor LR6.

Next, the operation will be described. The robot rhythmically moves the fourth encoder built-in motor LR6 and the second encoder built-in motor RR3 to generate left and right movements so that one foot floats alternately. Hereinafter, the floating foot is called the free leg, and the grounded foot is called the supporting leg. Second encoder built-in motor RP3 and third encoder built-in motor L
The front-rear positional relationship between the support leg, the body 4, and the free leg is controlled by P5 to perform the front-back movement. Each joint angle is detected by each encoder of each motor with built-in encoder 2, 3, 5, and 6. Also, the posture of the body is calculated from the output of the three-axis rate gyro 8, and each value is calculated from each joint angle. Find the angles of legs 1 and 7.

The outline of the forward and backward movement during walking is as shown in FIG. When the swing leg (left foot) rises to some extent, swing the swing leg forward. The supporting leg (right leg) falls forward while being affected by the speed and gravity until then, and the free leg lands. Here, the supporting legs change, and the right leg, which was the supporting leg, becomes a free leg and goes up. This operation is repeated to walk.

FIG. 4 is a block diagram showing a schematic configuration of a control unit of the form 1 of FIG. 3. Reference numeral 21 is a rate gyro as an inertial sensor, 22 is a posture detection unit that detects the posture (inclination with respect to the vertical axis) of the body of the biped robot based on the rate gyro 21, and 23 is each joint of the leg (right foot, left foot) The joint angle detection unit that detects the angle of the leg, 24 is the leg angle detection unit that detects the longitudinal angle of the leg (the angle with respect to the vertical axis), and 25 is the movement of the robot obtained from the posture detection unit 22 and the joint angle detection unit 23. A posture detection error estimation unit that estimates a posture detection error by the posture detection unit 22 is a posture detection value of the posture detection unit 22 that is a value obtained by multiplying the posture detection error estimated by the posture detection error estimation unit 25 by a predetermined coefficient. In addition, a coefficient multiplication unit for correction, and 27 is a sampler that supplies the output of the coefficient multiplication unit 26 to the posture detection unit 22.

Next, the error estimation method of the attitude detection error estimation unit will be described. The behavior of the stilt-type walking robot as shown in FIG. 1 when supporting one foot is as shown in FIG.
As you can see, it can be approximated as an inverted pendulum. In FIG. 4, θ (t) represents the angle of the supporting leg at time t (the vertical is 0), but when the inverted pendulum is linearly approximated, time t =
If θ (t) at 0 is θ (0) and the time derivative dθ (t) / dt at that time is θ ′ (0), then θ
(T)

[0018]

[Equation 1] Becomes However, k is a constant.

As shown in FIG. 5, t = 0 is the time immediately before the landing of the free leg in one step, and t = -T is the time immediately before the start of the free leg moving backward. At θ
If (t) is θ (-T),

[0020]

[Equation 2] Becomes (T = 0 is defined in this way because the noise of dθ (t) / dt is small in one step. In addition, t = -T is defined in this way because This is because there is little noise due to movement, and the vibration due to support leg replacement is attenuated.)

In the walking robot, the swing leg is largely moved for walking and moves by Δφ from t = -T to t = 0. However, since the initial movement is large, Δφ is instantaneously changed immediately after time t = -T. It can be approximated as moving. At this time, the accuracy can be improved by adding a term due to the reaction to the equation of the movement of the supporting leg,

[0022]

(Equation 3) α can be approximated as a constant.

Here, the measured value of the longitudinal angle of the supporting leg calculated from each joint angle and the detected posture value of the body is represented as Θ (t), and Θ (t) at time t = -T, t = 0. Θ (-
Represented as T) and Θ (0). At this time, assuming that the attitude detection error is constant from t = -T to t = 0, and the measurement error is set as e, the measured values Θ (-T) and Θ (0) are

[0024]

(Equation 4) It is expressed as

By substituting this into equation (3),

[0026]

(Equation 5) Becomes

Therefore

(Equation 6) And the angle of the supporting leg at t = 0 is estimated to be Θ (0) -e. This e is the output of the posture detection error calculation unit 25.

Embodiment 2 FIG. 6 is a configuration diagram showing another embodiment 2 of the present invention. In FIG. 6, 28 is a walking speed command unit that generates a walking speed command value for the leg, and 29 is 1
A walking speed calculation unit that calculates an actual average walking speed from the walking distance and the time required for it, and a correction value calculation unit 30 that obtains a posture detection correction value from the values of the walking speed command unit 28 and the walking speed calculation unit 29. Reference numeral 27 is a sampler, 21 is a rate gyro as an inertial sensor, and 22 is an attitude detector that calculates an attitude from the output of the rate gyro 21 and the output of the correction value calculator 30 that has passed through the sampler 27.

Next, the operation of this embodiment will be described. In an inverted pendulum type bipedal robot as shown in FIG. 1, the angle of the leg with respect to the ground is calculated from the output of the posture detection unit calculated from the value of the rate gyro attached to the body and each joint angle. . Therefore, when the front-back angle of the actual posture (the front is positive) is tilted before the value measured by the posture detection unit, the forward acceleration in the next one step is large as described later. And the walking speed (positive forward) increases. On the contrary, if it is tilted backward,
In the next one step, the forward acceleration is reduced and the walking speed is reduced. That is, when the error in the front-back angle of the posture detection unit 22 is positive, the walking speed decreases, and when it is negative, the walking speed increases. Therefore, each time the swing leg lands, the correction value calculation unit 30 subtracts the value of the walking speed command unit 29 from the value of the walking speed calculation unit 29, and a certain constant γ and one step are set for the value. It is multiplied by the required time to obtain the output of the correction value calculation unit 30. The output of this correction value calculation unit 22 is applied to the sampler 2
Once every 7 steps in addition to the front-back angle of the swing leg posture detection unit 24, it is set as a new value of the posture detection unit 22. As a result, it is possible to reduce the error in the front-rear angle of the posture detection unit 22 and prevent the error from accumulating, thereby realizing stable walking.

Now, the influence of the error of the posture detecting unit 22 on the next one step will be described. For example, as shown in Figure 7,
It is assumed that the body is controlled so that the front-back value of the output of the posture detection unit 22 becomes zero, and that the actual front-back angle (the front is positive) is inclined by η before the value of the posture detection unit 22. To do.
That is, the error in the front-back angle of the posture detection unit 22 is −η. Further, at this time, it is assumed that the angle (positive in the counterclockwise direction) with respect to the vertical axis during landing of the free leg for maintaining the speed command value is ζ. Since the control of the free leg is performed by the output of the posture detection unit 22 and each joint angle, when the posture detection unit 22 has the above-mentioned error, the angle at the time of landing of the free leg is as shown in the lower diagram of FIG.
ζ-η. In this way, the actual posture is determined by the posture detection unit 2
When leaning forward from the value of 2, the angle when the free leg lands and becomes a support leg leans forward, and the forward acceleration increases.

[0031]

As described above, according to the present invention, an inertial sensor such as a rate gyro is used for attitude detection, and the detection error is estimated by the attitude detection error estimation unit to detect the error detected by the inertial sensor. Since it can be automatically corrected, the posture detection error due to the inertial sensor does not accumulate, and the bipedal walking robot has an effect that it can walk even if the ground shape is unknown.

[Brief description of drawings]

FIG. 1 is a schematic view of a bipedal walking robot of the present invention.

FIG. 2 is a schematic diagram of a walking state of a bipedal walking robot as seen from the side.

FIG. 3 is a block diagram showing a first embodiment of the present invention.

FIG. 4 is an explanatory diagram of an inverted pendulum.

FIG. 5 is an explanatory diagram showing behavior during walking according to the first embodiment.

FIG. 6 is a block diagram showing a second embodiment of the present invention.

FIG. 7 is an explanatory diagram showing the influence of a posture detection error according to the second embodiment.

FIG. 8 is a schematic view of a conventional bipedal walking robot.

[Explanation of symbols]

21 rate gyro, 22 posture detection unit, 23 joint angle detection unit, 25 posture detection error estimation unit, 28 speed command unit, 29 walking speed calculation unit, 30 correction value calculation unit.

Claims (5)

[Claims] 1. A bipedal robot having a torso and two legs connected to the torso, and a posture detector that detects the posture of the torso using an inertial sensor, and each joint of the legs. A joint angle detection unit that detects the joint angle of the robot, and a posture detection error estimation unit that estimates a posture detection error by the posture detection unit from the movement of the robot obtained from the posture detection unit and the joint angle detection unit. A bipedal walking robot, characterized in that the posture value of the posture detection unit is corrected using the estimated value of the posture detection error estimation unit. 2. The posture detection error estimation unit estimates the posture of the body from the posture value detected by the posture detection unit and the joint angle of each joint detected by the joint angle detection unit. 2. The bipedal walking robot according to claim 1, wherein a difference between the estimated posture value obtained in step S3 and the posture value detected by the posture detection unit is obtained as a posture detection error. 3. A coefficient multiplication unit for adding a value obtained by multiplying the posture detection error estimated by the posture detection error estimation unit by a predetermined coefficient to the posture detection value of the posture detection unit to correct the value. The bipedal walking robot according to claim 2, wherein the bipedal walking robot is provided. 4. A bipedal walking robot having a torso and two legs connected to the torso, a posture detection unit for detecting the posture of the torso using an inertial sensor, and a walking speed command for the legs. A walking speed command unit that generates a value, and a walking speed calculation unit that detects the actual walking speed of the leg,
A bipedal walking robot, comprising: a correction value calculation unit that corrects the posture value of the posture detection unit based on the walking speed command value and the actual walking speed detected by the walking speed calculation unit. . 5. The correction value calculation unit subtracts the output value of the walking speed command unit from the output value of the walking speed calculation unit,
5. The bipedal walking robot according to claim 4, wherein the value obtained in this way is multiplied by a predetermined constant and the time required for one step to obtain a correction value.