JPH04201187A - Walk controller on leg type mobile robot - Google Patents

Abstract

PURPOSE:To smoothly continue change in attitude by correcting a control value by a control value correcting means so that a predetermined point of a leg part link may pass a predetermined reference position on the space when a change in walking is commanded. CONSTITUTION:A control value deciding means for setting walking of a robot 1 and for deciding a control value for driving a joint of a leg part link in conformity to the set walking and a walking change commanding means for commanding change in the set walking are provided. When a change in walking is commanded by this commanding means, the control value is corrected by the control value correcting means so that a predetermined point of the leg part link passes a predetermined reference position on the space. Then, joints 10R, 10L, 12R, 12L, 14R, 14L, 16R, 16L, 18R, 18L, 20R and 20L are driven by an actuator based on this corrected control value.

Description

[Detailed Description of the Invention] (Industrial Application Field) The present invention relates to a walking control device for a legged mobile robot, and more specifically, in a legged mobile robot such as bipedal walking, when the posture of the robot collapses. The present invention relates to a gait control device for a legged mobile robot that restores postural stability by changing its gait.

(Problems to be solved by conventional technology and the invention One of the methods for stabilizing the walking of a legged mobile robot is to detect the state of the robot while walking and determine the robot's gait according to the detected state quantity. However, there are robots that calculate in real time where to take the next step and make the robot walk.An example of a robot walking using this method is, for example, 'L
EggedRobots on Rough Terra
ain: Experi+wents in Adj-
using 5step Length: IEEE
1988. by Jessica Ho-dgins
It has been proposed in J. et al.

However, in this type of technology, the state of the robot is recognized, and then a very complex determinant is solved to determine the gait, the next optimal landing position is calculated, and the swing leg is actually adjusted to that position. Since a series of processes for moving to a position were performed in real time, the determinant would be large for a complex bipedal locomotion robot with, for example, six degrees of freedom in one leg, and the computer required to execute it would be large. Either a very large and high-performance device is required, or a compromise must be made such as simplifying the mechanism of the robot itself, in other words, configuring the legs with fewer degrees of freedom to reduce the burden on the computer. For example, the real-time processing described above was difficult.

On the other hand, one method for making a robot walk is to use a large computer to calculate the optimal values for each joint angle of the robot in advance, and then store the resulting gait in the computer installed in the robot. A method of outputting output while walking has also been proposed.
-97 (One example is the technology described in Publication No. 106. According to this method, the calculation of the gait itself is performed off-line, so it may take time. Therefore, even if the robot has many degrees of freedom, , the burden on the on-board computer is greatly reduced, and a computer that is small, lightweight, low-level (low-priced), and consumes little power is sufficient.

However, the biggest drawback of this method is that since the gait is set in advance, it cannot deal with cases where the actual walking conditions differ from the set conditions. If there are unexpected irregularities in the area where the robot actually walks, the robot will receive a reaction force of a magnitude different from the planned reaction force at an unexpected timing, and if the difference is large, the walking posture will be disturbed. If the disturbance exceeds the physical limit of the compensation of the bottom area of the robot's soles, the robot will fall.

Therefore, an object of the present invention is to eliminate the above-mentioned drawbacks and provide a gait control device for a leg-hole mobile robot that combines the advantages of the two methods described above. To provide a walking control device for a mobile robot with leg holes, which can easily change a set gait to restore walking stability when unexpected walking conditions are encountered during actual walking.

Furthermore, when setting the target gait in advance off-line, it is not only necessary to set it in detail down to joint angles as described above, but also to set it using more general gait data such as landing position or center of gravity position. Furthermore, even when a method is used to determine the gait in real time while walking, it is inevitable that the user will encounter unexpected and sudden changes in the walking conditions and be forced to change the gait.

Therefore, the second object of the present invention is to move the leg holes so that the gait can be easily changed in response to unexpected changes in walking conditions, regardless of whether or not gait data is set offline in advance. An object of the present invention is to provide a walking control device for a robot.

(Means for Solving the Problem) In order to solve the above-mentioned object, the present invention includes, for example, claim 1.
In the gait control device for a leg hole mobile robot comprising a base body and a plurality of leg links each coupled to the base body and each having at least one joint, the gait of the robot is set, and the gait of the robot is set. control value determining means for determining a control value for driving the joints of the leg links; gait change command means for instructing a change in the set gait; When the gait is changed, the control value correction means includes a control value correction means for correcting the control value, and an actuator for driving the joint based on the determined or corrected control value, and the control value correction means includes:
When a change in gait is commanded, the set value is corrected so that a predetermined portion of the leg link passes through a predetermined reference position in space.

(Function) In addition to setting a gait and determining joint drive control values to achieve that gait, when the gait needs to be changed, the control values are corrected so that the changed gait is obtained, and furthermore, at that time, Since the leg links are made to pass through a predetermined posture, even if the posture collapses due to an unexpectedly mixed road surface, the gait can be corrected to easily recover. In addition, since the configuration is configured so that a predetermined posture is taken when changing the gait, the change in gait becomes smooth, and there is no need for a special angle correction value for posture connection.For example, data can be set offline in advance. (In some cases, the amount of data can be reduced.

(Example) Hereinafter, the present invention will be described in detail by taking a bipedal walking robot as an example of a legged mobile robot. FIG. 1 is an explanatory skeleton diagram showing the entire robot 1, which has six joints (axes) on each of the left and right legs. The six joints (
From top to bottom, the joints (axis) for rotation of the legs at the waist (IOR)
, 10L (the right side is R1 and the left side is L. The same applies hereafter), waist pinch direction (7) Sekii! l'j (axis) 12R, 1
2L, joints (axes) in the same roll direction 14R, 14L, knee pi.

joints (axes) in the pitch direction 16R, 16L, joints (axes) in the pitch direction of the ankle 18R, 18L, joints in the roll direction (
shaft) 2OR, 2OL, and the foot part 2 is at the bottom.
2R and 22L are attached, and the topmost body part (
A control unit 26 is provided inside the base body 24.
is stored.

In the above, the hip joint is a joint (axis) IOR (L), 12
It is composed of R(L) and 14R(L), and their axes are all arranged to intersect at one point. Also joint (axis) 1
The axes of the ankle joints, which are composed of 8R(L) and 2OR(L), are perpendicular to each other, and the three pitch direction joints (axes) 12R(L) and 16R(L) mentioned above
, 18R(L) are arranged parallel to each other, and their relative positional relationship always remains unchanged regardless of the behavior of other joints (axes). As shown in the diagram, 6 for 0 legs on one side.
Even if the torso 24 is fixed, the legs 2 are given two degrees of freedom.
2R(L) can be placed at any position and in any direction. That is, while walking, these 6X2=12
By driving each joint (axis) at an appropriate angle, the entire leg can be given a desired movement, allowing the user to walk freely in three-dimensional space. The hip joint and knee joint are connected by thigh links 27R and 27L, and the knee joint and ankle joint are connected by lower leg links 28R and 28L.

These joints mainly consist of an electric motor and a reduction gear that boosts its output. The details of the parts below the knee joint will be explained below with reference to FIGS. 2 and 3, but the lumbar joint has a similar structure. It should be noted that since it is bilaterally symmetrical, the following explanation will be given for the right leg side. In FIG. 2, the output of an electric motor (not shown) attached to the midway position of the thigh link 27H is transmitted to the knee joint (
It is input to the input shaft of the harmonic reducer 84 mounted on the shaft) 16R. A recess 87 is also formed on the upper end side of the lower leg link 28R, and an electric motor 88 is housed therein.
The output is inputted via the heald 90 to a harmonic reduction gear 92 disposed at the ankle, which drives the foot 22R in the pitch direction about the axis 18R mentioned above. Further, the foot portion 22 is configured to be able to swing freely in the roll direction about the aforementioned axis 20R that is perpendicular to the axis 18R, and for this purpose, the harmonic reduction gear 94 and the electric motor 96 that supplies power thereto are directly connected. provided. Each electric motor is provided with a rotary encoder, and only the rotary encoder 89 for the electric motor 88 is shown in Figure C for detecting the rotation angle of the motor shaft).

A six-axis force sensor 98 is provided at the ankle, and x and y are transmitted to the robot via the foot.

The three-direction components of No. 2 and the three-direction components of the moment are separated and measured separately to detect whether or not the foot touches the ground and the magnitude and direction of the force applied to the supporting leg. Further, a known grounding switch 99 is provided at the four corners of the sole of the foot, which is formed substantially flat and is equipped with an elastic material 220 such as rubber to absorb the impact upon landing, to detect whether or not there is contact with the ground (second (Omitted in Figures 3 to 3).

Furthermore, as shown in FIG. 1, at appropriate positions on the body portion 24,
A pair of tilt sensors 1 (10, 102 are installed, x-z
The angle with respect to the Z axis in the plane and its angular velocity, similarly y-z
Detects angles and angular velocities with respect to two axes in a plane. These outputs are sent to the control unit 26 in the body section 24 described above.
sent to.

FIG. 4 is a diagram showing details of the control unit 26, which is composed of a microcomputer. There, the output of the tilt sensor 1 (10, 102, etc.) is converted into a digital value by the A/D conversion circuit 104, and the output is transferred to the bus 1.
06 to the RAM 108. Further, the output of the encoder 89 etc. is sent to the RAM 108 via the counter 110.
The outputs of the ground switch 99 and the like are similarly stored in the RAM 108 via the waveform shaping circuit 112. A CPU 114 is provided in the control unit, and as described later, the CPU 114 reads stored walking data, calculates a speed command value from the deviation from the actual measurement value sent from the counter 110, and performs D/A conversion. circuit 1
The signal is sent to the servo amplifier 120 via 18 to realize position control. Also, as shown in the figure (the encoder output is F/
The signal is sent to the servo amplifier via the V conversion circuit 122, realizing speed feed bank control as a minor loop. In addition, numeral 128 is a jiko stick for commanding gait changes such as course and stride length, and numeral 130 is an origin (
The reference numeral 132 indicates an origin switch for determining the upright (upright) posture, and a glue mint switch for preventing overrun.

The operation of this control device will be explained below with reference to the flow chart in FIG. The control shown in the figure assumes that the target joint angles are set offline in advance, and also uses virtual compliance control ("virtual compliance control for multi-degree-of-freedom robots") that realizes impedance control by velocity resolution control. (Proceedings of the Society of Instrument and Control Engineers, VOL.

22, NO. 3, March 1986).

First, after initializing each part of the device with SIO, 312
A walking pattern nθi is searched for (the gait determined from this walking pattern is referred to as a "standard gait" in this specification). This indicates the target value of each joint angle when the robot walks on an ideally flat road surface with uniform hardness. Here, the subscript i indicates the joint number, and the subscript indicates the angle at time n. The joint numbers are 2 for the left foot, starting from the bottom.
0L=1.18L=2. ,．． 10L=6, and similarly for the right foot 2OR=7.18R=8. ,.

Let 10R=12. These time series data are calculated in advance by a large computer and stored in the ROM 116 of the microcomputer in the control unit.

Next, in S14, parameters Kp, Kv, . . . are input. These are feedbank gains, details of which will be described later. Next, in S16, the timer value n, counter value c, U
NT, reset the joint number (counter) value i to zero,
Walking is started at 31B, and the joint number i counter value is set to 1 at S20. Next, parameters such as joint angle nθi (i=1) corresponding to the joint number set in S22 are read from the memory. Here, n+1 θi indicates the time next to the current time t, that is, the target joint angle at the next time the program is started. An example of the walking pattern data is shown in FIG. 6 (as shown in the figure, target joint angles are set offline in advance for joint numbers 01 to 12 at times n, t, u, and 0.). Also, in S22, ωDt is the target angular velocity (
(described later). Furthermore, Ft (ω-) is the bipedal support phase, F
t(ωs) is a one-leg support period, Ft(C) is a shock absorption control period, Ft(T) is a gait switching period, and Ft(S) is a flag (described later) indicating a detecting unit that is likely to fall (fall risk). Yes, it is determined from the output of the 6-axis sensor, etc. in the microcomputer described above, and the bit is set to 1 when the period is in question.

Next, in S24, the detected value of the tilt sensor etc. is read. Here, iθR represents the actual joint angle of the i-th joint, ωR represents the actual tilt angular velocity, and M represents the actual moment applied to the foot. Subsequently, in S26, it is determined from the flag described above whether or not there is a detecting unit that is likely to fall. This point will be explained with reference to FIG. In the lower part of the figure, "IC" means heel contact, ie, landing, and "TO" means toe-off, ie, getting off the ground. Therefore, regarding the left and right feet, the period shown by solid lines is the stance phase, the period shown by broken lines is the swing phase, and the period where the solid lines overlap is the two-leg support period. In the embodiment, as shown in the figure, the gait is switched during the swing phase, and the first half is used as a fall detection period, the risk of falling is detected there and a corrected gait is determined, and the second half is the corrected gait. I am trying to execute it.

Now, to explain the correction of the gait in this embodiment, in this embodiment, as shown in FIG. 8, a plurality of preliminary landing positions are prepared in addition to the standard landing position. In other words, the black circles in Fig. 8 are the expected landing positions according to the walking pattern calculated in advance by the offline large-scale computer shown in Fig. 6, but there are 24 preliminary landing positions indicated by white circles on the front, back, left, and right sides. A landing position is prepared, and when the robot is about to fall, it selects the optimal landing position to recover its posture, changes its gait to land there, and
That is, the target joint angles were corrected to restore postural stability.

This possibility of falling) is specifically detected by the 9th
As shown in the figure, the deviation between the target angular velocity ωDT and the actual angular velocity ωR is integrated to obtain the sums χθ and yθ in the x and y directions, and the gait is selected accordingly. That is, the ninth
In the figure, if the current sum is Xθa, the gait corresponding to the position orthogonal to it and the selection function is a, so in Figure 8, the gait a (landing position) is located slightly ahead of the standard gait. , and correct the target joint angles to match that gait. The details are shown in FIG.

That is, the ankle joint 20L (i = 01) is tOθOf,
If it is determined to switch the gait after driving tl θ01,..., t5 θi, the gait is corrected from that point on, and the target values are set to tiθO1a, tjθO1a,...
and amend it. In addition, at this time, not only the ankle joint 20L,
The remaining 11 joints were also revised in whole or in part,
In the illustrated example, 2OL, 16L, IOL, 2OR, and 12R.

Modify 10R.

What should be emphasized here is that a predetermined reference posture is guaranteed to be output at the beginning of the gait switching period ("tiθna" in FIG. 10). That is, the posture shown in FIG. 11 is changed once and then the gait is changed to the corrected gait. The reason for this configuration is to prevent the number of gaits calculated in advance and stored in the computer memory from increasing in proportion to the square of the landing positions as the number of preliminary landing positions increases. In other words, in the example shown in Figure 8, 25 landing positions are prepared, but this is done with an example in mind where the gait is first changed from a standard gait. If you correct it to a, the next landing will have a gait of a.
There is a possibility that the gait will be corrected from 1 to 2, and then the gait may be corrected to C for the next step. This means - If we prepare 25 candidate landing points for each landing, there are 25 landing points where the next step may land from the 25 landing positions where the foot may be currently placed. Considering all combinations, the gait for landing the swing leg in the position is 25X
25=625 different gaits must be set in advance. In consideration of this point, in this embodiment, a predetermined reference posture is taken once at the time of switching, that is, the joint angle is reset to a predetermined angle and then corrected. With this configuration, it is possible to suppress an increase in the number of gaits that should be set in advance.

Returning to the flow chart of FIG. 5 again, when it is determined in S26 that it is in the detection period that it is likely to fall, in 32B, the integral value of the inclination angular velocity, that is, the total sum of inclination angles Xθ is determined as described above.
, yθ is determined. If the answer is NO, the total sum is set to zero in S30. Next, in S32, it is determined whether or not the position is in the gait switching section, and when it is affirmative, the process proceeds to S34, where the total sum Xθ
, yθ, and read out the corrected target joint angles of the selected gait in S36.

Then, at 33B, the position feed hack control value iVl
is calculated, and a speed feedback (forward) control value iV2 is calculated in S40. That is, as shown in FIG. 12, the joint angle command value nθl and the actual joint angle i θ
A position feedback value obtained by multiplying the deviation Δθ from R by a proportional gain kp, and a feedforward value obtained by multiplying the deviation between the joint angle command value nθl at the current time and the joint angle command value n+1 θ1 at the next time by a gain are added. The speed command value is output to the servo amplifier 120. In addition, No. 12 is a block diagram of the joints excluding the ankle joint, and the ankle joint is based on compliance control as shown in the block diagram of FIG. 13 (control values etc. are also fed back, but this will be described later). do.

Next, in step 342, it is determined whether or not it is an ankle joint, and if the answer is affirmative, the process proceeds to step S44 and subsequent steps, where tilt angular velocity feedback and the like are first performed. Specifically, first, S44 to S46
Then, it is determined whether the flag Ft (ω-) or Ft (ωS) is in focus, that is, whether it is a two-leg support period or a one-leg support period, and based on the determination result, the target inclination angular velocity is determined at 348 or S50 as shown in FIG. The third velocity feedback control value iV3 is calculated by multiplying the deviation Δω between ωDt and the actual tilt angular velocity ωR by the gain. In other words, in this example, in addition to compliance control, aiming to further stabilize walking, there is a risk that the robot will fall if the inclination angular velocity deviates from the target value, or a moment is applied to the ankle joint from the outside. The system determined that the deviation was corrected by multiplying the deviation by an appropriate gain by ω to drive the ankle joint of the landing leg to generate a ground reaction force and correct the collapse of the robot's posture. In addition, when it is determined in the flow chart of FIG.
Set the control value to zero.

Subsequently, a virtual compliance control value of 354 or less is determined. That is, the joint drive speed is controlled according to the moment acting on the ankle joint to reduce the impact upon landing. Specifically, the predetermined time T from when the robot's free leg leaves the floor until it lands on the floor.
COMP is the shock absorption control period, and when it is determined in S54 that it is in that period, the process advances to S56, where the gain k is
Calculate c as kc = kcomp
If landing is detected at 0, the counter value is incremented at 362. In this way, the impact absorption gain is set as a function of the counter value c and UNT, and as shown, it is set so that it gradually decreases over time at the same time as the landing, and finally reaches zero.

If it is determined in S54 that the shock absorption control period is not in effect, the control value iV4 is set to zero in S64, and the counter value is reset to zero in S66.

Next, add all the control values calculated in 36B to get the total i
Find VcOMM and output it to the servo amplifier 120, S7
Increment the value of the joint number counter by 0, and proceed to S72.
It is determined whether the joint is the last joint or not. If the answer is negative, the timer value n is incremented to search for the next target joint angle in S74, and the control value is continuously updated for each joint unless it is determined that the walking has ended in S76. Determine.

As described above, in this embodiment, in addition to the standard gait, multiple gaits with different landing positions of the swing legs are prepared in advance, and when it is determined that the robot is likely to fall down, these gaits are prepared in advance. The system is configured to select the optimal gait from among the given gaits. In other words, the standard gait is calculated in advance, which reduces the burden on the onboard computer, and the gait with different landing positions is also calculated in advance, so the onboard computer stores this data. It is sufficient to select the optimal gait from among them according to a determined procedure, and as a result, it is possible to secure sufficient real-time performance even using a low-level computer, and to create a robot with multiple degrees of freedom. Able to walk stably.

In addition, when designing the above-mentioned gait, as is well known (generally, the movement is calculated by dividing it into two planes, a plane in the direction of movement and a plane in the left and right directions, and then combined again after calculation. , when using that method, the number of landing candidate positions described above is 2.
Instead of 5 pieces, 9 pieces will be enough. However, when using this method, the accuracy of the gait decreases as the gait moves away from the standard gait, such as from the right front to the left rear, resulting in an error. The more candidate points you have, the better your ability to respond to large changes in the situation, so if you generalize this idea and have m candidate points before and after the standard position, and n candidates to the left and right, you will have the ability to respond to large changes in the situation. If gaits are required, the total number of gaits to be stored is (2m)×(2n)=4mn. However, if a small error is tolerated, 2m+2n-1 is enough, and even if the number of candidate points increases in the future, the storage capacity of the computer can be significantly saved.

Furthermore, candidate points may be determined by linking them using a method such as linear interpolation, or by interpolation from gaits corresponding to a finite number of pre-stored landing positions. The data may be obtained by interpolating if necessary to save storage capacity.

Furthermore, when changing the gait, it is configured to pass through the reference posture shown in FIG. 11 in order to save memory capacity.Although this reference posture is shown only as an angle in the example, it is not limited to this. Instead, the reference posture may include angular velocity in addition to the angle, and may even include jerk, and the gait may be switched after passing through this state.

FIG. 14 shows a second embodiment of the invention. In the case of the first embodiment, multiple landing points were prepared, and all joints were modified to select one of them and land there, but in the case of the second embodiment, when a risk of falling was detected. As shown in the figure, the hip and ankle joints were modified to restore postural stability.

Specifically, in step 334 of the flowchart in FIG. 5, the correction angle δθ is determined based on the determination function shown in FIG.
Adjust the joint angle settings for the hip and ankle joints.

Specifically, if the sum is the value in the X direction, that is, Xθ, the joints in the pinch direction (12L (R), 18L (R)) are corrected, and if δθ determined in If it is a negative value, it is added, and if it is a negative value, it is subtracted. FIG. 16 shows the corrected data. The figure only shows corrections in the pitch direction, but
The same applies to the roll direction, and δθ is determined from yθ according to a function that is the same as or similar to the determination function shown in FIG. 15, and the corresponding joint (14L, (R)2OL(R)) is adjusted.

Note that when switching the gait, switching from the initially set standard gait to the above-mentioned modified gait at once will impair smooth walking, so the gait is gradually modified as shown in Figure 17. I tried to do it. In the example shown in the figure and FIG. 16, the correction was made in five steps. This also applies when the patient regains postural stability and returns to a standard gait.

In the case of the second embodiment, there is no need to prepare multiple landing positions, and only a few joints need to be corrected, so it is easier to adjust the posture even when the set walking data encounters unexpected changes in conditions. It is possible to correct the collapse and restore a stable posture. In the case of the second embodiment as well, the reference posture is assumed to be passed once during the gait switching period. As a result, the gaits to be stocked can be stored using independent time-series joint angle data, and special joint angle data for connection is not required, so that storage capacity can be saved.

FIG. 18 is an explanatory diagram showing a third embodiment of the present invention, which differs from the first embodiment in that the angle is corrected not by addition or subtraction but by multiplication or division. That is, as shown in the figure, the correction coefficient is set to kl and 2 when it is a positive value, and 1 /kl and 1 when it is a negative value, depending on the total sum Xθ (yθ).
/ was determined to be 2, and the correction was made by multiplying it by the joint in the corresponding pinch (roll) direction. FIG. 19 is an explanatory diagram showing the modification. In the third embodiment as well, as shown in FIG. 20, it is increased over time to achieve smooth gait changes.

In this third embodiment and the second embodiment described above, an example was shown in which the amount of correction is multiplied (divided) and added (subtracted), that is, corrected using a linear function. For the pitch direction, nθ1REV=nθ1sTDXAxθ+Bxθ°, °n
θi REV: joint angle of modified gait, nθ1STD: joint angle of standard gait, A, B: constants.

In any case, the second. In the third embodiment, various modifications are possible; for example, the direction of the correction amount was changed depending on the sign of the sum, but the magnitude was made the same, but each may be determined independently. In the third embodiment, another correction coefficient may be used instead of division. Furthermore, it does not have to be a linear function, and may be determined using, for example, fuzzy inference.

Further, in the first to third embodiments described above, the risk of falling (likelihood of falling) was calculated from the inclination angle, but it is not limited to this, and may also be calculated by the inclination angular velocity, for example, the point of action of road reaction force. may be calculated, and it may be determined that the user is likely to fall forward when the motion speed of the locus of the point of action moves faster from the heel-equivalent region to the toe-equivalent region than in the normal state.

Further, in each of the above embodiments, only the joints on the swinging leg are subject to correction, but it is not necessary to limit the correction to the swinging leg. For example, the joints on the stance leg and the swinging leg may be corrected equally.

Furthermore, although the method is based on a method in which the target joint angles are calculated offline in advance, the method is not limited to this. Even if the target data is similarly set offline, it is possible to calculate the target joint angles in advance, such as the landing position or center of gravity position. It is also appropriate when using a more general concept, and even when assuming a method of determining gait in real time, it is appropriate as far as circumstances allow, such as when sudden changes in road surface conditions occur.

Although the present invention has been described with respect to a legged mobile robot that walks on two legs, it is not limited thereto, and is also applicable to legged mobile robots with three or more legs.

(Effect of the invention) The walking control device for a legged mobile robot according to claim 1 has the following features:
Set the robot's gait and make sure the gait is as set as possible.
control value determining means for determining a control value for driving the joints of the leg links; gait change command means for instructing a change in the set gait; Preferably, the control value correction means includes a control value correction means for correcting the control value and an actuator for driving the joint based on the determined or corrected control value, and the control value correction means is configured such that when a change in gait is commanded, Since the control value is corrected so that a predetermined point of the leg link passes through a predetermined reference position l in space, it is possible to immediately correct the posture with the corrected gait in response to changes in walking conditions. This allows you to walk stably. In addition, since the leg links pass through the reference posture to change the gait, changes in posture are smoothly continuous, and even when storing the gait as data in advance, for example, special connection data The amount of data can be reduced without the need for
As a result, the control device can be implemented using a relatively low-level computer.

A walking control device for a legged mobile robot according to claim 2,
control value determining means for determining control values for driving the joints of the leg links so as to achieve a preset gait; landing candidate position setting means for presetting a landing candidate position; gait change command means for commanding a change in the set gait; and when a change in gait is commanded, one of the plurality of landing positions is selected. In order to select and land at that position, the configuration includes a control value correction means for correcting the drive-side a value of the joint and an actuator for driving the joint based on the determined or corrected control value. Even when the set gait encounters an unexpected change, it is possible to quickly correct the posture and continue walking stably.Also, when implementing the control device with a computer, it is possible to memorize the data calculated offline in advance. Since it is only necessary to select and output the information at the appropriate time, it can be configured with relatively low-level information.

A walking control device for a legged mobile robot according to claim 3,
A control value determining means for determining a control value for driving the joints of the leg links so as to achieve a preset gait; a gait change command means for instructing a change in the preset gait; and a gait change. When commanded, a gait corresponding to an arbitrary landing position is generated by interpolation and extrapolation from gaits corresponding to a finite number of landing positions stored in advance, and the joints are moved in order to land at that position. Since it is configured to include a control value correction means for correcting the drive control W value of and an actuator for driving the joint based on the determined or corrected control value,
Even when the set gait encounters an unexpected change, it is possible to quickly correct the posture and continue walking stably, and even when implementing the control device on a computer, the data calculated offline in advance is memorized. Since it is only necessary to select and output the information at the appropriate time, it can be configured with relatively low-level information.

The walking control device for a legged mobile robot according to claim 4 comprises:
The control value correction means, when commanded to change the gait,
Since the control value is corrected so that a predetermined point of the leg link passes through a predetermined reference position in space, the number of gaits to be set can be effectively reduced, and - The burden on the computer can be reduced. Furthermore, it is possible to realize a smooth change in posture. More specifically, the number thereof is set as described in claim 5.

A walking control device for a legged mobile robot according to claim 6,
A control value determining means for determining a control value for driving the joints of the leg links to achieve a preset gait; a gait change command means for instructing a change in the set gait; When commanded, the controller includes a control value correction means for correcting the drive control value of the joint using a zero-order function so as to achieve a changed gait, and an actuator for driving the joint based on the determined or corrected control value. With this configuration, even if the preset gait encounters an unexpected change in walking conditions, the joint drive control values can be quickly corrected to restore the posture, allowing stable walking to continue, and the computer When configuring a control device, it can be realized with a relatively low-level device.

A walking control device for a legged mobile robot according to claim 7,
The control value correction means, when commanded to change the gait,
Since the control value is corrected so that the leg link passes through a predetermined point or a predetermined reference position in space, special data for connecting joint angles is not required when switching gait. , the amount of data to be stored can be saved, and smooth changes in gait can be realized.

A walking control device for a legged mobile robot according to claim 8,
When the control value correction means corrects the control value, the set gait should be changed to a temporally smooth gait (
Since the control value is corrected, the gait can be changed smoothly, and stable walking can be achieved.

A walking control device for a legged mobile robot according to claim 9,
Since the control value correction means is configured to correct the control value during a certain predetermined period during one cycle of the gait, the control period can be made clear and the gait can be switched in a timely manner. It is possible to achieve stable walking.

In the walking control device for a legged mobile robot according to claim 10, the gait change command means includes a detection means for detecting an n-th differential value of the inclination angle of the base body and/or the leg link, and the detected value Since the structure is configured so that a change in the gait is commanded when the robot's gait is greater than a predetermined value, it is possible to reliably detect the risk of the robot falling over, and the posture can be repositioned accordingly to achieve stable walking. be able to.

[Brief explanation of the drawing]

Fig. 1 is a schematic diagram showing the overall walking control device for a legged mobile robot according to the present invention, Fig. 2 is an explanatory side view showing the parts below the knee joint in Fig. 1, and Fig. 3 is the - ■ Line partial sectional view, Figure 4 is an explanatory block diagram of the control unit, Figure 5 is a flow chart showing the operation of this control device, and Figure 6 is a walking pattern (standard gait) set in advance offline. FIG. 7 is an explanatory diagram showing the swing phase and stance phase of the left and right legs on a time axis, as well as the fall detection phase and corrected gait output phase, and FIG. 8 is an explanatory diagram showing the landing candidate position. 9 is an explanatory diagram showing the selection function of the corrected gait; FIG. 10 is an explanatory diagram showing the correction of the walking pattern of FIG. 6 by it; FIG. 11 is an explanatory diagram showing the reference posture; FIG. 12 is a block diagram explaining the control algorithm shown in the flow chart in FIG. 5 for joints other than the ankle joint, FIG. 13 is a Bronk diagram similarly explaining the control algorithm for the ankle joint, and FIG. 14 is an explanatory diagram showing the modification of the control value in the second embodiment of the present invention, FIG. 15 is an explanatory diagram showing the determination of the amount of modification (angle), and FIG. 16 is the modification of the walking pattern of FIG. 6 accordingly. FIG. 17 is an explanatory diagram showing the temporal characteristics of the correction amount, FIG. 18 is an explanatory diagram showing the correction coefficient of the control value according to the third embodiment of the present invention, and FIG. FIG. 6 is an explanatory diagram showing the modification of the walking pattern, and FIG. 20 is an explanatory diagram showing the temporal characteristics of the modification coefficient of the third embodiment. 1... Legged mobile robot (biped walking robot), IO
R, IOL... Joint (axis) for leg rotation, 12R, 1
2L... Joint (axis) in the pinch direction of the crotch, 14R, 1
4L...The point in the roll direction of the crotch! fff (axis), 16
R, 16L... joint (axis) in the pinch direction of the knee, 18
R, 18L... joint (axis) in the pinch direction of the ankle, 2
OR, 2OL... joint (axis) in the roll direction of the ankle,
22R. 22L... Leg part, 24... Torso part, 26... Control unit, 27R, 27L... Thigh link, 28R,
28L...Lower leg link, 84,92.94...Harmonic reducer, 82.90...Held, 88.96
...Electric motor, 89...Low reencoder, 8
7.°゛ recess 98... 6-axis force sensor, 99... ground switch, 1 (10, 102... tilt sensor, 104
...A/D conversion circuit, 106...bus, 108...
・RAM, 110... Counter, 112... Waveform shaping circuit, 114... CPU, 116... ROM, 1
18... D/A conversion circuit, 120... Servo amplifier, 122... F/V conversion circuit, 12B... Jigoi Steink, 130... Origin switch, 132... Limit switch,

Claims (10)

[Claims] (1) a substrate and at least one
In a walking control device for a legged mobile robot comprising a plurality of leg links each having joints, a. control value determining means for determining a control value for driving the gait; b. gait change command means for instructing a change in the set gait; control value correction means for correcting the control value; and d, an actuator for driving the joint based on the determined or corrected control value; A walking control device for a legged mobile robot, wherein the set value is corrected so that a predetermined location of the leg link passes through a predetermined reference position in space. (2) a substrate and at least one
In a walking control device for a legged mobile robot comprising a plurality of leg links each having a plurality of joints, a) determining a control value for driving the joints of the leg links to achieve a preset gait; control value determining means; b. landing candidate position setting means for presetting a plurality of landing candidate positions including the landing position of the preset gait for each landing of the leg link; c. d. gait change command means for instructing a change in gait; d. when a change in gait is commanded, selecting one of the plurality of candidate landing positions and driving the joints in order to land at that position; A walking control device for a legged mobile robot, comprising: control value correction means for correcting a control value; and (e) an actuator for driving the joint based on the determined or corrected control value. (3) a substrate and at least one
In a walking control device for a legged mobile robot comprising a plurality of leg links each having a plurality of joints, a) determining a control value for driving the joints of the leg links to achieve a preset gait; control value determining means; b. gait change command means for instructing a change in the preset gait; c. corresponding to a finite number of pre-stored landing positions when a change in gait is commanded; control value correction means for generating a gait corresponding to an arbitrary landing position by interpolation or extrapolation from the gait, and correcting the drive control value of the joint in order to land at that position; A walking control device for a legged mobile robot, comprising: an actuator that drives the joint based on a control value. (4) The control value correction means is characterized in that when a change in gait is commanded, the control value is corrected so that a predetermined location of the leg link passes through a predetermined reference position in space. A walking control device for a legged mobile robot according to claim 2 or 3. (5) The joint control value is calculated by dividing the motion in two planes, a plane in the advancing direction of the robot and a plane in the left-right direction, and the landing candidate position is moved forward and backward by m. 5. The walking control device for a legged mobile robot according to claim 4, wherein the total number of the plurality of candidate landing points is m+n-1 when the number of candidate landing points is n in the left and right direction. (6) a substrate and at least one
In a walking control device for a legged mobile robot comprising a plurality of leg links each having a plurality of joints, a) determining a control value for driving the joints of the leg links to achieve a preset gait; control value determining means; b. gait change command means for instructing a change in the preset gait; A walking control device for a legged mobile robot, comprising: control value correction means for correcting using an n-th order function; and d, an actuator for driving the joint based on the determined or corrected control value. (7) The control value correction means is characterized in that when a change in gait is commanded, the control value is corrected so that a predetermined location of the leg link passes through a predetermined reference position in space. A walking control device for a legged mobile robot according to claim 5 or 6. (8) When the control value correction means corrects the control value,
8. Walking control for a legged mobile robot according to any one of claims 1 to 7, wherein the control value is corrected so that the gait changes temporally smoothly from the preset gait. Device. (9) The leg-based locomotion according to any one of claims 1 to 8, wherein the control value correction means corrects the control value during a certain predetermined period during one cycle of the gait. Robot walking control device. (10) The gait change command means includes a detection means for detecting an nth differential value of the inclination angle of the base body and/or the leg link, and commands a change in the gait when the detected value is a predetermined value or more. A walking control device for a legged mobile robot according to any one of claims 1 to 9.