JP2008229847A - Leg type robot and walking control method of leg type robot - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a leg type robot and its walking control method, for performing stable walking without using a tiptoe sensor. <P>SOLUTION: This leg type robot has a body trunk, a leg link rockably connected to the body trunk, a means 210 for storing tiptoe walking figure data for describing a change with the lapse of time of target tiptoe motion, a means 210 for storing body trunk walking figure data for describing a change with the lapse of time of target body trunk motion for realizing target ZMP for following a change in a target tiptoe position, a body trunk motion detecting means 218, 220 for detecting actual body trunk motion, a deviation calculating means 312 for calculating a deviation between the target body trunk motion and the actual body trunk motion, a correction quantity calculating means 308 for determining a correction quantity on the basis of a predetermined transfer function from the calculated deviation, and a correcting means 306 for correcting data for describing the target body trunk motion on the basis of the determined correction quantity. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to a technique for controlling and walking a machine (a mechanical system of a legged robot) in which a leg link is swingably connected to a trunk (torso). Specifically, the present invention relates to a legged robot that can stably walk even when an unexpected disturbance force is applied, and a walking control method thereof.

Legged robots have been developed that walk by changing the relative postures of the left foot link, trunk, and right foot link.
In order for a legged robot to walk, data indicating the motion of the left foot tip, trunk and right foot tip is required. Among them, the trunk position needs to be an appropriate value with respect to the foot position, and if the value is not appropriate, the robot falls.
In order to obtain a trunk position where a legged robot does not fall down, complicated calculations that take into account the dynamics of the robot are required. The calculation process is roughly as follows.
(1) Designate temporal data indicating the positions of the left and right foot tips of the robot.
(2) Designate the position where the ZMP of the robot must exist in consideration of the position of the toe. ZMP (zero moment point) refers to a point on the floor where the moment of the resultant force of gravity, floor reaction force and inertial force acting on the robot becomes zero. If the ZMP is within the foot of the grounded leg, the robot will not tip over. Conversely, in order for the robot not to fall, the ZMP must be in the foot of the grounding leg. Therefore, a target ZMP satisfying the following relationship is designated in consideration of the foot position of the grounding leg. That is, while one leg link (for example, the left leg) is a free leg, it exists in the foot of the ground leg (right leg), and the one leg (left leg) is grounded and both legs are grounded. At the time of movement, it starts to move toward the foot of the newly grounded leg (left leg), and before the leg (right leg) that was grounded before becomes a free leg, Specify the ZMP to finish moving in the foot. The ZMP designated in this way is called a target ZMP. If the actual ZMP moves as the target ZMP, the robot continues to walk without falling.
(3) When a change in the toe position and a target ZMP that changes following the change are specified, the dynamics of the robot is calculated on the assumption that the trunk position changes with time. At the time of calculation, since the toe trajectory is specified, assuming the position of the trunk of the robot, the posture of the robot is determined. When the posture of the robot is determined, the position of the ZMP in that posture can be calculated. In order to calculate the position of the ZMP, in addition to static elements, the influence of inertial force acting on the robot must be taken into account. By including the temporal change of the assumed trunk position in the calculation, it becomes possible to calculate the position of the ZMP in consideration of the dynamics of the robot. Since it is possible to calculate the position of the ZMP assuming a temporal change (trunk trajectory) of the trunk position, the trunk trajectory (change in trunk position over time) that realizes the ZMP that matches the target ZMP is searched. be able to.
The data indicating the temporal change of the trunk position searched for by the above is called trunk gait data, the data showing the temporal change of the originally specified foot position is called the foot gait data, and both Collectively called gait data. If the robot walks according to the gait data, the actual ZMP matches the target ZMP, and the robot can continue to walk without falling.
Gait data is given as a change in position with respect to time, but position, speed, and acceleration are related, and one of these can be used to calculate another, so position, speed, or acceleration Good. In this specification, since the gait data is described in any one of position, velocity, and acceleration, it is referred to as data describing movement.
The technique for calculating the trunk movement that brings about the ZMP that matches the target ZMP is a technique for calculating the change over time of the target trunk movement that allows the robot to continue walking following the change in the target toe movement. This is a typical example and is not limited thereto. Generally speaking, legged robots have a leg link that is swingably connected to the trunk, and a footstep that describes the change over time of the target foot movement. When the condition data is instructed, the trunk gait data describing the change over time of the target trunk movement that allows the user to keep walking following the change in the target toe movement is calculated, and the indicated foot Walk using the previous gait data and the calculated trunk gait data. The calculated trunk gait data may be calculated up to data on the trunk inclination angle in addition to data on the trunk position (if the trunk inclination angle follows a predetermined pattern, the trunk It may not be necessary to calculate the tilt angle). The data on the position and inclination angle of the trunk is the position and angle itself, the changing speed, and the acceleration.

Unexpected disturbance may act on the robot. An unexpected disturbance force may act due to an unexpected unevenness of the road surface, and an external force that pushes or pulls the robot from the outside may be applied. If it deviates from the gait data due to the structural deflection of the robot, the backlash of the robot joints, the response delay of the robot, etc., it is the same as an unexpected disturbance force.
If unexpected disturbance force acts, the robot may lose balance. Therefore, a walking control device described in Patent Document 1 has been proposed.
In the walking control device described in Patent Document 1, the floor reaction force or the floor reaction force moment acting on the robot is measured to actually measure the ZMP position. The measured ZMP position is compared with the target ZMP position, and the deviation is fed back to the target positions and postures of both feet to stabilize the walking.
JP-A-5-305579

In the walking control device described in Patent Document 1, it is essential to detect the floor reaction force or the floor reaction force moment. In order to detect the floor reaction force or the floor reaction force moment, the walking control device is provided near the toes of the right leg and the left leg. It is necessary to mount a high-performance force sensor such as a six-axis force sensor.
A high-performance force sensor such as a six-axis force sensor is expensive. Also, the foot tip is a part where a landing impact load repeatedly acts as the robot walks. If a sensor is mounted near the foot tip, the sensor is likely to break down due to the impact load. If the force sensor mounted on the toe fails, the robot cannot continue walking. Moreover, when a force sensor is mounted on the tip of a foot, many components are required to transmit a signal from the force sensor to the calculation unit. If any one of the components fails, the signal from the force sensor is not correctly transmitted to the calculation unit, and as a result, the force sensor does not function.
Against this background, in order to reduce the manufacturing cost of legged robots and improve reliability, it is not necessary to mount force sensors on the toes, and technology is required to stabilize walking with a simpler configuration. Is done.
The present invention was created to solve the above-described problems of the prior art, and a legged robot that can continue to walk stably without a force sensor mounted on the tip of the foot, and walking control therefor Provide a method.

The legged robot created in the present invention includes a trunk and a leg link that is swingably connected to the trunk. Furthermore, it has a storage means for foot gait data describing the change over time of the desired toe movement, and a storage means for trunk gait data describing the change over time of the target trunk exercise. . The target trunk exercise is set to a relationship that allows the user to keep walking following the change in the target foot exercise. For example, the trunk exercise | movement which makes actual ZMP correspond with the target ZMP which tracks the change of a target foot tip position is memorize | stored.
The robot according to the present invention further includes a trunk motion detecting means for detecting an actual trunk motion and a deviation calculating means for calculating a deviation between the target trunk motion and the actual trunk motion. Further, a correction means is provided for obtaining a correction amount from the calculated deviation based on a predetermined transfer function, and correcting the trunk gait data stored in the trunk gait data storage means based on the correction amount. Yes.

It is also possible to understand the present invention as a walking control method. In the gait control method of the present invention, a step of storing foot gait data describing a change over time of a target toe movement, and a body gait data describing a change over time of a target trunk exercise The process of memorizing is performed. The target trunk exercise is set to a relationship that allows the user to keep walking following the change in the target foot exercise. For example, the trunk exercise | movement which makes actual ZMP correspond to the target ZMP which tracks the change of a target foot tip position is memorize | stored.
The walking control method of the present invention includes a step of detecting an actual trunk exercise and a step of calculating a deviation between the target trunk exercise and the actual trunk exercise. Further, a correction amount is obtained from the calculated deviation based on a predetermined transfer function, the trunk gait data is corrected based on the correction amount, and the joint angle of the robot is calculated from the corrected trunk gait data. A process of instructing the apparatus is performed. A robot is provided with a device for calculating a joint angle for each joint of the robot.

In the above legged robot and its walking control method, the actual trunk movement of the robot is monitored, and the trunk gait data describing the target trunk movement is corrected from the deviation from the target trunk movement. Find the amount. Since the trunk movement is directly feedback controlled, the trunk movement is stabilized. The robot can continue to walk stably without losing balance.
Through experiments, it has been confirmed that the actual trunk movement is monitored and feedback-controlled, so that walking can be continued without actually measuring the ZMP position and performing feedback control. It has been confirmed that walking can continue even when disturbance force or the like acts.
According to this control method, it is not necessary to actually measure the ZMP position, and a force sensor for that purpose is not required. A highly reliable walking robot can be realized at low cost.

It is preferable that the deviation calculating means calculates a deviation between the target trunk acceleration and the actual trunk acceleration. In this case, the magnitude of the unexpected disturbance force acting on the robot can be calculated from the deviation in acceleration. Therefore, it becomes possible to calculate the correction amount of the target trunk motion necessary for causing the robot to exert a force sufficient to compensate for the disturbance force. In the correction amount calculation process, the correction amount is obtained based on the disturbance force calculated from the acceleration deviation and a predetermined transfer function.
When the robot is instructed to perform the target trunk motion corrected with the calculated correction amount, the robot can continue walking while compensating for the influence of disturbance force. This method is very practical because an acceleration sensor can be mounted on the trunk relatively easily.

When calculating the correction amount of the target trunk position from the acceleration deviation using the transfer function, the transfer function preferably includes a proportional element. In this case, when the actual trunk position deviates from the target trunk position, a correction amount proportional to the deviation is calculated, and as a result, a phenomenon of pulling toward the target trunk position with a spring is obtained.
More preferably, the transfer function includes a first-order lag element and / or a second-order lag element in addition to the proportional element. In this case, when the actual trunk position deviates from the target trunk position, a correction amount proportional to the change speed and / or acceleration of the position deviation is calculated, and as a result, converges toward the target trunk position. Phenomenon to be obtained.

If the actual trunk position can be detected, the target trunk position is further corrected by adding the amount of feedback processing to the deviation between the target trunk position corrected based on the acceleration deviation and the actual trunk position. It is preferable to do.
In this case, even if the robot walks on rough terrain and the entire robot tilts, the actual trunk position can be feedback controlled to the target trunk position, and the robot can stabilize the rough terrain with a simple mechanism. You can walk.

  According to the present invention, it is possible to realize a legged robot that can stably walk without using a force sensor at the toes. A legged robot that is inexpensive and highly reliable can be realized.

Example 1
The robot 6 shown in FIG. 1 moves the left leg as a swing leg 7 while the right leg is in contact with the ground, and when the left leg touches, this time moves the right leg as a swing leg 7 like the path 8. In the same manner, the left leg is moved as a free leg as in the track 9, and then the right leg is moved as a free leg as in the track 10 to continue walking.
In order for the robot 6 to continue walking, the toe trajectory data describing the trajectories 7, 8, 9, 10 of the toe reference points L and R, the arm tip trajectory data, and the trajectory of the reference point W of the trunk 12 Gait data consisting of trunk trajectory data describing is required. Foottip trajectory data is specified by the operator. The arm tip trajectory data is calculated corresponding to the foot tip trajectory data. The trunk trajectory data is set so that the ZMP trajectory calculated using the dynamic model of the robot matches the target ZMP trajectory.
In order to control the walking of the robot 6, a computer device 14 is mounted on the trunk 12 of the robot 6. The computer device 14 includes a CPU, a ROM, a RAM, a hard disk, and the like. The hardware configuration of the computer device 14 is the same as that of a general-purpose computer, and a description thereof will be omitted. The computer device 14 stores gait data, and controls the joint group of the robot 6 based on the gait data.
In this embodiment, as shown in FIG. 1, the walking direction of the robot 6 is the x axis, the body side direction is the y axis, and the height direction is the z axis.

  FIG. 2 is a block diagram showing the functions of the computer device 14. Among the elements shown in FIG. 2, a part excluding the robot mechanical system 216 and the trunk acceleration sensor 218 is a component of the computer device 14. The computer device 14 may be physically contained in one device as a whole, or may be accommodated separately for each physically separated device.

  The gait data storage device 210 stores gait data of the robot 6 calculated in advance. Gait data refers to arm tip trajectory data (arm tip gait data), foot tip trajectory data (foot tip gait data) and trunk trajectory data (trunk gait data). Of the gait data stored in advance, the arm tip trajectory data and the foot tip trajectory data are directly input to the joint angle group calculation device 212. The trunk gait data is input to the trunk gait data correction calculation unit 222, corrected each time the robot 6 walks, and the corrected data indicating the trunk trajectory is input to the joint angle group calculation device 212. The

  The joint angle group calculation device 212 solves so-called inverse kinematics based on the input gait data, thereby solving each joint angle θ1 (t), θ2 (t), θ3 (t),. Calculate The calculated joint angle group data is input to the actuator control unit 214.

  The actuator control unit 214 controls an actuator group that rotates the joint group of the robot 6 of FIG. The actuator group exists in the mechanical system 216 of the robot. By controlling the actuator group, the joint angle of the robot 6 can be adjusted to the calculated joint angle. The robot 6 walks according to the gait data.

  A trunk acceleration sensor 218 is mounted on the trunk 12 and detects accelerations of the trunk in the x direction (walking direction), the y direction (body side direction), and the z direction (height direction). The trunk acceleration calculation device 220 outputs the actual trunk accelerations axr, ayr, and azr detected by the trunk acceleration sensor 218 to the trunk gait data correction calculation unit 222.

  The trunk gait data correction calculation unit 222 includes trunk trajectory data (trunk gait data) stored in the gait data storage device 210 and the actual body of the robot calculated by the trunk acceleration calculation device 220. Based on the trunk acceleration, the trunk trajectory data is corrected. Details of the correction of the trunk trajectory data will be described later. The corrected trunk trajectory data is input to the joint angle group calculation device 212.

FIG. 3 is a block diagram showing details of the trunk gait data correction calculation unit 222. In FIG. 3, only events in the x direction are shown for clarity of illustration. The same applies to the y direction and the z direction. The toe trajectory data and the arm tip trajectory data are directly instructed to the joint angle group calculation device 212 without being corrected.
In the trunk gait data correction calculation unit 222, target trunk positions xo, yo, and zo (each of which changes with time) indicated by the trunk gait data stored in the gait data storage device 210 are converted into time. The target trunk accelerations axo, ayo, and azo are calculated by the device 304 that performs the second-order differential calculation with respect to.
The deviation calculating device 312 calculates acceleration deviations Δax, Δay, Δaz by subtracting the target trunk accelerations axo, ayo, azo from the actual trunk accelerations axr, ayr, azr.
The disturbance force calculation device 310 calculates the magnitudes of the disturbance forces Drx, Dry, Drz estimated to have acted on the robot by multiplying the calculated acceleration deviations Δax, Δay, Δaz by the mass M of the robot. If the robot joints are loose, the control is delayed, or an external force is applied from the outside world, the actual trunk acceleration will not become the target trunk acceleration, and the acceleration deviation Δax due to an unexpected disturbance force , Δay, Δaz are generated. From the acceleration deviations Δax, Δay, Δaz, the magnitudes of the disturbance forces Drx, Dry, Drz acting on the robot can be calculated.

The target correction amount calculation device 308 receives the calculated disturbance forces Drx, Dry, Drz, and calculates the target trunk position correction amounts xd, yd, zd based on the transfer function 1 / G (s). The transfer function 1 / G (s) is composed of a delay element in which the coefficient of the second-order term is Mi, the coefficient of the first-order term is Ci, and the coefficient of the 0th-order term is Ki. Mi, Ci, and Ki are desired parameters and are designated in advance. The processing of the target correction amount calculation device 308 is performed in the x direction, for example.
Drx = Mi · xd (2) + Ci · xd (1) + Ki · xd (0)
Is equivalent to finding xd by solving the differential equation. Here, (2) indicates the second derivative with respect to time, (1) indicates the first derivative with respect to time, and (0) indicates xd itself.
Depending on how to sign the correction amount xd,
−Drx = Mi · xd (2) + Ci · xd (1) + Ki · xd (0)
Sometimes it is more correct to say.

  The target trunk position correction device 306 adds the correction amounts xd, yd, zd to the target trunk positions xo, yo, zo before correction stored in the gait data storage device 210, and the corrected target The trunk positions xref, yref, and zref are calculated, and an instruction is given to the joint angle group calculation device 212.

The effect of the control law for correcting the target trunk trajectory data will be described.
FIG. 4 shows a dynamic model of the robot. A trunk when the trunk 12 is at the target trunk position is indicated by 402, and a trunk when the trunk 12 deviates from the target trunk position is indicated by 404.
Equation (1) in FIG. 4 shows a general equation of motion of the robot. F represents the total external force acting on the trunk, M represents the trunk mass, and a represents the trunk acceleration.
The trunk has an ideal trajectory (target trajectory) suitable for walking. The target trajectory is a trunk trajectory that realizes a ZMP that matches the target ZMP unless a disturbance force acts on the robot.
Equation (2) represents an equation of motion of the trunk 402 of the robot that moves along the target trunk trajectory. Fo represents the total external force acting on the trunk 402, and ao represents the target acceleration of the trunk 402.
Fo is a total external force acting on the trunk 402 when moving on the target trunk trajectory, and an internal force acting on the trunk from the leg link and a gravity acting on the trunk of the robot in an ideal walking state of the robot. It is the combined power of.
Since the actual trunk movement is affected by disturbance forces such as unpredictable road surface irregularities and robot structural deflection, it is difficult to control the movement according to the target trunk trajectory. The actual trunk movement is called a real trajectory.
Equation (3) represents an equation of motion of the trunk 404 of the robot that moves along the actual trajectory. Fr represents the total external force acting on the trunk 404, and ar represents the actual acceleration of the trunk 404.
Equation (4) represents an estimation formula for the disturbance force acting on the trunk of the robot. Dr represents a disturbance force expressing the influence of the disturbance force acting on the robot on the movement of the trunk by the external force acting on the center of gravity of the trunk. Fo is the total external force that acts on the trunk when moving on the target trunk trajectory, and the internal force that acts on the trunk from the leg link and the gravity that acts on the trunk of the robot while the robot is walking ideally. It is a power. Fr is the total external force acting on the trunk moving in the actual trajectory, the internal force acting on the trunk from the leg link in the ideal walking state of the robot, and the gravity and disturbance force acting on the trunk of the robot It is the combined power of. Therefore, the disturbance force Dr acting on the trunk can be estimated by taking the difference between Fr and Fo.
Equations (5) and (6) are equations for calculating the target trunk position correction amount xd used in the embodiment. The transfer function G (s) describes the “spring-mass point-damper system” shown in FIG.

FIG. 5 is a diagram schematically showing a relationship in which a force for returning to the ideal trajectory is applied to the trunk of the robot by correcting the target trunk position.
The target trunk position 510 before correction is an ideal trunk position that realizes a ZMP that matches the target ZMP.
Actually, the actual trunk position 512 is deviated from the target trunk position 510 because the disturbance force Dr is applied.
The target trunk position 506 after correction is a point separated by xd from the target trunk position 510 before correction.
When the target trunk position is corrected in (6), the corrected target trunk position 506 and the uncorrected target trunk position 510 are coupled via the spring 502 having the spring constant Ki and the damper 508 having the damping coefficient Ci, This is equivalent to a “spring-mass-damper system” in which the mass point 504 of mass Mi is present at the corrected target trunk position 506 and the disturbance force Dr is acting on the actual trunk position 512. The transfer function 1 / G (s) of the “spring-mass point-damper system” includes a delay element indicated by the coefficient Mi of the second order term, the coefficient Ci of the first order term, and the coefficient Ki of the zeroth order term. In addition, it acts as a shock absorber that exerts a return force to return to the target trunk position against a disturbance force that changes suddenly and suppresses a sudden position change.
The zeroth-order coefficient Ki derives a correction amount of the target trunk position that is proportional to the disturbance force Dr. The first-order coefficient Ci derives a correction amount of the target trunk position that is proportional to the first-order lag element of the disturbance force Dr. The second-order coefficient Mi derives a target trunk position correction amount that is proportional to the second-order lag element of the disturbance force Dr.
Among these, the coefficient Ki of the proportional element is indispensable, but one or both of the coefficient Ci of the primary delay element and the coefficient Mi of the secondary delay element can be omitted. It has been confirmed that even if one or both of the coefficient Ci of the first-order lag element and the coefficient Mi of the second-order lag element are zero, the robot continues to walk stably.

FIG. 6 schematically shows the effect of correcting the target trunk movement.
FIG. 6A shows a situation where a robot that does not correct the target trunk motion walks. When the disturbance force Dr acts on the trunk, the trunk position 602 of the robot follows the disturbance force Dr and deviates from the ideal trajectory. When the disturbance force changes abruptly, the actual trunk movement also changes abruptly, so that walking becomes unstable.
FIG. 6B shows a situation where the robot that corrects the target trunk motion walks. When the disturbance force Dr acts on the robot, the target trunk position is corrected by xd on the side opposite to the direction in which the disturbance force Dr acts. As a result, the same force is generated as when the mass point of the mass Mi is moved to the opposite side by the acceleration xd (2). The robot is coupled to the mass point of the mass Mi via the spring 502 having the spring constant Ki and the damper 508 having the damping coefficient Ci, and the same force as when the mass point of the mass Mi is separated by the distance xd acts. . This force compensates for the disturbance force Dr. The trunk position of the robot returns toward the target trunk position. When approaching the target trunk position before correction, the correction amount xd also decreases. The actual trunk position of the robot is returned to the target trunk position before correction by correcting the target trunk position.
From the viewpoint of the dynamic characteristics of the robot, the disturbance force Dr is compensated via the delay element 610, which corresponds to maintaining the robot in the vicinity of the target position 606 using a spring and a shock absorber.
Obviously from the equivalent dynamic system, according to the control technique of the present embodiment, the abrupt fluctuation is removed from the control for compensating for the disturbance force, and it can be handled by an actuator having a limited capacity. The robot can follow the target trunk trajectory well without rapidly changing. When actually tested, it can be confirmed that the trunk of the robot walks stably. The observer can observe the robot walking with peace of mind.

FIG. 7 shows a processing procedure of the walking control method of the robot 6.
In step S <b> 2, gait data describing temporal changes in target foot tip position, arm tip position and trunk position is stored in the gait data storage device 210.
In step S4, the robot starts walking. When walking is started, the robot 6 repeatedly executes step S8 and subsequent steps. That is, the processing from step S8 to step S16 is repeatedly executed every predetermined time (Δt).
In step S8, based on the difference between the target trunk exercise and the actual trunk exercise, data that describes the target trunk exercise (trunk gait data) so that the robot's walking is stabilized. to correct. Details of the processing performed in step S8 will be described later.
In step S12, the toe trajectory data and arm tip trajectory data stored in the gait data storage device 210 and the trunk trajectory data corrected in step S8 are input to the joint angle group calculation device 214.
In step S14, the joint angle group calculation device 212 calculates the joint angle of each joint of the robot 6 based on the input gait data. The calculated joint angle group data is instructed to the actuator control unit 214 of the robot. The actuator control unit 214 controls an actuator group that rotates each joint of the robot 6. Thereby, the joint angle of the robot 6 is adjusted to the calculated value.
As already described, the processing from step S8 to step S16 is repeatedly executed every predetermined time (time interval Δt) (step S6).

FIG. 8 shows in detail the processing procedure of the trunk gait data correction calculation in step S8.
In step S22, a target trunk acceleration is calculated from the trunk gait data stored in the gait data storage device 210.
In step S24, the detection value of the trunk acceleration sensor 218 is captured.
In step S <b> 26, the trunk acceleration calculation device 220 calculates the actual trunk acceleration from the detection value of the trunk acceleration sensor 218.
In step S 28, the deviation between the target trunk acceleration and the actual trunk acceleration is obtained by the deviation calculation device 312.
In step S30, the disturbance force calculation device 310 obtains the magnitude of the disturbance force Dr that caused the acceleration deviation from the calculated acceleration deviation.
In step S32, the target correction amount calculation device 308 calculates the trunk position correction amount based on the magnitude of the disturbance force Dr and the transfer function 1 / G (s).
In step S34, the correction device 306 adds the trunk position correction amount to the target trunk position before correction stored in the gait data storage device 210, calculates the corrected target trunk position, and calculates the joint angle. Instructs the group calculation device 212.

A part or all of the computer device 14 can be mounted outside the robot. For example, the gait data storage device 210 and the trunk gait data correction calculation unit 222 can be placed outside the robot. In this case, the actual trunk acceleration detected by the robot is sent to the trunk gait data correction calculation unit 222 by wireless or wired signal, and the corrected gait data is sent to the robot by wireless or wired signal. The joint angle group calculation device 212 may also be mounted outside the robot. The calculated joint angle data is transmitted to the actuator control unit 214 by a wired or wireless signal.
The walking robot according to the present embodiment stabilizes the trunk against a disturbance force and keeps walking even if there is no force sensor at the tip of the foot.
In the above description, the inclination angle of the trunk changes according to a planned change pattern. The trunk gait data includes only data relating to the trunk position at which the actual ZMP matches the target ZMP when the inclination angle of the trunk changes according to a planned change pattern. On the other hand, the “trunk position and trunk inclination angle” that matches the actual ZMP with the target ZMP may be calculated and used as trunk gait data.
The trunk inclination angle may be a planned change pattern or a calculated change pattern, but in any case, data indicating the change over time of the target trunk inclination angle is given. It has been. The trunk inclination angle here refers to an inclination angle around the x axis and an inclination angle around the y axis.
The technique of the present invention is effective not only for the trunk position but also for the trunk angle. Calculate the deviation between the second-order differential value (angular acceleration) of the target trunk angle and the angular acceleration of the actual trunk angle, the angle proportional to the deviation, the angle proportional to the first derivative of the deviation, and the secondary of the deviation If the angle proportional to the differentiation is calculated and the correction amount of the target trunk angle is calculated therefrom, it is possible to compensate for the shift of the trunk angle from the target trunk angle due to the influence of the disturbance force. The present invention can be generally applied to exercises relating to the position and posture of the trunk, such as the trunk position and trunk inclination angle.

(Example 2)
The robot may be equipped with a device that detects the trunk position. For example, if a gyro is mounted, not only actual trunk acceleration but also actual trunk position can be detected. Even if a GPS measurement sensor is installed, the actual trunk position can be detected. The trunk speed can be obtained by integrating the trunk acceleration, and the trunk position can be detected by further integrating the trunk speed.
If the robot can detect the actual trunk position, a feedback control loop can be utilized. By utilizing the feedback control loop, it is possible to cope with the case where the robot tilts as a whole because the robot has walked on rough terrain. Hereinafter, the second embodiment will be described, but only points different from the first embodiment will be described.
FIG. 9 is a block diagram showing the functions of the computer device of the robot according to the second embodiment. Both the actual trunk acceleration and the actual trunk position are input to the trunk gait data correction calculation unit 224. The
FIG. 10 is a block diagram showing details of the trunk gait data correction calculation unit 224. In FIG. 3, only events in the x direction are shown for clarity of illustration. The same applies to the y direction and the z direction.
Since the trunk position sensor 904 is mounted on the robot, the actual trunk position is measured. The trunk acceleration calculation device 220 second-order differentiates the actual trunk position with respect to time to obtain the actual trunk acceleration. Based on the obtained trunk acceleration, the same processing as in the first embodiment is executed, and a correction amount xd based on the acceleration deviation is calculated.
The corrected target trunk position xo + xd obtained by correcting the uncorrected target trunk position xo with the correction amount xd based on the acceleration deviation is input to the deviation calculation device 906, and the deviation xo + xd−xr from the actual trunk position xr. Is calculated. The trunk position deviation xo + xd-xr is input to the feedback processing block 908 and processed using the transfer function C (s). A value C (s) (xo + xd-xr) obtained by processing the deviation xo + xd-xr by using the transfer function C (s) is added to the target trunk position xo + xd corrected with the correction amount based on the acceleration deviation, whereby the target Further correct the trunk position. The finally corrected target trunk position is corrected to xo + xd + C (s) (xo + xd−xr).
FIG. 11 shows a processing procedure of trunk gait data correction calculation of the second embodiment.
In step S54, the detection value of the trunk position sensor 904 is captured.
In step S56, the actual trunk position and the actual trunk acceleration are detected from the detection value of the trunk position sensor 904.
In step S58, a deviation xo + xd−xr between the target trunk position xo + xd obtained by correcting the target trunk position xo before correction with the correction amount xd based on the acceleration deviation and the actual trunk position xr is calculated.
In step S60, the transfer function C (s) is processed to calculate a feedback processing amount C (s) (xo + xd-xr).
In step S62, the target trunk position xref instructed to the joint angle group calculation device is corrected to xo + xd + C (s) (xo + xd−xr).

The legged robot shown in the second embodiment calculates the deviation xo + xd-xr and performs feedback control so that the deviation becomes smaller. Therefore, even when the robot walks on rough terrain and the robot as a whole tilts, Feedback control can be performed toward the trunk position.
The trunk position sensor may be a gyro that measures the inclination angle of the trunk. When the actual posture angle of the trunk is determined by the gyro, the trunk position relative to the center of the foot of the ground leg can be measured by multiplying the distance R from the foot center of the ground leg to the center of gravity of the trunk.

As mentioned above, although embodiment of this invention was described in detail, these are only illustrations and do not limit a claim. For example, in the gait data, an example was described in which changes over time were described in terms of time and position. However, time and speed, or the relationship between time and acceleration, time-dependent movements of feet, arms, and trunk Changes may be described. The technology described in the claims includes various modifications and changes of the specific examples illustrated above.
The technical elements described in this specification or the drawings exhibit technical usefulness alone or in various combinations, and are not limited to the combinations described in the claims at the time of filing. In addition, the technology exemplified in the present specification or the drawings achieves a plurality of objects at the same time, and has technical usefulness by achieving one of them.

The figure which shows the outline of the walk of a legged mobile robot. FIG. 3 is a block diagram illustrating functions of the computer apparatus according to the first embodiment. The figure which blocks and shows the function of the trunk gait data correction | amendment calculating part of Example 1. FIG. The figure which shows the dynamic model of a trunk. The figure which shows typically the restoring force to the ideal track | orbit which acts on a trunk. The figure which shows typically the effect which correction | amendment of target trunk exercise | movement has on the trunk exercise | movement. The figure which shows the process sequence of the walking control method of the robot of Example 1. FIG. The figure which shows the process sequence of the correction | amendment calculation of trunk gait data of Example 1. FIG. FIG. 6 is a block diagram illustrating functions of a computer device according to a second embodiment. The figure which shows the function of the trunk gait data correction | amendment calculating part of Example 2 as a block. The figure which shows the process sequence of correction | amendment calculation of trunk gait data of Example 2. FIG.

Explanation of symbols

6 ... legged robots 7, 8, 9, 10 ... foot trajectory 12 ... trunk 14 ... computer device 210 ... gait data storage device 212 ... joint angle group calculation device 214 ... Actuator control unit 216 ... Robot mechanical system 218 ... Trunk acceleration sensor 220 ... Trunk acceleration calculation device 222 ... Trunk gait data correction calculation unit 304 ... Second order differentiation Arithmetic devices 306, 907 ... Correcting device 308 ... Target correction amount calculating device 310 ... Disturbance force calculating device 312, 906 ... Deviation calculating device 402 ... Body of the robot that moves along the target trajectory Trunk 404: Robot trunk moving along a real trajectory 502 ... Spring 504 ... Mass point 506, 604 ... Corrected trunk position 508 ... Damper 510, 606 ... Ideal Core position 12 ... Actual trunk position 602 ... Trunk position 608 when no correction is performed ... Disturbance force action point 610 ... Delay element 904 ... Trunk position measurement sensor 908 ... Feedback Processing block

Claims (7)

A trunk link and a leg link that is swingably coupled to the trunk;
Storage means for foot gait data that describes changes over time in target foot movement;
A means for storing trunk gait data that describes changes over time in a target trunk movement that enables walking while following a change in a target foot tip movement;
Trunk motion detection means for detecting actual trunk motion;
Deviation calculation means for calculating the deviation between the target trunk exercise and the actual trunk exercise,
Correction means for obtaining a correction amount based on a predetermined transfer function from the calculated deviation, and correcting trunk gait data stored in the trunk gait data storage means based on the correction amount;
Legged robot equipped with.   The legged robot according to claim 1, wherein the deviation calculating means calculates a deviation between a target trunk acceleration and an actual trunk acceleration.   3. The legged robot according to claim 2, wherein the correction means obtains a correction amount based on a disturbance force calculated from the acceleration deviation and a predetermined transfer function.   The legged robot according to claim 3, wherein the transfer function includes a proportional element.   The legged robot according to claim 4, wherein the transfer function includes a first-order lag element and / or a second-order lag element.   The correction means further corrects the target trunk position by adding an amount of feedback processing to the deviation between the target trunk position corrected based on the acceleration deviation and the actual trunk position. Item 6. The legged robot according to any one of items 3 to 5. Storing foot gait data describing a change over time of the desired foot toe movement;
Storing trunk gait data describing changes over time in the target trunk movement that enables walking while following the change in the target foot movement;
Detecting actual trunk movements;
Calculating a deviation between the target trunk movement and the actual trunk movement;
A correction amount is obtained from the calculated deviation based on a predetermined transfer function, the trunk gait data stored based on the correction amount is corrected, and the corrected trunk gait data is used as a joint angle group of the robot. Instructing the computing device;
Control method for legged robots equipped with

Images